Opinion
Plaintiffs and defendant-intervenors, SKF USA Inc. and SKF GmbH (collectively “SKF”) and FAG Kugelfischer Georg Schafer AG and FAG Bearings Corporation (collectively “FAG”), move pursuant to Rule 56.2 of the Rules of this Court for judgment on the agency record challenging the Department of Commerce, International Trade Administration’s (“Commerce”) final determination, entitled Antifriction Bearings (Other Than Tapered Roller Bearings) and Parts Thereof From France, Germany, Italy, Japan, Singapore, Sweden, and the United Kingdom; Final Results of Antidumping Duty Administrative Reviews and Partial Termination of Administrative Reviews (“Final Results”), 61 Fed. Reg. 66,472 (Dec. 17, 1996), as amended, Anti-friction Bearings (Other Than Tapered Roller Bearings) and Parts Thereof From Germany, Italy, Japan, and the United Kingdom: Amended Final Results of Antidumping Duty Administrative Reviews, 62 Fed. Reg. 3,003 (Jan. 21, 1997). Defendant-intervenor and plaintiff, The Torrington Company (“Torrington”) also moves pursuant to Rule 56.2 of the Rules of this Court for judgment on the agency record challenging Commerce’s Final Results.
SKF claims that Commerce erred in: (1) disregarding SKF’s negative home market billing adjustment number two values in calculating foreign market value (“FMV”); and (2) including SKF’s zero-value United States transactions in its margin calculations.
FAG claims that Commerce erred in: (1) disregarding transactions that had not failed its profit comparison test when calculating constructed value (“CV”) for cylindrical roller bearings (“CRBs”); (2) including general and administrative (“G&A”) expenses unrelated to FAG’s further manufacturing activities in its calculation of increased value for further manufacturing; and (3) including losses related to the sale of FAG’s Korean joint venture facility in its calculation of FAG’s G&A ratio.
Torrington claims that Commerce erred in: (1) accepting FAG’s reported research and development (“R&D”) costs because they were not reported on a product-specific or product-line basis; and (2) treating SKF’s home market billing adjustment number two values as indirect selling expenses (“ISE”) in calculating FMV
NTN Bearing Corporation of America and NTN Kugellagerfabrik (Deutschland) GmbH did not file a response brief to Torrington’s Rule *90756.2 motion for judgment on the agency record. SNR Roulements did not file any papers.
Background
This case concerns the fifth administrative review of the antidumping duty order on antifriction bearings (other than tapered roller bearings) and parts thereof (“AFBs”)imported to the United States during the review period of May 1, 1993 through April 30, 1994.1 Commerce published the preliminary results of the subject review on December 7, 1995. See Antifriction Bearings (Other Than Tapered Roller Bearings) and Parts Thereof From France, Germany, Japan, Singapore, Sweden, Thailand, and the United Kingdom; Preliminary Results of Antidump-ing Duty Administrative Reviews, Partial Termination of Administrative Reviews, and Notice of Intent to Revoke Order (“Preliminary Results”), 60 Fed. Reg. 62,817. Commerce published the Final Results on December 17, 1996. See 61 Fed. Reg. at 66,472.
Standard of Review
The Court will uphold Commerce’s final determination in an administrative review unless it is “unsupported by substantial evidence on the record, or otherwise not in accordance with law.” 19 U.S.C. § 1516a(b)(l)(B) (1994).
Discussion
I. Jurisdiction
The Court has jurisdiction over this matter pursuant to 19 U.S.C. § 1516a(a)(2) and 28 U.S.C. § 1581(c) (1994).
II. SKF’s Claims
A. SKF’s Home Market Billing Adjustment Number Two Values
Title 19, United States Code, §§ 1677a and 1677b require Commerce to determine the price actually charged to a customer both in the home market, that is, FMV and in the United States for the merchandise at issue. See 19 U.S.C. §§ 1677a, 1677b (1988). The actual price charged to a customer necessarily includes adjustments for discounts or rebates paid by the company to the customer. SKF reported billing adjustment two in the German home market which was used for debits and credits related to multiple invoices, invoice lines or products. Credits to customers were reported as negative values and decreased FMV Debits to customers were reported as positive values and increased FMV
In the Final Results, Commerce differentiated between SKF’s positive and negative billing adjustment values by making upward adjustments to the home market price for customer numbers that were positive and disregarding the reported values for negative numbers. See 61 Fed. Reg. at 66,498.
