MICHEL, Circuit Judge.
This consolidated appeal concerns the Department of Commerce, International Trade Administration’s (“Commerce’s”) fourth annual administrative review of the antidumping order on certain antifriction bearings and parts thereof (the “antifriction bearings”). See Antifriction Bearings (Other Than Tapered Roller Bearings) and Parts Thereof From France, et al.; Final Results of Antidumping Duty Administrative Reviews, Partial Termination of Administrative Reviews, and Revocation in Part of Antidumping Duty Orders, 60 Fed. Reg. 10,900 (Dep’t Commerce, 1995) (“Final Results ”). On appeal, the Court of International Trade granted in part and denied in part various parties’ motions for judgment on the agency record and remanded to Commerce for various redeter-minations in accordance with its opinion. See NSK Ltd. v. United States, 969 F.Supp. 34 (Ct. Int’l Trade 1997) (“NSK /”). Commerce made the redetermina-tions as ordered on remand. See Final Results of Redetermination Pursuant to Court Remand, NSK Ltd., et al. v. United States, slip op. 97-74 (June 17, 1997), (Dep’t Commerce Apr. 28,1998) (“Remand Results ”). The Remand Results were subsequently affirmed by the Court of International Trade in their entirety. See NSK Ltd. v. United States, 4 F.Supp.2d 1264 (Ct. Int’l Trade 1998) (“NSK II”).
Plaintiffs-Appellants Koyo Seiko Co., Ltd. and Koyo Corporation of U.S.A. (together “Koyo Seiko”), Plaintiffs-Appellants NTN Bearing Corporation of America, American NTN Bearing Manufacturing Corp., NTN Corporation, NTN Driveshaft, Inc., and NTN-Bower Corporation (collectively “NTN”), and Defen-danWCross Appellant The Torrington Company (“Torrington”) now appeal and cross appeal to, this court from various aspects of NSK I and NSK II. We affirm the judgment of the Court of International Trade with respect to NTN’s appeal because we find no error, legal or factual, with regard to (i) Commerce’s inclusion of certain sample sales and sales with a sporadic sales history as home market sales in its calculation of foreign market value (“FMV”); (ii) Commerce’s exclusion of related party sales [1325]*1325from the home market sales used in its calculation of FMV; (iii) Commerce’s refusal to adjust FMV to take account of NTN’s reported home market discounts; and (iv) Commerce’s comparison of sales across different levels of trade in its calculation of FMV. However, with respect to the Court of International Trade’s rejection of the home market warranty expense factor reported by Koyo Seiko and accepted by Commerce as a “circumstances of sale” adjustment to FMV, we reverse on the grounds that Commerce’s acceptance of the adjustment was based upon a reasonable interpretation of the governing statute, accorded with applicable precedent, and its rejection by the court was therefore error. Finally, we affirm with respect to Torrington’s cross appeal of the determination of the United States price of the antifriction bearings bought and resold by Defendants-Appellees Honda Motor Co., Ltd., American Honda Motor Co., Inc., Honda of America Mfg., Inc. and Honda Power Equipment Mfg., Inc. (collectively, “Honda”). Like the Court of International Trade, we hold that Commerce reasonably interpreted the term “reseller” in the governing antidumping statute and that substantial evidence supports Commerce’s determination that Honda constitutes such a “reseller” with regard to its sales of subject antifriction bearings.
