At&T Corp. & Includible Subsidiaries v. Department of Revenue

*693 BALDWIN, J. Appellant AT&T (together with its includible subsidiaries) appeals a Tax Court judgment that denied AT&T’s claim for a refund of a portion of the Oregon corporate excise taxes it paid for tax years 1996 through 1999. AT&T Corp. v. Dept. of Rev., 20 OTR 299 (2011) (Tax Court’s opinion). The dispute concerns AT&T’s sale of interstate and international phone and data transmissions. We must decide whether those sales are counted in determining the fraction of AT&T’s income that Oregon can tax. Under the relevant statute, ORS 314.665(4) (1999), those sales count as Oregon sales if the “income-producing activity” was entirely performed in Oregon or if Oregon was the state with the greatest share of the “costs of performance” for that activity. Based on its interpretation of what constituted an “income-producing activity,” AT&T presented a cost study that purported to show that Oregon did not have the greatest share of the “costs of performance.” The Department of Revenue (department) challenged AT&T’s interpretation of “income-producing activity” and attacked the validity of its cost study. The Tax Court ruled in favor of the department. As we will explain, we conclude that AT&T did not use a correct definition of “income-producing activity.” AT&T’s proposed interpretation is network-based; it focused on the operation of its network as a whole. The correct understanding, however, is transaction-based; it examines individual sales to customers. AT&T thus failed to meet its burden of proof, because it did not correctly calculate the “costs of performance” for the correct “income-producing activities.” Accordingly, we affirm. I. BACKGROUND A. The Taxpayer AT&T, the taxpayer, is a global telecommunications company. It provides voice and data telecommunications services over the global AT&T network, which AT&T constructed, maintains, and operates. The network works together as an integrated whole, and it is used to provide all of AT&T’s services at issue. In the United States, the network provides multiple pathways that a call may take, *694 depending on traffic on the overall AT&T network. The path a particular call may take is not known in advance and is determined by a complex computer system that chooses the path in milliseconds. The network is managed from the AT&T Global Network Operations Center in Bedminster, New Jersey. AT&T does not generally provide “last-mile service” to its customers over its own facilities. Instead, it provides service on its own facilities only from one “point of presence” to another “point of presence.” The last-mile service from the point of presence to the caller or the recipient is provided over the facilities of a local exchange carrier. AT&T pays the local exchange carriers an access charge for this last-mile service. B. The Statute and Rule This case involves income tax. For those taxpayers that do business in more than one jurisdiction, states generally may tax a representative fraction of the taxpayer’s total income. The fraction of income that a particular state can tax is determined, in part, by the amount of a taxpayer’s sales in that particular state. Specifically, that part of the calculation requires comparing “in state” sales to the taxpayer’s total sales. In this case, we must determine whether sales are considered to be “in this state” under ORS 314.665(4) (1999). 1 For purposes of income tax, that statute defines certain sales as Oregon sales depending on the location of the “income-producing activity.” The sales are in Oregon if (1) the income-producing activity occurs entirely in Oregon, or (2) Oregon’s share of the “costs of performance” of that income-producing activity is greater than any other state. Specifically, ORS 314.665(4) provides: “Sales, other than sales of tangible personal property, are in this state if (a) the income-producing activity *695 is performed in this state; or (b) the income-producing activity is performed both in and outside this state and a greater proportion of the income-producing activity is performed in this state than in any other state, based on costs of performance.” The correct application of that statute depends in large part on the meaning of two terms: “income-producing activity” and “costs of performance.” The department has promulgated a rule that defines those terms, among other things. Briefly, the definition of “income-producing activity” is (in part): “The term ‘income-producing activity’ applies to each separate item of income and means the transactions and activity directly engaged in by the taxpayer in the regular course of its trade or business for the ultimate purpose of obtaining gains or profit.” OAR 150-314.665(4) (2). The second term, “costs of performance,” is defined in part as follows: “The term ‘costs of performance’ means direct costs determined in a manner consistent with generally accepted accounting principles and in accordance with accepted conditions or practices in the trade or business of the taxpayer.” OAR 150-314.665(4) (4). C. Procedural Posture 1. AT&T’s Request for Refund This action began when AT&T sought a refund on the taxes that it paid for tax years 1996 through 1999. AT&T filed amended tax returns seeking a refund. AT&T argued that the relevant statute and rule, correctly understood, indicated that its “income-producing activities” were some of the business activities associated with its network operations, because the network was necessary for AT&T to provide interstate and international transmissions of voice and data for consumers and businesses. 2 *696 Based on its interpretation of the law, AT&T offered a cost study of the “costs of performance” for what it considered to be its “income-producing activities.” The cost study was prepared for AT&T by BI Solutions Group, LLC — more specifically, by James Allen, a manager at BI Solutions and a specialist in cost accounting, who testified as an expert witness. The printed version of the study, including the associated schedules, is almost 300 pages long; our discussion here gives only a brief overview, often simplifying details that are not relevant to our holding. Because AT&T’s understanding of “income-producing activity” indicated that its activities were performed both in-state and out-of-state, the cost study analyzed the “costs of performance” for each activity. It then compared the costs incurred in Oregon to those incurred in New Jersey. 