*908SFK complains that Commerce’s treatment of billing adjustment two had two adverse effects. First, SKF contends that Commerce’s disparate treatment of negative and positive values distorted the calculation of FMV so that it does not fairly represent the price actually paid by German customers. See SKF’s Br. Supp. Mot. J. Agency R. at 8. Specifically, SKF argues that by rejecting the negative values, Commerce did not properly take into account the credits granted to customers and, therefore, did not decrease FMV to the extent it should have. See id. at 8-9. SKF claims that the price distortion results in a skewed comparison between home and United States prices. See id. at 17.
Second, SKF asserts that Commerce included all positive values as direct adjustments in the margin calculations without determining whether they include out-of-scope merchandise. See id. at 30. SKF contends that Commerce had deviated from its principle of rejecting values derived from allocations by accepting the positive values. See SKF’s Br. Supp. Mot. J. Agency R. at 14. SEIF contends that denying both the positive and negative billing adjustments would have been “more consistent with the Department’s general position that billing adjustments derived from allocations should not be allowed.” Id. at 20.
Commerce’s position in the Final Results, however, is that by retaining positive price adjustments and rejecting negative ones, it provides to respondents a disincentive “to report positive billing adjustments on an allocated (e.g., customer-specific) basis in order to minimize their effect on the margin calculations.” 61 Fed. Reg. at 66,498.
In response to Commerce’s incentive rationale, SKF asserts that because it “does not know in advance for which customer numbers it will report increases [or decreases] to FMV * * * SKF could not manipulate individual values in order to achieve a beneficial result.” SKF’s Br. Supp. Mot. J. Agency R. at 18-19. Furthermore, SKF claims that the billing adjustment “is granted in a manner such that transaction-specific reporting is not feasible.” Id. at 5. The method used by SKF consists of totaling the credits and debits issued to a customer, dividing that total by the total sales to the customer number and applying the resulting factor to each reported sale to that customer number. See id.
SKF stipulates that it made no attempt to exclude out-of-scope merchandise from the positive or negative values because such action was not feasible. See id. at 31 n.25. Consequently, SKF demands that if Commerce determines that billing adjustment two contains out-of-scope merchandise, it must reject all values, whether positive or negative. See SKF’s Br. Supp. Mot. J. Agency R. at 30-31. Alternatively, SKF asks the Court to require Commerce to accept both positive and negative values as direct adjustments to FMV See id. at 31.
Commerce argues that because SKF did not tie the adjustments to specific transactions nor grant them as a fixed percentage across sales, the negative adjustments were properly denied. See Def.’s Partial Opp’n to Pis.’ Mots. J. Agency R. at 2-3. Commerce also argues that the acceptance of the positive values was proper because they increased FMV “to *909SKF’s detriment consistent with the principle that a party should not benefit from its non-compliance with Commerce’s request for information.” Id. at 3.
In essence, SKF’s main argument is that, because Commerce chose to accept SKF’s positive adjustments, it must accordingly accept the negative adjustments. In the alternative, SKF argues that Commerce should have rejected the positive adjustments since it rejected the negative adjustments. The Court finds that these propositions, however, are not reflected in the law. There is no requirement that Commerce treat modifications that increase respondent’s dumping margin and adjustments that decrease the margin in the same manner. Rather, the law supports the opposite conclusion. See SSAB Svenskt Stal AB v. United States, 21 CIT 1007, 1011-12, 976 F. Supp. 1027, 1032 (1997) (upholding Commerce’s selection of highest packing costs reported by respondent for United States sales with no accompanying deduction of packing expenses for FMV); see also INA Walzlager Schaeffler KG v. United States, 21 CIT 110, 126-29, 957 F. Supp. 251, 265-68 (1997) (remanding to Commerce to deny negative billing adjustments with no corresponding instructions regarding positive adjustments), opinion after remand, Slip Op. No. 97-141, 1997 WL 614300 (CIT Sept. 29, 1997), aff’d sub nom, SKF USA Inc. v. INA Walzlager Schaeffler KG, 180 F.3d 1370 (Fed. Cir. 1999). This is particularly true when Commerce is given data that is not responsive to its request for information, or when the respondent submits information in an improper form.
In INA, for example, Commerce treated certain home market expenses, including negative billing adjustments reported by a respondent on a customer-specific basis, as indirect billing expenses. Commerce treated positive billing adjustments as direct expenses to be deducted from FMV See id. at 265. This Court held that negative home market adjustments could not be treated as indirect expenses because, by their very nature, the adjustments constituted direct expenses. See id. at 267. The Court, therefore, remanded to Commerce to deny any adjustment to FMV for the respondent’s negative billing adjustment because the adjustment was improperly recorded. See id. at 268.