BACKGROUND
The fourth annual administrative review of the antidumping order at issue covered antifriction bearings entered during the period May 1, 1992, through April 30, 1993. Although the review concerned imports from eight countries, the judgments on appeal here concern only imports from Japan. Because the review was initiated prior to January 1, 1995, the applicable antidumping law and regulations are those that were in effect prior to the changes made by the Uruguay Round Amendments Act, Pub.L. No. 103-465, 108 Stat. 4809 (1994) (the “URAA”). See URAA § 291(a)(2), (b); Cemex, S.A. v. United States, 133 F.3d 897, 899 n. 1 (Fed.Cir.1998). Under the then-applicable law, the antidumping duty calculated by Commerce in the Final Results and the Remand Results was imposed “in an amount equal to the amount by which the foreign market value exceeds the United States price for the merchandise.” 19 U.S.C. § 1673 (1988). Broadly speaking, “[f]or-eign market value is the price of the merchandise in the producer’s home market or its export price to countries other than the United States.” Thai Pineapple Pub. Co. v. United States, 187 F.3d 1362, 1365 (Fed.Cir.1999); see generally 19 U.S.C. § 1677b (1988). United States price (“USP”) is the price at which the merchandise is purchased from the manufacturer or a reseller for exportation to the United States or the price at which the merchandise is sold or agreed to be sold in the United States for the account of the exporter. See 19 U.S.C. § 1677a (1988).
The parties appealed numerous aspects of the Final Results to the Court of International Trade. In NSK I, the Court of International Trade affirmed the Final Results, other than with respect to thirteen issues which were remanded back to Commerce for redetermination. See NSK I, 969 F.Supp. at 66. Commerce’s subsequent Remand Results were then affirmed in their entirety by the Court of International Trade in NSK II.
In this court, NTN appeals with respect to four issues. Specifically, NTN argues (i) that substantial evidence does not support Commerce’s inclusion in its FMV calculation of certain sales identified by NTN as sample sales and sporadic sales; (ii) that Commerce unreasonably excluded from its calculation of FMV certain of NTN’s home market sales to related parties; (iii) that Commerce in its Remand Results erroneously refused to adjust its [1326]*1326calculation of FMV for NTN’s reported home market discount as direct selling expenses; and (iv) that Commerce’s comparison of sales across different levels of trade in its calculation of FMV is not supported by substantial evidence.
Koyo Seiko appeals only with respect to one issue. In NSK I, the Court of International Trade held that Commerce’s acceptance in the Final Results of Koyo Seiko’s proposed circumstance of sale adjustment to FMV for its home market warranty expenses was erroneous due to the adjustment including warranty expenses on out-of-scope merchandise. Koyo Seiko argues that Commerce’s interpretation of the statute providing for circumstance of sale adjustments to FMV was reasonable, comported with precedent, and thus should have been upheld in NSK I.
Finally, Torrington cross appeals with respect to Commerce’s determination in its Final Results, as affirmed by the Court of International Trade in NSK I, that Honda constitutes a “reseller” as that term is used in the statute governing the calculation of USP and the accompanying definitional regulation. Torrington argues that Commerce unreasonably premised its determination that Honda was a “reseller” upon its finding that Honda’s suppliers neither knew nor had reason to know that their sales to Honda would be exported to the United States.
DISCUSSION
I. Standard of Review
“In reviewing a decision by the Court of International Trade, this court applies anew the statutory standard of review applied by that court to the agency’s decision.” Thai Pineapple, 187 F.3d at 1365. Thus, like the Court of International Trade, we must examine the underlying findings of Commerce to determine if they are “unsupported by substantial evidence on the record, or otherwise not in accordance with law.” 19 U.S.C. § 1516a(b)(l)(B)(i) (1994). While we defer to Commerce’s reasonable interpretation of the antidumping statutes if not contrary to an unambiguous legislative intent, see Timex V.I., Inc. v. United States, 157 F.3d 879, 881-82 (Fed.Cir.1998) (citing Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-43 & n. 9, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984)), and we accord substantial deference to Commerce’s interpretations of its own regulations, see Torrington Co. v. United States, 156 F.3d 1361, 1363-64 (Fed.Cir.1998), we review the Court of International Trade’s interpretation of the antidumping laws de novo, see Cemex, 133 F.3d at 900.