3 The study purported to show that the costs incurred in New Jersey exceeded the costs incurred in Oregon in every instance. Accordingly, AT&T argued that the requirement of ORS 314.665(4) for including the relevant sales as Oregon sales had not been met: Oregon did not have a “greater proportion” of the costs of performance than any other state. As a result, AT&T’s position is that none of its relevant revenues from sales — not even those revenues from Oregon customers — count as being “in” Oregon. The study consists of two parts. In the first part, the cost study analyzed AT&T’s costs based on its interpretation of the statute and rule. The second part of the study, anticipating the department’s position, analyzed AT&T’s costs incurred for “Oregon demand.” The first part of the cost study began by identifying what it considered to be the “separate item[s] of income,” a term used in the rule defining “income-producing activity.” The cost study concluded that the items of income were “the Network Telecommunications Services” that AT&T provided to its customers. Because AT&T provided many such *697 services, however, the cost study categorized them into “service families.” The service families used in the cost study were consumer voice, business voice, and business data, each of which was divided into interstate and international services. The cost study then identified what AT&T considered to be the relevant income-producing activities, which were certain of AT&T’s business activities. Fundamentally, these activities were based around AT&T’s operation of the network as a whole. Of those, the cost study focused on those business activities whose costs would, in AT&T’s estimation, qualify as “direct costs” for the “costs of performance” part of the analysis. The three applicable income-producing activities identified by AT&T were: “service activation” (which Allen summarized as “setting up the customers and having them able to utilize AT&T’s network telecommunication services”), “service assurance” (meaning “work activity associated with maintaining the availability of the network”), and “service execution” (meaning “the physical network,” “the assets by which AT&T provides its network telecommunications services”). In other words, AT&T’s analysis of its income-producing activities was system-based or network-based. 4 Based on those classifications, AT&T’s cost study generally analyzed the “costs of performance” for each *698 “income-producing activity” in the following manner: The study first identified all of AT&T’s costs that it believed qualified as “direct costs” under the rule, and specifically identified those costs incurred in Oregon and New Jersey. In particular, the study excluded the access charges, which it concluded did not qualify as a “direct cost.” The cost study then attributed the direct costs to the “income-producing activities” that it had concluded were recognizable under the statute and rule. Finally, the study attributed those costs to the service families that it had identified as “separate item[s] of income.” The result of those calculations purported to identify the direct costs for each income-producing activity as to each item of income. The summary calculations, however, were only by year, by service family (that is, AT&T’s classification of “item of income”), and by costs in Oregon and New Jersey. Here, for example, are the cost study’s summary calculations for 1999: Oregon Costs New Jersey Costs Business Voice Interstate $ 15,770,292 260,731,804 Int’l $ 911,698 $ 18,963,486 Data Interstate 15,715,660 $ 198,100,125 Int’l 437,258 $ 5,511,749 Consumer Voice Interstate $ 3,698,467 $ 70,274,114 Int’l $ 1,113,686 $ 20,778,559 Because the cost study showed similar results for the other tax years, AT&T argued that it had demonstrated that the New Jersey “costs of performance” exceeded Oregon costs. Accordingly, it maintained, the receipts from interstate and international voice and data should not be included in the numerator of the Oregon sales factor. *699 As we noted, the cost study contained two parts. In the second part of its cost study, the study analyzed the costs associated with “Oregon demand,” to address the department’s anticipated legal analysis. 5 This second part of the cost study functionally paralleled the first part. Again, in general terms, the study took the figures for direct costs that it had attributed to the “income-producing activities” for each item of income, and then calculated (1) the share of Oregon costs that had been incurred to support Oregon demand, and (2) the share of New Jersey costs that had been incurred to support Oregon demand. AT&T’s cost study again purported to show that, even under an “Oregon demand” analysis, the “costs of performance” in New Jersey were greater than they were in Oregon. Here, for example, are its calculations for 1999: OR Cost for OR NJ Costs for OR Demand Demand Business Voice Interstate 634,702 $ 2,638,468 Int’l 55,649 $ 190,560 Data Interstate $ 391,277 $ 1,994,880 Int’l $ 10,887 $ 55,504 Consumer Voice Interstate $ 318,818 709,725 Int’l $ 88,889 210,008 If AT&T’s analysis was both legally and factually sound, then the relevant sales would not count as Oregon *700 sales, and Oregon would not be entitled to tax as much of AT&T’s income. For example, AT&T’s original 1996 tax return indicated that Oregon could tax 0.7621 percent of AT&T’s total income of $7.3 billion, while AT&T’s amended 1996 tax return suggested that Oregon could tax only 0.4303 percent of that $7.3 billion. Similar differences would have applied to the other tax years. The department opposed the claim for a refund. The department argued for a transaction-based application of “income-producing activity”; namely, that the relevant “income-producing activity” here was the activity that produced each individual interstate and international phone and data transmission billed to an Oregon customer. As the department stated in its pretrial memorandum to the Tax Court: “The proper construction of‘income producing activity’ * * * applies to each item of income from each call billed on a per minute basis or each flat rate subscription. Thus, the income producing activity is AT&T’s connection of each Oregon long distance call to the network.” The department maintained that the only “costs of performance” for those individual transactions — meaning “direct costs,” OAR 150-314.665(4) (4) — were (1) the cost of the electricity for the calls and data transmission, and (2) the access charges levied by the local exchange carrier for connecting the customer with AT&T’s network. The majority of those costs were incurred in Oregon, the department argued, and so the sales should be counted as Oregon sales. The result would be that AT&T’s tax liability to Oregon would remain unchanged from its initial tax returns. In addition, the department offered expert testimony critical of AT&T’s cost study. Michael Starkey, a consultant who specialized in telecommunications cost analysis, testified and prepared a rebuttal report. He asserted that the department’s transaction-based interpretation of “income-producing activity” was correct, and he disagreed with AT&T’s network-based interpretation. The Tax Court agreed with the department and denied AT&T’s request for a refund. The court first determined that the statutes and regulations required a three-step analysis: *701 “[T]he analysis must begin with transactions that are or include ‘income producing activities].’ The next step is to determine the gross receipt from that transaction. The final step is to determine where the direct costs of performance occurred geographically, as to the transaction or activity.” 20 OTR at 304-05 (second alteration in original; footnote omitted). At the first step that it had identified, the court concluded that the statutes and regulations required a focus on individual transactions with customers, while AT&T instead had “focuse[d] on entire groups or classes of transactions.” Id. at 306. AT&T’s analysis was deficient, the court explained, because it “ha[d] begun its analysis with the wrong cost object [and] ha[d] not attempted to determine where costs of particular transactions are incurred.” Id. at 307. 6 Although the court concluded that that was enough to deny AT&T’s refund claim, it went on to consider other questions. Id. Because the court had agreed with the department that each separate phone call or flat-rate billing was an income-producing activity, the court rejected AT&T’s argument that the costs of performance — by rule, the direct costs — were those associated with operating the network generally. Id. at 307-09. The court apparently agreed with the department that the direct costs were only those costs that AT&T incurred in carrying a particular individual phone call — basically the electricity consumed by the call and the access charges that AT&T paid to the local exchange carrier on each end of the transaction. See id. at 309 (concluding that the department’s interpretation “has a number of points in its favor”). The Tax Court specifically rejected AT&T’s argument that the access charges paid to local exchange carriers did not count as direct costs for purposes of the “costs of performance” analysis. Id. at 309-11. *702 II. ANALYSIS A. Standard of Review and Interpretive Methodology AT&T has appealed the decision of the Tax Court. Our review is “limited to errors or questions of law or lack of substantial evidence in the record to support the tax court’s decision or order.” ORS 305.445. The appeal involves AT&T’s request for a refund, and as the party seeking affirmative relief, AT&T bears the burden of proof. See ORS 305.427 (in Tax Court proceedings “and upon appeal therefrom,” “[t]he burden of proof shall fall upon the party seeking affirmative relief’). To address the issues before us, we must interpret Oregon statutes and administrative rules. In both instances, we apply the same methodology: We first consider the text and context, and then (in the case of statutes) examine any relevant legislative history. See State v. Gaines, 346 Or 160, 171-72, 206 P3d 1042 (2009) (explaining that methodology for statutes); Wetherell v. Douglas County, 342 Or 666, 678, 160 P3d 614 (2007) (court applies its statutory interpretation methodology to administrative rules). In the case of administrative rules, we ordinarily defer to the adopting agency’s interpretation if that interpretation is plausible and is not inconsistent with the rule in its context or with some other source of law. See Crystal Communications, Inc. v. Dept. of Rev., 353 Or 300, 311, 297 P3d 1256 (2013) (stating principle); Don't Waste Oregon Com. v. Energy Facility Siting, 320 Or 132, 142, 881 P2d 119 (1994) (same). B. Uniform Division of Income for Tax Purposes Act The statute at issue here, ORS 314.665(4), is a part of Oregon’s version of the Uniform Division of Income for Tax Purposes Act (UDITPA). In this state, UDITPA is codified at ORS 314.605 to 314.675. Strictly speaking, Oregon’s UDITPA statutes apply here only by administrative rule. AT&T is a public utility, see ORS 314.610(6), 7 and so it is taxed under ORS 314.280(1). 8 *703 The latter statute generally authorizes the department to determine whether and when a taxpayer may use what it terms the “the segregated method of reporting or the apportionment method of reporting.” 9 The department promulgated rules under ORS 314.280 that, for purposes of the apportionment method, have largely incorporated Oregon’s version of UDITPA. Crystal Communications, 353 Or at 303-04 (so noting); see OAR 150-314.280-(A) - (F) (incorporating various UDITPA statutes and administrative rules). AT&T here used the apportionment method of reporting, and the department does not question the propriety of AT&T using that method. Accordingly, we turn to UDITPA. A brief overview of UDITPA may provide some helpful background. UDITPA was intended to address the problem of how to attribute income to taxpayers who do business in more than one state. It is often difficult or impossible for a taxpayer doing business in more than one state to attribute a particular dollar of income to a particular state. See, e.g., Walter Hellerstein, State Taxation of Corporate Income from Intangibles: Allied-Signal and Beyond, 48 Tax L Rev 739, 745 (1993). The types of businesses that cannot attribute dollars to particular states are generally known as unitary businesses; they are “businesses in which a portion of the business done within the state is dependent upon or contributes to the operation of the business without the state.” Crystal Communcations, 353 Or at 303 (internal quotation marks and citations omitted); see Hellerstein, 48 Tax L Rev *704 at 745-46. 10 To tax unitary businesses, states use what is known as formulary apportionment. Rather than attempt to attribute individual dollars of income to particular states, formulary apportionment uses a formula to estimate the fraction of the taxpayer’s total income that can be attributed to each state. See Hellerstein, 48 Tax L Rev at 745; Walter Hellerstein, State Income Taxation of Multijurisdictional Corporations: Reflections on Mobil, Exxon, and H.R. 5076, 79 Mich L Rev 113, 117 (1980); Frank M. Keesling and John S. Warren, The Unitary Concept in the Allocation of Income, 12 Hastings L J 42, 43 (1960). 11 UDITPA was intended to provide a uniform method for the states to determine the specific fraction to be used for the formulary apportionment. See, e.g., William J. Pierce, The Uniform Division of Income for State Tax Purposes, 35 Taxes 747, 748 (1957) (noting the “amazing variety of formulas” then being used by the states, creating the danger that a taxpayer would be either double taxed or not taxed on everything). If UDITPA were adopted by every state, then its common formula would mean that all of a taxpayer’s income would be taxed in some state, but none of it would be taxed by more than one state. Pierce, 35 Taxes at 748 (so noting). The apportionment formula used by UDITPA utilizes certain factors to estimate the extent to which a taxpayer’s income may be attributed to a particular state. See, e.g., Steve Christensen, Formulary Apportionment: More Simple — On Balance Better? 