INA held that both positive and negative adjustments have the same nature, that is, both types are direct adjustments to FMV and must be reported in a particular manner. Id. at 267. Although INA did not expressly address the issue of disparate treatment of positive and negative billing adjustments, the Court’s order in INA remanding to Commerce to deny adjustments to FMV for respondent’s negative billing adjustments only clearly indicates the Court’s position that the law does not require either a blanket denial or a uniform acceptance of positive and negative billing adjustments to FMV
Having decided that positive and negative adjustments need not be treated in the same manner, the Court addresses Commerce’s denial of the negative billing adjustments first. The Court recognizes that it is well-established that Commerce’s decision to deny a direct adjustment *910to FMV is reasonable and proper if the adjustment sought is not reported on either a transaction-specific basis or as a fixed and constant percentage of the sales price of all transactions for which it was reported. See SKF USA Inc. v. United States, 19 CIT 625, 633, 888 F. Supp. 152, 159 (1995); SKF USA Inc. v. United States, 19 CIT 79, 86, 875 F. Supp. 847, 853 (1995); SKF USA Inc. v. United States, 19 CIT 54, 65, 874 F. Supp. 1395, 1405 (1995). “The party seeking a direct price adjustment bears the burden of proving entitlement to such an adjustment.” SKF USA Inc. v. United States, 180 F.3d 1370, 1377 (1999) (citing Fujitsu General Ltd. v. United States, 88 F.3d 1034, 1040 (Fed. Cir. 1996)).
Because SKF’s improper reporting made it impossible for Commerce to determine if the claimed adjustment pertained to the subject merchandise, Commerce determined that SKF had not met its burden. The Court finds, therefore, that Commerce properly declined to make the negative adjustments because of SKF’s failure to tie the expenses to specific transactions or products. See Torrington Co. v. United States, 82 F.3d 1039, 1050-51 (Fed. Cir. 1996) (“Torrington I”).
The Court, however, finds that Commerce properly accepted the positive billing adjustment. SKF itself indicated that there were positive billing adjustments which increased the dumping margins. Commerce exercised its discretion to grant the adjustment as reported. Prohibiting Commerce from granting the positive adjustment in this case, especially when the adjustment was reported by respondent, would limit Commerce’s ability to obtain the information it requires in the appropriate form. As Commerce stated in the Final Results, if Commerce disregarded positive billing adjustments, “respondents would have no incentive to report these adjustments on a transaction-specific basis, as requested.” 61 Fed. Reg. at 66,498.
The Court, therefore, finds Commerce’s application of the billing adjustments to be a proper exercise of its authority to grant or deny adjustments. Because Commerce’s decision to grant the positive adjustment and to deny the negative adjustment was in accordance with law, Commerce’s determination is affirmed.
B. SKF’s Zero-Value United States Transactions
SKF argues that in light of NSK Ltd. v. United States, 115 F.3d 965, 975 (Fed. Cir. 1997), the Court should remand the matter to Commerce to exclude SKF’s zero-value transactions from the margin calculations. See SKF’s Br. Supp. Mot. J. Agency R. at 32. SKF’s rationale is that United States transactions at zero value, such as prototypes and samples, do not constitute true sales and, therefore, should be excluded from the margin calculations pursuant to NSK. See id. at 32-33. The identical issue was decided by this Court in SKF USA Inc. v. United States, 23 CIT 402, Slip Op. 99-56, 1999 WL 486537 (June 29, 1999).
Torrington concedes that a remand may be necessary in light of NSK, but argues that further factual inquiry by Commerce is necessary to determine whether the zero-price transactions were truly without consid*911eration or if they were'matched to sales above fair value in an effort to allow the customer to purchase merchandise below fair value. See Tor-rington’s Opp’n to SKF and FAG’s Mots. J. Agency R. at 12-13. Torring-ton argues that only if the transactions are truly without consideration can they fall within NSK’s exclusion. See id.
Commerce concedes that the case should be remanded to Commerce to exclude the sample transactions for which SKF received no consideration from SKF’s United States sales database. See Def.’s Partial Opp’n to Pis.’ Mots. J. Agency R. at 28.
Commerce is required to impose antidumping duties upon merchandise that “is being, or is likely to be, sold in the United States at less than its fair value.” 19 U.S.C. § 1673(1) (1988). A zero-priced transaction does not qualify as a “sale” and, therefore, by definition cannot be included in Commerce’s FMV calculation. See NSK, 115 F.3d at 975 (holding “that the term sold * * * requires both a transfer of ownership to an unrelated party and consideration”). Thus, the distribution of AFBs for no consideration falls outside the purview of 19 U.S.C. § 1673. Consequently, the Court remands to Commerce to exclude any transactions that were not supported by consideration from SKF’s United States sales database, and to adjust the dumping margins accordingly.