II. NTN’s Appeal
A. The Alleged Sample and Sporadic Sales
Under 19 U.S.C. § 1677b(a)(l)(A) (1988), FMV is the price “at which such or similar merchandise is sold or, in the absence of sales, offered for sale in the principal markets of the country from which exported, ... in the ordinary course of trade for home consumption” (emphasis added). As defined by statute:
The term “ordinary course of trade” means the conditions and practices which, for a reasonable time prior to the exportation of the merchandise which is the. subject of an investigation, have been normal in the trade under consideration with respect to merchandise of the same class or kind.
19 U.S.C. § 1677(15) (1988).
In the Final Results, Commerce included as home market sales in the ordinary course of trade certain sales identified by NTN as sample sales and sales with sporadic sales histories. Commerce reasoned that NTN’s mere designation of certain [1327]*1327sales as sample sales did not satisfy its burden of proving that the sales were made outside of the ordinary course of trade. See Final Results, 60 Fed.Reg. at 10,947. Similarly, Commerce rejected NTN’s claim that certain sales of small quantities of products with sporadic sales histories were outside of the ordinary course of trade, explaining that “such sales histories are typical of certain types of products.” Id.
In NSK I, the Court of International Trade remanded to Commerce for a rede-termination in light of this court’s ruling in NSK, Ltd. v. United States, 115 F.3d 965, 973-75 (Fed.Cir.1997), that samples given without consideration do not constitute sales under the antidumping statute. In its Remand Results, Commerce explained that certain of the transactions labeled by NTN as “sample and other similar transfers” were, in fact, transfers in which monetary consideration was paid for the bearings. Consequently, the Court of International Trade amended its prior ruling in NSK I to order Commerce only to exclude those transactions for which NTN received no consideration. Commerce complied with this instruction on remand and its determinations were sustained by NSK II.
We are not persuaded that substantial evidence does not support Commerce’s determination that NTN failed to meet its burden to prove that the alleged sample and sporadic sales were sales outside of the ordinary course of trade. The evidence submitted by NTN consisted of its records indicating sales of samples under the term “SAMPLEH” and sporadic sales under the term “NORDCH”. The sales designated by NTN as sporadic were simply those in which there were seven or less transactions in the total review period with three or less units per transaction. No evidence was submitted that the process of ordering or shipping these alleged sample or sporadic sales differed from “ordinary” sales. Nor was evidence submitted that any of these sales required, for example, unique engineering specifications. Moreover, no evidence was submitted that the alleged sample sales for which monetary consideration was paid, differed in any material respect, other than by their designation as “SAMPLEH”, from “ordinary” sales of the same merchandise for monetary consideration. Thus, given our deferential standard of review, we cannot say that Commerce’s finding of a failure of proof was unsupported by substantial evidence.
B. The Related Party Sales
Under 19 U.S.C. § 1677b(a)(3) (1988), Commerce may base FMV on related party transactions. To implement this statute, Commerce promulgated a regulation requiring importers to demonstrate that their sales to related parties are comparable in price to unrelated party transactions. Specifically, the regulation provides:
If a producer or reseller sold such or similar merchandise to a person related as described in section 771(13) of the Act, the Secretary ordinarily will calculate foreign market value based on that sale only if satisfied that the price is comparable to the price at which the producer or reseller sold such or similar merchandise to a person not related to the seller.
19 C.F.R. § 353.45(a) (1993).
In the fourth annual administrative review, NTN responded to Section C of Commerce’s questionnaire by listing its sales and annotating them with a “1” or a “2” to indicate whether they were, respectively, related or unrelated party sales. After conducting an “arm’s-length test” using weighted average prices for each class of merchandise to determine whether the prices that NTN charged related parties were representative of market prices, [1328]*1328Commerce excluded the reported related party sales of ball and cylindrical roller bearings due to the related party prices being lower than the corresponding unrelated party prices. See Final Results, 60 Fed.Reg. at 10,946-47. However, Commerce included the reported related party sales of spherical roller bearings due to the reported prices not being less than the corresponding prices to unrelated parties. See id. In NSK I, the Court of International Trade affirmed, reasoning that NTN had presented no evidence that Commerce’s test was unreasonable. See NSK I, 969 F.Supp. at 54-55.