28 L & Pol’y in Int’l Bus 1133, 1147 (1997). The three factors used in UDITPA *705 apportionment are the property factor, the payroll factor, and the sales factor. Each factor is itself a fraction: • the property factor is the ratio of in-state property to total property, • the payroll factor is the ratio of in-state payroll to total payroll, and • the sales factor is the ratio of in-state sales to total sales. See UDITPA §§ 10, 13, 15 (defining each factor); ORS 314.655 - 314.665 (Oregon’s version of those factors). Under UDITPA, the fractions representing the three factors are added together and divided by three. UDITPA § 9; compare ORS 314.650(1) (during the relevant tax years, Oregon doubled the sales factor before adding, and then divided by four). The result is the fraction of the taxpayer’s total income that can be taxed by this state. The first two factors, the property and payroll factors, estimate the state’s share of responsibility for the income stream by focusing on production. Those two factors track in-state capital and labor respectively, so they reflect obvious connections with the taxing state. See, e.g., Hellerstein, 48 Tax L Rev at 754. The third factor, sales, generally tracks the extent to which the taxpayer takes advantage of the taxing state’s market. See Powerex Corp. v. Dept. of Rev., 357 Or 40, 64-65, 346 P3d 476 (2015); John A. Swain, Reforming the State Corporate Income Tax: A Market State Approach to the Sourcing of Service Receipts, 83 Tul L Rev 285, 288 (2008); Christensen, 28 L & Pol’y in Int’l Bus at 1134. C. UDITPA’s Sales Factor As noted, this case involves the sales factor. We must begin by defining sales. Generally, “sales” are “gross receipts of the taxpayer.” UDITPA § 1(g); ORS 314.610(7) (both excluding gross receipts that are not allocated under other statutes). 12 *706 The sales factor generally is the taxpayer’s in-state sales divided by the taxpayer’s total sales. UDITPA section 15 and ORS 314.665(1), which are identical, specifically state: “The sales factor is a fraction, the numerator of which is the total sales of the taxpayer in this state during the tax period, and the denominator of which is the total sales of the taxpayer everywhere during the tax period.” See also Powerex, 357 Or at 42, 60. The denominator of the sales factor begins and ends by counting the gross receipts from all sales by the taxpayer everywhere. The numerator, however, is more complex. As we will explain, the relevant statute begins at the same point as the denominator: with all sales by the taxpayer everywhere. It then sets out an analytical method to determine which of those sales will be counted as sales “in this state.” This analytical method divides all of taxpayer’s sales into two classes: sales of tangible personal property, and all other sales, i.e., “sales, other than sales of tangible personal property.” See ORS 314.665(2), (4); Powerex, 357 Or at 43 n 3, 60-61. For each class, the statute then prescribes which of those sales count as in-state sales. If the sales are of tangible personal property, then one set of criteria are used to determine which of those sales count as sales “in this state.” ORS 314.665(2); see Powerex, 357 Or at 43. If the sales are of anything other than tangible personal property, then another set of criteria are used to determine which of those sales count as sales “in this state.” ORS 314.665(4); see Powerex, 357 Or at 43. If all sales in both classes meet the relevant criteria so that they are all considered to be “in this state,” then the numerator of the sales factor will be identical to the denominator, and the sales factor will be 1.0 (that is, 100 percent). 1. Sales of Tangible Personal Property Sales of tangible personal property are controlled by ORS 314.665(2) (UDITPA section 16). Although that provision is not at issue here, it serves as useful context. The Oregon statutory text provides: “(2) Sales of tangible personal property are in this state if: *707 “(a) The property is delivered or shipped to a purchaser, other than the United States Government, within this state regardless of the f.o.b. point or other conditions of the sale; or “(b) The property is shipped from an office, store, warehouse, factory, or other place of storage in this state and (A) the purchaser is the United States Government or (B) the taxpayer is not taxable in the state of the purchaser.” ORS 314.665(2); see also UDITPA § 16 (essentially identical). Thus, ORS 314.665(2) sets out the criteria to be applied to all of a taxpayer’s sales of tangible personal property to determine which (if any) count as sales “in this state.” The statute focuses on individual sales to purchasers. The general rule counts an individual sale as “in this state” if this state is where the property is delivered or shipped to the purchaser. See Powerex, 357 Or at 46-52 (analyzing statute in detail). As a result, this part of the sales factor generally reflects that state’s contribution as a market toward the taxpayer’s income. The statute does not always reflect the market state’s contribution, however. Sales of tangible personal property are not attributed to the market state in two instances: when the United States Government is a purchaser, and when the taxpayer cannot be taxed in the purchaser’s state. The latter exception is particularly significant, because it shows that UDITPA emphasized the interest of making all of the taxpayer’s income taxable somewhere over the interest of reflecting the market state’s contribution to the transaction. See Jerome R. Hellerstein, Walter Hellerstein, and John A. Swain, 1 State Taxation ¶ 9.18 [1] [b] [i], 9-259 (3d ed 2014) (noting policy justification to avoid possibility of “‘nowhere’ income”). In short, sales of tangible personal property are generally — but not invariably — apportioned to the market state. 