III. FAG’s Claims
A. The CV Profit Calculation
In making its CV calculation, Commerce is required to include an amount for profit. See 19 U.S.C. § 1677b(e)(l)(B). The amount for profit must be “equal to that usually reflected in sales of merchandise of the same general class or kind as the merchandise under consideration which are made by producers in the country of exportation, in the usual commercial quantities and in the ordinary course of trade.” Id.
Congress has also provided that certain transactions may be disregarded in the calculation of CV See 19 U.S.C. § 1677b(e)(2). Specifically, the statute provides:
a transaction directly or indirectly between [related parties] may be disregarded if, in the case of any element of value required to be considered, the amount representing that element does not fairly reflect the amount usually reflected in sales in the market under consideration of merchandise under consideration. If a transaction is disregarded under the preceding sentence and there are no other transactions available for consideration, then the determination of the amount required to be considered shall be based on the best evidence available as to what the amount would have been if the transaction had occurred between [unrelated parties].
Id.
Commerce conducts the arm’s-length test on a class-or-kind basis and compares sales prices to related parties with sales prices to unrelated parties in the same market. See Final Results, 61 Fed. Reg. at 66,493. Commerce does not automatically reject related-party sales that fail the arm’s-length test. See id. Such sales, however, may be disregarded if the *912amount of an element of value that must be considered “‘does not fairly reflect the amount usually reflected in sales in the market under consideration.’” Id. (quoting 19 U.S.C. § 1677b(e)(2)). Commerce employed the following method to examine that possibility:
[W]e compared profit on sales to related parties that failed the arm’s-length test to profit on sales to unrelated parties. If the profit on sales to related parties varied significantly from the profit on sales to unrelated parties, we disregarded related-party sales for the purposes of calculating profit for CV We first calculated profit on sales to unrelated parties on a class-or-kind basis. If the profit on these sales was less than the statutory minimum[,] * * * we used the * * * minimum in the calculation of CV If the profit on these sales was equal to or greater than the * * * minimum, we calculated profit on the sales to related parties that failed the arm’s-length test and compared it to the profit on sales to unrelated parties as described above. If the profits on such sales to related parties varied significantly from the profits on sales to unrelated parties, we excluded those related-party sales for the purpose of calculating profit on CV
Final Results, 61 Fed. Reg. at 66,493.
FAG argues that 19 U.S.C. § 1677b(e)(2) requires “Commerce to review sales for use in constructed value calculations on a transaction-by-transaction basis, and Commerce is only authorized to disregard those particular transactions that do not fairly reflect the usual profit amount or other value element under consideration.” FAG’s Br. Supp. Mot. J. Agency R. at 7-8. FAG complains that instead of merely disregarding transactions that failed the arm’s-length test, Commerce discarded FAG’s entire class or kind profit calculation and recalculated CV profit for CRB sales. See id. at 5. FAG claims that Commerce had the ability to isolate transactions that did not reflect normal profit levels and should have exercised that ability to disregard only those transactions and use the remaining sales instead of recalculating profit. See id. at 9-10. FAG argues that Commerce’s resort to best evidence available was unlawful. See id. at 9.
Torrington maintains that Commerce was correct in resorting to FAG’s home-market sample to calculate a profit for use in the CV calculations. See Torrington’s Opp’n to SKF and FAG’s Mots. J. Agency R. at 16. Torrington contends that Commerce properly refused to use FAG’s class or kind data because it possibly contained transactions that failed the arm’s-length test. See id. Torrington maintains that Commerce had reasonable grounds “to base its CV profit calculation on FAG’s home-market sales database minus sales failing the arm’s-length test.” Id. at 17.
*913In its brief, Commerce supports its method of calculating profit, but requests a remand for another reason. See Def.’s Partial Opp’n to Pis.’ Mots. J. Agency R. at 31-32. Specifically, Commerce contends that:
[t]he fact that Commerce performed the arm’s-length test on a customer-specific basis during the fifth review * * * necessitates a remand so that the basis of the profit-variance test can be consistent with the basis of the arm’s-length test. Commerce’s profit-variance test in this review grouped together all of the customers which failed the arm’s-length test rather than applying the profit-variance test to each individual customer which fañed the arm’s-length test. If Commerce were permitted to do the latter upon remand, then, where the sales to a related customer “fail” the arm’s-length test, the profit on all of the sales to that related customer will be compared to the profit on unielated party sales.