On appeal, NTN argues that Commerce’s use of weighted averages to compare related and unrelated party sales unreasonably distorted the comparison. To illustrate, NTN provides a hypothetical example in which the actual per unit price to both the related and unrelated parties is the same but, because different quantities are sold to the related and unrelated parties, the weighted average price for the goods sold to the unrelated party is higher.
In light of “the substantial discretion accorded Commerce when interpreting and applying its own regulations,” Torrington Co. v. United States, 156 F.3d 1361, 1363 (Fed.Cir.1998), we are unpersuaded that Commerce’s use of weighted averages was unreasonable. See also Smith-Corona Group v. United States, 713 F.2d 1568, 1571, 1 Fed. Cir. (T) 130, 132 (Fed.Cir.1983) (discussing Commerce’s “broad discretion in executing the law”). Commerce’s methodology was reasonable because, without the use of weighting, small sales that might be “outliers” could be given undue weight in the calculation. Without employing such a weighting, Commerce’s approach would be susceptible to the “perceived danger that a foreign manufacturer will sell to related companies in the home market at artificially low prices, thereby camouflaging true FMV and achieving a lower antidumping duty margin.” NEC Home Elecs., Inc. v. United States, 54 F.3d 736, 739 (Fed.Cir.1995). Moreover, although NTN’s hypothetical suggests a potential anomaly of using a weighted average test, NTN has not provided evidence, nor even suggested, that such an anomalous result occurred here. Thus, NTN’s hypothetical does not demonstrate that Commerce’s test was unreasonable as applied to NTN’s actual situation, only, at the very most, to circumstances not here present.
C. The Home Market Discounts
In the Final Results, Commerce determined that NTN’s reported home market discounts would be treated as indirect selling expenses, not direct adjustments to price, because they were not reported on a sale-specific basis. See Final Results, 60 Fed.Reg. at 10,934. In NSK I, the Court of International Trade held that Commerce erroneously treated the home market discounts as indirect selling expenses. See NSK I, 969 F.Supp. at 46. The court reasoned that “they are direct selling expenses because they represent expenses related to particular sales of in-scope merchandise.” Id. However, the court concluded that “they are not entitled to a direct adjustment because of [NTN’s] failure to tie them to specific transactions .... ” Id. Consequently, Commerce denied NTN a direct adjustment to FMV in its Remand Results and this denial was subsequently affirmed by the Court of International Trade in NSK II.
NTN argues on appeal that its home market discounts can be traced directly to specific transactions through its internal accounting system of product classification and recording of discounts. Accordingly, NTN maintains that its reported home market discounts were not pooled or dis-tortive and thus should have qualified for a direct adjustment to FMV.
We hold that there was no legal or factual error in Commerce’s determination [1329]*1329that NTN’s reported home market discounts do not qualify for a direct adjustment to FMV. NTN’s home market discounts were reported on an allocable, not transaction specific, basis. However, as Commerce explained, “[allocated price adjustments have the effect of distorting individual prices by diluting the discounts or rebates received on some sales, inflating them on other sales, and attributing them to still other sales that did not actually receive any at all.” Final Results, 60 Fed. Reg. at 10,929. As demonstrated by NTN’s responses to Commerce, substantial evidence supports not only Commerce’s finding that NTN’s reported home market discounts were reported on a product and customer specific basis, but also Commerce’s implicit finding that NTN’s reporting method was not equivalent to transaction specific reporting.