2. Sales Other Than Sales of Tangible Personal Property The second sales factor provision of UDITPA, and the one at issue here, is the catchall provision. While the first factor deals with sales of tangible personal property, the second *708 factor deals with all other sales — “sales, other than sales of tangible personal property.” Again, ORS 314.665(4) provides: “(4) Sales, other than sales of tangible personal property, are in this state if (a) the income-producing activity is performed in this state; or (b) the income-producing activity is performed both in and outside this state and a greater proportion of the income-producing activity is performed in this state than in any other state, based on costs of performance.” See also UDITPA § 17 (essentially identical). The framers of UDITPA did not provide any commentary for UDITPA section 17. In any particular case, how ORS 314.665(4) will apply depends largely on the precise meaning of two terms: “income-producing activity” and “costs of performance.” Here, the department urges that “income-producing activity” is transaction focused and applies to each individual phone call or separate monthly billing. AT&T, on the other hand, contends that “income-producing activity” refers to broad swaths of its business. Its system-based or network-based interpretation, if accepted, would cause the statute to assign large chunks of revenue to a single state. We make two observations about the text of ORS 314.665(4). First, AT&T correctly notes that the provision does not look to the market where the sales occur. There is nothing specified about the geographic location of the taxpayer’s customers, which one would expect from a factor focused on a state’s contribution to the market. Instead, the provision looks to where the taxpayer effectively produces the income. The state where the taxpayer conducts its “income-producing activity” for a sale or class of sales may or may not happen to be the market state. Here, just as we noted in connection with certain sales of tangible personal property, the drafters of UDITPA apparently chose to run the risk that those sorts of sales would not reflect the market state’s contribution. The second point, by contrast, is that ORS 314.665(4) seems to connect the term “income-producing activity” with particular “sales.” The statutory purpose is to assign the income from sales to particular states, and to do that, the *709 provision directs us to identify the activity that produces that income — the income from that particular sale. The statute thus suggests that the focus may be on individual sales. Such a reading would parallel the treatment of sales of tangible personal property under ORS 314.665(2): Just as that statute attributes each individual sale of tangible personal property to a particular state, so ORS 314.665(4) arguably attributes other individual sales to a particular state. That would imply, then, that the income-producing activity in ORS 314.665(4) means the activity that produces the income associated with a particular sale. In saying that, we do not suggest that that is the only possible way to read the provision. Commentators have routinely criticized UDITPA section 17 for its ambiguity: “[T]he commentators have found the rules to be ‘confusing and indefinite’ and plagued by ‘vagueness,’ ‘ambiguity,’ ‘substantial debate,’ ‘lack of clear guidance,’ ‘whipsaw [ing],’ ‘tremendous flexibility, and hence [tax planning] opportunity,’ ‘frequent litigation,’ ‘inconsistency,’ and ‘confusion for taxpayers and taxing authorities alike.’” Swain, 83 Tul L Rev at 306 (second and third alterations in original; footnotes omitted). While there is room for doubt at the statutory level, then, it appears to us that the department’s interpretation of the statute is more likely to be within the range of permissible interpretations. D. OAR 150-314.665(4) As noted, the department adopted an administrative rule, OAR 150-314.665(4) (2007), that defines “income-producing activity” and “costs of performance” and that otherwise attempts to clarify how ORS 314.665(4) should be applied. That rule is almost identical to a model regulation that had been previously promulgated by the Multistate Tax Commission. See Multistate Tax Commission Allocation and Apportionment Regulation IV.17 (1997), available at http://bit.ly/lgbN3zv (accessed July 1, 2015) (essentially identical to OAR 150-314.665(4)). 13 Neither party asserts *710 that the department’s rule is invalid or inconsistent with ORS 314.665(4). Accordingly, we now turn to that rule for guidance. The first subsection of OAR 150-314.665(4) provides a detailed and informative restatement of ORS 314.665(4) itself: “(1) In General. ORS 314.665(4) provides for the inclusion in the numerator of the sales factor of gross receipts from transactions other than sales of tangible personal property (including transactions with the United States Government); under this section gross receipts are attributed to this state if the income producing activity that gave rise to the receipts is performed wholly within this state. Also, gross receipts are attributed to this state if, with respect to a particular item of income, the income producing activity is performed within and without this state but the greater proportion of the income producing activity is performed in this state, based on costs of performance.” OAR 150-314.665(4) (1). We read that subsection to establish the appropriate framework to be used to determine whether sales other than sales of tangible personal property are included in the numerator of the sales factor. The subsection first directs us to begin with the gross receipts from sales of other than tangible personal property. Next, we identify the income-producing activity that generated those gross receipts. We then determine where that income-producing activity was performed. If the activity was entirely performed in Oregon, then the associated gross receipts are counted in the Oregon sales factor’s numerator. If the activity was performed in more than one state, then we determine which state had the largest share of that income-producing activity, based on the direct costs of producing the sale or sales. 14 If more of that income-producing activity occurred in Oregon than in any other state, then all of the income is counted in the Oregon numerator. Otherwise, none of it is. *711 The first question, then, involves the meaning of “income-producing activity.” Even before we turn to the subsection of the rule that defines the term, we can make certain generalizations about it. The statutory term itself indicates that an “income-producing activity” is something that produces income for the taxpayer. Similarly, the subsection that establishes the analytical framework adds that an “income-producing activity” is something that generates “gross receipts” and results in an “item of income.” See OAR 150-314.665(4) (1) (focusing on whether “the income producing activity that gave rise to the receipts” meets certain conditions or whether, “with respect to a particular item of income, the income producing activity” meets other conditions); see also OAR 150-314.665(4) (2) (“the term ‘income producing activity’ applies to each separate item of income”). The subsection defining “income-producing activity” provides as follows: “(2) Income Producing Activity; Defined. The term ‘income producing activity’ applies to each separate item of income and means the transactions and activity directly engaged in by the taxpayer in the regular course of its trade or business for the ultimate purpose of obtaining gains or profit. Except as provided otherwise in this rule, such activity does not include transactions and activities performed on behalf of a taxpayer, such as those conducted on its behalf by an independent contractor. Accordingly, income producing activity