Id. at 31.
The issue confronting the Court is whether Commerce may properly disregard FAG’s class or kind data and recalculate profit on the basis of arm’s-length sales to related parties and sales to unrelated parties. There is nothing in the statute that prevents Commerce from relying on arm’s-length sales to related parties and sales to unrelated parties when Commerce finds that reported transactions do not fairly reflect the value of the transactions being considered. See 19 U.S.C. § 1677b(e)(2). Additionally, there is nothing in the legislative history of § 1677b(e)(2) which prohibits Commerce from resorting to such sales as the best evidence available. Because Congress is silent on the particular issue facing the Court, the Court’s next task is to examine Commerce’s actions to determine whether they reflect a reasonable interpretation of § 1677b(e)(2). See Koyo Seiko Co. v. United States, 36 F.3d 1565, 1573 (Fed. Cir. 1994) (citing Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984)).
This Court’s decision in INA Walzlager Schaeffler KG v. United States, 21 CIT 110, 957 F. Supp. 251 (1997), opinion after remand, Slip Op. No. 97-141, 1997 WL 614300 (CIT Sept. 29, 1997), aff’d sub nom, SKF USA Inc. v. INA Walzlager Schaeffler KG, 180 F.3d 1370 (Fed. Cir. 1999), is instructive. In that case, INA opposed Commerce’s use of such or similar merchandise in its calculation of profit. See INA, 21 CIT at 114, 957 F. Supp. at 257. INA argued that “home market sales of such or similar merchandise are not representative of home market sales of the entire class or kind of merchandise,” as required by 19 U.S.C. § 1677b(e). Id. Commerce replied that it could not use the class or kind data because the variance in profit amounts “tainted the profit calculated by INA for the class or kind of merchandise.” Id. at 258. This Court agreed, and approved of Commerce’s resort to such or similar sales as the best evidence available. See id. at 259. In doing so, this Court rejected INA’s argument that the best evidence available was INA’s submitted general class or kind profit information since it would be illogical to “have Commerce use the very information that [it] rejected as unrep*914resentative of the sales under consideration as best evidence available.” INA, 21 CIT at 118, 957 F. Supp. at 259.
Similarly, the Court rejects FAG’s argument that Commerce is not permitted to resort to arm’s-length sales to related parties and sales to unrelated parties. Commerce was unable to use FAG’s reported profit because FAG’s data contained sales to related parties “at below arm’s-length prices and because the profit in the aggregate on these sales was significantly lower than the profit on other sales.” Def.’s Partial Opp’n to Pis.’ Mots. J. Agency R. at 29. Additionally, Commerce could not calculate profit on a class or kind basis (i.e., on all of FAG’s sales during the period of review) because Commerce only had data on record for the reported sales. See id. at 33. Thus, Commerce “recalculated profit on the basis of the other sales (both arm’s-length sales to related parties and sales to unrelated parties).” Id. at 29.
As in INA, it would be illogical for this Court to require Commerce to use FAG’s class or kind data as the best evidence available since Commerce found that the data contained sales to related parties that failed the arm’s-length test and whose profit varied from the profit on other sales. FAG’s suggestion that Commerce extract the transactions that do not reflect normal profit levels from the class or kind data is not viable since Commerce only has data on record for the reported sales. Thus, the Court finds that Commerce’s resort to arm’s-length sales to related parties and sales to unrelated parties as the best evidence available is in accordance with the law.
Although the Court finds that Commerce’s method of calculating profit is proper, the Court grants Commerce’s request for a remand. Commerce is directed to apply the profit-variance test to each customer who failed the arm’s-length test in order to ensure that the profit-variance test is consistent with the arm’s-length test.
B. G&A Expenses in the Calculation of Increased Value for Manufacturing
Title 19, United States Code, § 1677a(e)(3) provides that the exporter’s sale price shall be reduced by the amount of:
any increased value, including additional material and labor, resulting from a process of manufacture or assembly performed on the imported merchandise after the importation of the merchandise and before its sale to a person who is not the exporter of the merchandise.
Commerce’s regulations provide that additional adjustments to the exporter’s sales price may consist of reductions for:
[a]ny increased value resulting from a process of production or assembly performed on the merchandise after importation and before sale to a person who is not the exporter of the merchandise, which value the Secretary generally will determine from the cost of mate*915rial, fabrication, and other expenses incurred in such production or assembly.
19 C.F.R. § 353.41(e)(3) (1989).2
FAG “included the portion of group administrative expenses related to production” in its calculation of increased value for further manufactured products, but “did not include that portion of home market incurred group G&A related to sales.” FAG’s Br. Supp. Mot. J. Agency R. at 10. FAG maintains that the group market G&A related to sales, which consisted of group-level headquarters expenses and broadly-based R&D, was incurred in the home market and, therefore, did not relate to the further manufacturing in the United States. See id. at 11. FAG asserts that “only those G&A expenses directly related to the further manufacturing activity in the United States should be included in the calculation of ‘increased value’ under section 1677a(e)(3).” Id. at 12.
Torrington argues that a general expense cannot be attributed to a particular sale but may benefit the entire company, including the further manufacturing operation in the United States. See Torrington’s Opp’n to SKF and FAG’s Mots. J. Agency R. at 23. Torrington contends, therefore, that Commerce’s decision to allocate a portion of FAG’s headquarter expenses to the United States operations was reasonable since the expenses were allocable to operations in the United States. See id.
Commerce included the sales-related G&A in the further manufacturing calculation, finding that:
group-level headquarters expenses and broadly based R&D benefit all group members, including U.S. subsidiaries engaged in adding value. While FAG Germany reported such expenses for the cost of the parts imported, it did not include such expenses in the cost of further processing in the United States. In addition, we consider these expenses to affect the processing cost in the United States as well as support sales. Therefore, we have recalculated the G&A expenses for further processing in the United States to include group-level headquarters expenses and broadly based R&D expenses.
Final Results, 61 Fed. Reg. at 66,506. Commerce maintains that it is proper to include expenses not directly related to production or assembly, such as salaries of non-sales personnel, that “are by their very nature not directly incurred in production or assembly activities.” Def.’s Partial Opp’n to Pis.’ Mots. J. Agency R. at 34. Commerce argues that including such expenses is proper because they indirectly benefit the production operations. See id.
The issue confronting the Court is whether Commerce may properly include group-level headquarters expenses and broadly-based R&D in the calculation of increased value for further manufacturing. Neither the antidumping statute nor Commerce’s own regulations prohibit the reduction of exporter’s sales price by any increased value attributable to *916group-level headquarters expenses and broadly-based R&D. See 19 U.S.C. § 1677a; 19 CFR § 353.41. Additionally, there is nothing in the legislative history of § 1677a which prohibits such action. Because Congress is silent on the particular issue facing the Court, the Court’s next task is to examine Commerce’s actions to determine whether they reflect a reasonable interpretation of § 1677a(e)(3). See Koyo Seiko, 36 F.3d at 1573 (citing Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984)).
The Court finds that including group-level headquarters expenses and broadly-based R&D in the calculation of increased value for further manufacturing is a reasonable construction of § 1677a(e)(3) and that Commerce’s decision was in accordance with law. Commerce included the group-level headquarters expenses and broadly-based R&D because it determined that they imputed a benefit to United States subsidiaries for both the processing cost and support sales. See Final Results, 61 Fed. Reg. at 66,506. Thus, the increase in value from the expenses was directly linked to the “process of manufacture and assembly” as contemplated by the applicable statute. 19 U.S.C. § 1677a(e)(3). Accordingly, Commerce’s determination is affirmed.
C. The Losses Related to the Sale of FAG’s Korean Joint Venture Facility
FAG argues that the loss from the sale of a joint-venture facility in Korea should not have been included in G&A because the facility was not involved with the production of the merchandise at issue. See FAG’s Br. Supp. Mot. J. Agency R. at 13. In response, Torrington argues that the loss was properly included because the company’s activities as a whole, including those of a parent company or other division, should be accounted for in G&A. See Torrington’s Opp’n to SKF and FAG’s Brs. Supp. Mots. J. Agency R. at 26.
Title 19, United States Code, § 1677b(e)(l)(B) requires Commerce to calculate G&A expenses for CV using an amount “equal to that usually reflected in sales of merchandise of the same general class or kind as the merchandise under consideration which are made by producers in the country of exportation.” The statute does not define G&A nor specify how it should be allocated. See id. When Congress remains silent on a specific issue, the Court will defer to Commerce’s interpretation of a statutory provision, as long as it is based on a permissible construction of the statute. See Chevron, 467 U.S. at 843-44.
In the Final Results, Commerce had determined that the loss from the sale of the Korean facility was to be included in FAG’s G&A expenses.3 Commerce included the loss because it “relates to the overall operation of the company [and,] [t]herefore, it is most appropriately characterized *917as a G&A expense.” 61 Fed. Reg. at 66,497. Commerce allocated the loss “on the basis of all costs incurred by the company during the [period of review], including non-subject merchandise.” Id.