includes but is not limited to the following: “(a) The rendering of personal services by employees or the utilization of tangible and intangible property by the taxpayer in performing a service. “(b) The sale, rental, leasing, franchising, licensing or other use of real property. “(c) The rental, leasing, franchising, licensing or other use of tangible personal property. “(d) The sale, franchising, licensing or other use of intangible personal property. The mere holding of intangible personal property is not, of itself, an income producing activity.” OAR 150-314.665(4) (2). *712 That definition adds that an income-producing activity itself includes “transactions and activity * * * by the taxpayer.” OAR 150-314.665(4) (2). The “transactions and activity” that constitute the “income-producing activity” must have three qualities: they must be “directly engaged in by the taxpayer”; they must be done “in the regular course of [the taxpayer’s] trade or business”; and they must have the “ultimate purpose of obtaining gains or profit.” Id. There is, however, an additional requirement. OAR 150-314.665(4) (2) states, “The term ‘income producing activity’ applies to each separate item of income [.]” The term “item of income” is not defined in the rules. Its ordinary meaning is not helpful in this context, and we cannot find any evidence that “item of income” is a term of art in the accounting industry. The department takes the position that “item of income” means an individual exchange between a buyer and a seller: “Each particular or separate item of income is a sale — a transaction — an exchange between a buyer and a seller.” Whether the transmissions are sold individually or are sold in bulk, so to speak (for a flat monthly rate), is irrelevant to the analysis. The department’s position regarding the meaning of “item of income” is plausible and not inconsistent with the text of the rule in its context, or with the statute, or with any other source of law. Accordingly, we defer to that interpretation. See Don’t Waste Oregon Com., 320 Or at 142. That narrow interpretation of “item of income” necessarily limits the definition of “income-producing activity.” To identify the income-producing activity, we must identify for each “item of income” — for each individual sale — the “transactions and activity directly engaged in by the taxpayer in the regular course of its trade or business for the ultimate purpose of obtaining gains or profit.” OAR 150-314.665(4) (2). The analysis thus generally begins with each item of income — each individual sale — and determines the relevant transactions and activity associated with that sale. The result is an “income-producing activity.” Whatever ambiguity may exist in the statutory text of ORS 314.665(4) alone, then, the rule removes it. As we *713 noted earlier, it seems likely that Oregon’s UDITPA sales factor statutes were intended to allocate individual sales: sales of tangible personal property by where the goods are delivered or shipped (ORS 314.665(2)) and other sales by where the income-producing activity is mainly or entirely performed (ORS 314.665(4)). The rule, as permissibly interpreted by the department, removes any doubt. Having identified the income-producing activity, we must next determine where the taxpayer performs it. The first part of the inquiry involves an ostensibly simple question: Was “the income producing activity that gave rise to the receipts *** performed wholly within this state”? If so, then the analysis ends, and the receipts from that individual sale are counted as in-state sales for purposes of the numerator of the sales factor. If the income-producing activity is performed in more than one state, however, the Tax Court must then determine whether “the greater proportion of the income producing activity is performed in this state, based on costs of performance.” To know whether “the greater proportion” of the activity occurs in this state, the Tax Court must necessarily compare the costs of performance in different states. As noted, the rule defines “costs of performance” as follows: “(4) Costs of Performance; Defined. The term ‘costs of performance’ means direct costs determined in a manner consistent with generally accepted accounting principles and in accordance with accepted conditions or practices in the trade or business of the taxpayer. For purposes of this rule, direct costs do not include costs that are not part of the income producing activity itself, such as accounting or billing expenses.” OAR 150-314.665(4) (4). To identify the “costs of performance,” then, the Tax Court must identify the “direct costs.” The identification of direct costs will depend both on “generally accepted accounting principles” as well as “accepted conditions or practices in the trade or business of the taxpayer.” Expert testimony at trial indicated that the proper identification of the “income-producing activity” would drive *714 whether costs would count as “direct costs” for the “costs of performance” analysis. Witnesses on both sides — Allen for AT&T, and Starkey and Michelle Henney, accounting experts, for the department — all agreed in substance that one must know what is being “costed” in order to identify what constitute the “direct costs” of that thing. Allen testified: “A direct cost is that cost which is directly attributable or directly related to that which I’m trying to cost. So I’m not trying to be obtuse here.” Henney explained: “[W]e have to specify what are we talking about before we can find the causal relationship that is direct or indirect.” And Starkey noted: “[F]or purposes of a separate call, as we have talked about earlier, a single call, those [direct] costs would be very different than they would be for all calls.” E. Application With that legal background, we turn to the issues presented in this case. Again, as noted, the burden of proof is on AT&T to demonstrate its entitlement to a refund. See ORS 305.427. As a practical matter, that means AT&T had to introduce evidence showing that, in connection with its sales of interstate and international voice and data transmission, a greater share of the “costs of performance” for each “income-producing activity” was incurred in a state other than Oregon. The parties here offered two competing interpretations of what constitutes AT&T’s income-producing activity. AT&T’s interpretation focused on the operation of the network broadly, which was part of its justification for treating network costs as “costs of performance.” The department’s interpretation focused on individual transactions with customers, and that was part of its justification for concluding that network costs should be left out of the “costs of performance” analysis. As we have stated, we agree with the department’s interpretation. OAR 150-314.665(4) (2) indicates that an *715 “income-producing activity” is something