The issue is whether Commerce properly included the loss from the sale of the Korean facility in FAG’s G&A even though the facility was not involved in the production of the subject merchandise. The Court finds that U.S. Steel Group v. United States, 22 CIT 104, 998 F. Supp. 1151 (1998), is instructive. The issue in U.S. Steel was whether Commerce properly included an offset for “miscellaneous income” in calculating G&A for cost of production where the “miscellaneous income” was related to the general operations of the company. U.S. Steel, 22 CIT at 106, 998 F. Supp. at 1153. The Court found that G&A expenses “ ‘relate to the activities of the company as a whole rather than to production process.’” Id. at 107, 998 F. Supp. at 1154 (quoting Rautaruukki Oy v. United States, Slip Op. No. 95-56, 1995 WL 170399 (CIT Mar. 31, 1995)). The Court upheld “Commerce’s decision that offsets to G&A expenses should also be related to. the company’s general operations.” Id. In upholding Commerce’s determination, the Court agreed with Commerce that “limiting offsets to G&A expenses to income from activities related to ‘production of the subject merchandise’ would be inconsistent with the accounting allocation concept of G&A expenses.” Id. at 106-107, 998 F. Supp. at 1153-54.
Similarly, the Court finds that FAG’s request to exclude the loss from the sale of the Korean facility from the calculation of G&A in this case would be inconsistent with the accounting allocation concept of G&A expenses. G&A expenses are those which relate to the overall operation of the company; by their very nature, therefore, they are not directly traceable to any one specific product or group of merchandise. Contrary to FAG’s contentions, there is no requirement that the Korean facility have been involved with the production of the subject merchandise in order for the expense from its loss to be included in G&A expenses. All that Commerce needed to find was that the loss could be properly characterized as a G&A expense in order to allocate part of the expense to the subject merchandise, and this is exactly what Commerce did. See Final Results, 61 Fed. Reg. at 66,497. Accordingly, the Court finds that Commerce’s decision to include the loss because it relates to the overall operation of the company results from a reasonable construction of § 1677b(e)(l)(B). Commerce’s determination is affirmed.
IV. Torrington’s Claims
A. FAG’s Reported R&D Costs
Torrington argues that FAG failed to comply with Commerce’s questionnaire in calculating R&D costs by means of a broadly based factor rather than on a product-specific or product-line basis. See Torrington’s Br. Supp. Mot. J. Agency R. at 4. Specifically, Torrington claims that “FAG reported product-line R&D based on expenses for all roller bearing products!,] * * * fail[ing] to distinguish among specific products or *918product lines, as the questionnaire requested,” or even among the different types of bearings. Id. at 7. Torrington contends that “aggregate allocation of product-specific R&D expenses is allowable only where the record contains evidence that R&D expenses relating to one product line equally benefitted other product lines,” and that is not the situation here. Id. at 7-8.
FAG disputes Torrington’s contentions, arguing that Torrington took Commerce’s instruction that the respondent should “report ‘specific product or product line’ R&D for all products under review” out of context. FAG’s Resp. to Torrington’s Br. Supp. Mot. J. Agency R. at 3-4 (internal quotations omitted). FAG contends that the questionnaire actually instructs the respondent to calculate R&D with a broadly based factor by requiring that “‘[g]eneral R&D for the company which is not related to a specific product or product line should be allocated to all products of the company and included in general expenses.’” Id. at 4. FAG maintains that it cannot isolate R&D expenses to any specific group of goods. See id. at 6.
The Court finds that Commerce’s acceptance of FAG’s allocation was in accordance with law. Commerce found upon verification that “allocating FAG Germany’s R&D expenses on a product-specific basis would not be feasible because a large portion of R&D projects are on-going and benefit more than one product or category of products.” Final Results, 61 Fed. Reg. at 66,491. Commerce found that FAG’s allocation method was appropriate because “to the extent possible, R&D expenses have been assigned directly to particular manufacturing and distribution cost-center areas.”4 Id. Thus, Commerce made the most appropriate allocation of FAG’s R&D expenses under the circumstances. The Court does not find that such action was unreasonable, nor will it require a more specific allocation where Commerce has determined that none is possible. Accordingly, Commerce’s determination is affirmed.
B. SKF’s Home Market Rebate Number Two Values
SKF made rebate payments to certain of its “dealer/distributor customers to compensate them for competitive conditions in the German market.” Final Results, 61 Fed. Reg. at 66,502. By granting them support rebate payments if they did not meet the minimum profit level, SKF insured that its dealers/distributors would reap an overall minimum profit level. See SKF’s Resp. to Torrington’s Br. Supp. Mot. J. Agency R. at 17. Because the rebates were based on sales by SKF’s customers rather than to SKF’s customers, SKF states that “payment can only be allocated over the entire sales base to the dealer/distributor.” Final Results, 61 Fed. Reg. at 66,502. SKF asks the Court to remand to *919Commerce with instructions to allow rebate two as a direct deduction from price instead of as an ISE. See SKF’s Resp. to Torrington’s Br. Supp. Mot. J. Agency R. at 21.
Commerce determined that the rebate expense was not directly related to sales by SKF to its customers and, therefore, was treated as an ISE adjustment pursuant to Zenith Elecs. Corp. v. United States, 77 F.3d 426 (Fed. Cir. 1996). See Final Results, 61 Fed. Reg. at 66,502. Commerce explained that the rebate was a “promotional expense that does not relate to any particular sale by SKF Germany and does not vary with the quantity of merchandise that SKF Germany sells.” Id.
Torrington argues that Commerce improperly treated SKF’s home market rebate number two values as ISEs in calculating FMV and it should have denied the claimed adjustment to FMV See Torrington’s Br. Supp. Mot. J. Agency R. at 12. Torrington cites Torrington I, 82 F.3d at 1050-51, for the proposition “that Commerce may deduct direct expenses from FMV when they are properly reported on a transaction-specific basis, but that, if not so reported, Commerce may not treat such expenses as if they were indirect selling expenses.” Torrington’s Br. Supp. Mot. J. Agency R. at 13.
The issue centers around the proper categorization of rebate two as either a direct or indirect expense. A direct expense applied as an adjustment to FMV is inherently an expense which either varies with the quantity sold, Zenith, 77 F.3d at 431, or is “related to a particular sale,” Torrington Co. v. United States, 68 F.3d 1347, 1353 (Fed. Cir. 1995) (“Torrington II”). An indirect expense is one that does not vary with the quantity sold, Zenith, 77 F.3d at 431, or is “not related to a particular sale,” Torrington II, 68 F.3d at 1353.
SKF concedes that rebate two is “not ‘direct’ in a transaction-specific sense” since it does not directly relate to particular sales by SKF nor does it vary with the quantity of merchandise sold by SKF. SKF’s Resp. to Torrington’s Br. Supp. Mot. J. Agency R. at 18. Because “rebate 2 is structured to be paid in relation to SKF’s customers’ resales, it is impossible to establish a direct correlation between a single SKF sale and the later resales by that SKF customer (which resales may include a bearing sold by SKF in the initial transaction).” Id. at 6. Nevertheless, SKF argues that rebate two should be treated as a direct expense because it is “ ‘direct’ in a customer-specific sense.” Id. at 18.
SKF misses the mark. SKF itself conceded that the rebate expense cannot properly be classified as a direct expense, since it does not directly relate to particular sales by SKF nor does it vary with the quantity of merchandise sold by SKF. Commerce’s decision to treat SKF’s rebate two as an indirect expense is in accordance with law since the record shows that rebate two does not vary with the quantity sold, Zenith, 77 F.3d at 431, nor is it “related to a particular sale,” Torrington II, 68 F.3d at 1353.
The Court finds that the fact that the rebate varies with the quantity of merchandise sold by SKF’s dealers/distributors is irrelevant. SKF’s *920references to the way the rebate is structured to be paid and its allocation method do not alter the indirect character of the rebate because the direct or indirect character of an expense has nothing to do with the particular allocation method chosen by the respondent to report the expense. See Torrington I, 82 F.3d at 1051.
Although Commerce correctly treated the rebate as an indirect expense, it made no determination that the rebate was granted for in-scope merchandise only. Payment was “allocated over the entire sales base to the dealer/distributor.” Final Results, 61 Fed. Reg. at 66,502. The Court, therefore, remands the matter to Commerce to use a method which removes rebates paid on sales of out-of-scope merchandise from any adjustments made to FMV or, if no viable method can be developed, to deny such an adjustment in its calculation of FMV See Torrington Co. v. United States, 19 CIT 403, 421, 881 F. Supp. 622, 640 (1995).
Conclusion
The case is remanded to Commerce to: (1) exclude any transactions that were not supported by consideration from SKF’s United States sales database and to adjust the dumping margins accordingly; (2) apply the profit-variance test to each customer who failed the arm’s-length test before calculating the profit element of CV for FAG; and (3) remove rebates paid on sales of out-of-scope merchandise from any adjustments made to SKF’s FMV or, if no viable method can be developed, to deny such an adjustment in the calculation of FMV Commerce is affirmed in all other respects.