associated with an individual “item of income.” The department interprets “item of income” to relate to individual sales, either per-minute charges for phone calls or flat-rate monthly subscriptions (or similar data transmission sales). That interpretation is plausible and is not inconsistent with any source of law that AT&T has identified. That narrow definition of “item of income” correspondingly affected the meaning of “income-producing activity.” An “income-producing activity” consists of the “transactions and activity” that the taxpayer performs for each individual transaction — “each separate item of income.” OAR 150-314.665(4) (2). In other words, the “income-producing activity” consists of those transactions and activity that produced each individual sale, i.e., each per-minute phone call or transmission or each monthly flat-rate bill. AT&T’s network-focused interpretation of the term is too broad. That alone would justify affirming the Tax Court’s decision. For AT&T to meet its burden of proof in this case, it had to first calculate the costs of performance for its income-producing activities. But AT&T’s cost study did not identify the correct income-producing activities. The cost study thus did not show what the costs of performance were for the correct income-producing activities. In addition, AT&T’s erroneous interpretation of “income-producing activity” also distorted its calculation of the “costs of performance.” The experts here agreed that a transaction-based interpretation of “income-producing activity” would also narrow what constitute the “direct costs” for the “costs of performance” part of the analysis. AT&T’s own expert, Allen, acknowledged that, if the income-producing activity here was individual phone calls, then the “direct costs” arguably would not include network costs. After first explaining AT&T’s position that one should determine the “income-producing activity” at the level of the network or system — in which case all the costs of the network would be “direct costs” — Allen then explained that the “direct costs” might be different if one examined the “income-producing activity” at the level of individual calls: *716 “If * * * I’m looking at a call-by-call basis, doing a different type of modeling, I might be able to consider something less direct in that nature. So an example might be if I were to determine that direct cost is based on what I’ll call this notion of the incremental cost. “So as an example, the network is there. It exists. If I place a call on that network, I’m using it. However, if I don’t place that call, that network is still there. I don’t — that cost I have still incurred. “So the argument might be made that, therefore, that call, the network is not a direct cost associated with that call. The argument I make is that’s an inappropriate level, especially as it relates to the questions we are attempting to answer.” That is precisely the department’s position: The direct costs of the income-producing activity, each individual phone call or monthly flat-rate billing, are only those incremental costs associated with each individual call or billing, not overall network costs. Starkey testified: “It is an issue of — just to preface, it is an issue of the increment that you are choosing. For one call — and kind of think of it from a marginal versus sunk cost perspective. If you are looking at the cost of one call and the network is there, then, as I think I describe in the report, the cost associated with that one call is just the additional expense you incur to carry that one call. “And in this circumstance, that’s likely something very small. It is going to be the access charge you pay for purposes of using someone else’s network to help carry that call and then probably — and * * * also just the electricity necessary to carry that particular call.” Thus, AT&T’s cost study again failed to meet AT&T’s burden of proof. AT&T’s network-based interpretation of “income-producing activity” implied that network costs counted as direct costs. But a transaction-based interpretation of income-producing activity means that network costs do not qualify as direct costs. AT&T’s cost study did not identify the relevant non-network costs for each income-producing activity. Without a correct calculation of those direct costs, the Tax Court could not tell where the greater part of those costs were incurred. The court certainly had *717 no basis to find that the greater part of those costs were incurred in some state other than Oregon. The same problem also undermined the second part of AT&T’s cost study, which analyzed the costs of “Oregon demand.” As with the first part of the cost study, AT&T counted overall network costs as part of the “direct costs” of Oregon demand. However, the income-producing activities are the individual transactions, and so the “direct costs” are only those incremental costs associated with each transaction. Because the second part of the cost study treated network costs as “direct costs,” it failed to show what the true “costs of performance” were, and this resulted in AT&T not meeting its burden of proof. 15 III. CONCLUSION To succeed on its claim for a refund, AT&T was required to (1) show the costs of performance for each income-producing activity, and then (2) show that the greater part of those costs had been incurred in some state other than Oregon. The cost study that AT&T submitted did not identify the correct income-producing activities and it did not correctly calculate the costs of performance for those activities. AT&T thus failed to make the showing required to meet its burden of proof. Accordingly, we conclude that the Tax Court properly denied AT&T’s request for a refund. We need not, and do not, address the other questions presented by the parties, including whether the access charges paid to local exchange carriers should be counted as part of the relevant income-producing activity. The judgment of the Tax Court is affirmed. Footnotes [1] In general, all Oregon statutes and administrative rules discussed in this opinion are the 1999 versions, which were unchanged in their relevant parts during the tax years at issue here. The exception is OAR 150-314.665(4), in which we refer to the 2008 version containing amendments effective January 1, 2007: AT&T conceded in the Tax Court that the 2007 amendments to that rule applied retroactively to the tax years at issue here, and the parties and the Tax Court used that version. [2] AT&T admitted that some of its other business activities effectively did not qualify as “income-producing activities.” It also admitted that the “costs of performance” for intrastate consumer and business calls and data were primarily incurred in Oregon, and so those sales were taxable here. [3] AT&T chose to compare Oregon costs to New Jersey costs because New Jersey was where its Global Network Operations Center was located. Under the statute, AT&T only needed to show that some state somewhere had a “greater proportion” of the costs of performance for the income-producing activity. ORS 314.665(4). [4] It is not clear to what extent the cost study actually depended on this particular classification of “income-producing activities.” Although the cost study’s written description asserts that the relevant income-producing activities are “service activation,” “service assurance,” and “service execution,” the cost study itself breaks costs down according to a completely different set of categories: “network and other costs” (including network engineering, network support, network services, and field operations), “depreciation and amortization,” and “customer care.” Those are the only categories for which the cost study actually calculates costs; nowhere does the cost study identify the direct costs per year for “service activation,” for example. And the cost categories themselves (network, depreciation, customer care) are not simply different names for AT&T’s identified income-producing activities. The study indicates that those cost categories apply to more than one income-producing activity: Both “service activation” and “service assurance” include some share of “customer care” costs, for example, while “network and other costs” seemingly apply to all three identified “income-producing activities.” (The cost study contains a table that lists the cost categories involved in each “income-producing activity,” but that table does not contain any row identifying the cost categories for “service execution.” Presumably “service execution” involves both some share of “network and other costs” as well as “depreciation and amortization.”) [5] The cost study does not specifically define what it considers “Oregon demand,” but Allen in his testimony described what it meant, at least in the context of phone calls. Based on rules derived from sales and use taxes, a call would be counted as an Oregon call if Oregon was the state for at least two of the following three criteria: the state where the call began, the state where the call ended, or the state where the transaction was billed. [6] The department had argued that one should begin with Oregon calls only. The Tax Court characterized that as a permissible “shortcut” that the department could take with “its analysis”; “the department is simply starting with a smaller census” of possible revenues, and doing so could only benefit AT&T. 20 OTR at 305. We note, however, that the department did not present its own cost study. The only cost study before the court was that prepared by AT&T. [7] ORS 314.610(6) defines a “public utility” to include “any business entity whose principal business is * * * the transmission of communications.” [8] A provision of Oregon’s UDITPA adds that UDITPA does not apply to public utilities. ORS 314.615. “Taxpayers engaged in activities as a financial organization or public utility shall report their income as provided in ORS 314.280 and 314.675.” [9] ORS 314.280 provides, in part: “(1) If a taxpayer has income from business activity * * * as a public utility (as defined respectively in ORS 314.610 *** (6)) which is taxable both within and without this state (as defined in ORS 314.610(8) and 314.615), the determination of net income shall be based upon the business activity within the state, and the Department of Revenue shall have power to permit or require either the segregated method of reporting or the apportionment method of reporting, under rules and regulations adopted by the department, so as fairly and accurately to reflect the net income of the business done within the state. “(2) The provisions of subsection (1) of this section dealing with the apportionment of income earned from sources both within and without the State of Oregon are designed to allocate to the State of Oregon on a fair and equitable basis a proportion of such income earned from sources both within and without the state. * * *” [10] Unitary businesses are in contrast to nonunitary businesses, which are “business entities that are connected by common ownership but that exist independently and in different states.” Crystal Communications, 353 Or at 303. Nonunitary businesses are taxed using the segregated method of reporting mentioned in ORS 314.280. Id. [11] If a taxpayer does not operate a unitary business, then a state generally lacks the constitutionally necessary basis to use apportionment to tax the income that the taxpayer earned outside of the state. See, e.g., OAR 150-314.610(1)-(A) (6) (discussing constitutional limits and “unitary business principle”). On the other hand, if the taxpayer does operate a unitary business, then taxing only those parts of the business operations that are located in-state is to “endeavor!] to treat separately what is, in fact, inseparable.” George H. Weissman, Unitary Taxation: Its History and Recent Supreme Court Treatment, 48 Albany L Rev 47, 50-51 (1983) (internal quotation marks and citation omitted). [12] That is the general definition. For purposes of the sales factor, some other types of gross receipts are excluded from the definition of “sales.” See ORS 314.665(6) (listing additional exclusions). This case does not involve any dispute over those exclusions from the definition of “sales.” For purposes of this opinion, it is adequate to treat “sales” as simply being gross receipts. [13] The Multistate Tax Commission amended its model regulation in 2007, but the department had not incorporated those changes into the version of OAR 150-314.665(4) applicable to this case. See Multistate Tax Commission Allocation and Apportionment Regulations at 1, available at http://bit.ly/lRSr38z (accessed July 14, 2015) (noting amendments to Regulation IV.17). [14] This does not necessarily require evidence of the costs of performance in every state. If a taxpayer is attempting to prove that sales of other than tangible personal property should not be counted in the Oregon numerator, then it is sufficient for the taxpayer to show that at least one state has a greater share of the costs of performing the related income-producing activity. [15] We note that, if the department’s interpretation of OAR 150-314.665(4) (2) leads to unfair results in particular cases, then either the taxpayer or the department may seek an alternative method of apportionment. See ORS 314.670 (UDITPA section 18; if ordinary method of apportionment under UDITPA does not “fairly represent the extent of the taxpayer’s business activity in this state,” then alternative method may be used); see also OAR 150-314.280-(M) (2) (for public utilities being taxed under ORS 314.280, if ordinary methods of apportionment “do not fairly and accurately reflect the net income of the business done within Oregon,” then alternative method of apportionment may be used).