D. The Comparison of Sales Across Different Levels of Trade
As set forth more fully above, under 19 U.S.C. § 1677b(a)(l)(A) (1988), FMV is the price “at which such or similar merchandise” is sold or offered for sale in the country of exportation for home consumption. The term “such or similar merchandise” is defined in 19 U.S.C. § 1677(16) (1988), which sets forth a tripartite hierarchy for Commerce’s selection of “such or similar merchandise.” Under part (A) of that provision the most preferable category of merchandise is merchandise “identical” to that subject to the investigation in physical characteristics and origin. Under part (B) of section 1677(16), the next most preferable category of merchandise is merchandise produced by the same person in the same country as the merchandise under investigation, but merely “like” that merchandise in terms of components and purpose and “approximately equal in commercial value to that merchandise.” Finally, under part (C) of section 1677(16), the least preferable category of “such or similar merchandise” is merchandise produced by the same person in the same country as the merchandise under investigation, “like” that merchandise in terms of purpose, and which Commerce “determines may reasonably be compared with that merchandise.”
None of these statutory provisions make reference to the term “levels of trade” at which the merchandise is sold. However, this term is. used in Commerce’s implementing regulation, which provides:
The Secretary normally will calculate foreign market value and United States price based on sales at the same commercial level of trade. If sales at the same commercial level of trade are insufficient in number to permit an adequate comparison, the Secretary will calculate foreign market value based on sales of such or similar merchandise at the most comparable commercial level of trade as sales of the merchandise and make appropriate adjustments for differences affecting price comparability.
19 C.F.R. § 853.68 (1993). Under this regulation, Commerce “provide[s] adjustments to account for the difference between wholesale and retail market prices.” Zenith Elecs. Corp. v. United States, 988 F.2d 1573, 1584, 11 Fed. Cir. (T) 57, 68 (Fed.Cir.1993).
Here, Commerce was unable to compare NTN’s United States sales to home market sales at the same level of trade and thus “attempted to find matches at the next most similar level of trade.” Final Results, 60 Fed.Reg. at 10,940. To accomplish this goal, Commerce compared the variable cost of manufacturing of the proposed home market model with that of the variable cost of manufacturing the United States model. Commerce determined that if the difference between these variable manufacturing costs was greater than twenty percent, then the two models were not of “approximately equal commercial [1330]*1330value” and would not be used for price comparison purposes. Ultimately, Commerce denied NTN’s claim for a level of trade adjustment. Commerce rejected NTN’s claim for a level of trade adjustment to account for price differences, explaining that:
In order for the Department to make a level-of-trade adjustment, respondents must quantify any price differences that are attributable to differences in levels of trade. NTN has failed to demonstrate what portion, if any, of those price differences is attributable to differences in levels of trade.
Id. Commerce also rejected NTN’s claim for a level of trade adjustment based upon differences in indirect selling expenses. Commerce reasoned that NTN’s allocation of a common pool of expenses to all sales, irrespective of levels of trade, using relative sales values demonstrated that “such expenses were not unique to, nor disproportionally attributable to, any level of trade.” Id. The Court of International Trade subsequently affirmed Commerce’s denial of NTN’s claimed level of trade adjustment. See NSK I, 969 F.Supp. at 58-54.
Like the Court of International Trade, we are unpersuaded by NTN that Commerce’s denial of a level of trade adjustment was unsupported by substantial evidence. Although NTN submitted evidence that merchandise at different levels of trade had different prices and selling expenses, NTN did not provide evidence to prove that those differences were not caused by other factors, such as volume sold or arbitrary pricing practices. In other words, NTN did not present evidence to establish that the difference in the level of trade caused the differences in price and selling expenses. Moreover, with respect to indirect selling costs, substantial evidence also supports Commerce’s denial of a level of trade adjustment due to NTN’s flawed expense allocation methodology. This methodology was based on several unproven presumptions, including direct correlations between the volume of sales at a particular level of trade and the number of employees at that level and between the number of employees at each level and the quantum of selling expenses at each level. Given that NTN bore the burden before Commerce of establishing its entitlement to a level of trade adjustment, its failure to provide Commerce with the requisite causal evidence compels us to conclude that it has not met its burden on appeal of persuading this court that Commerce’s denial was not supported by substantial evidence.
III. Koyo Seiko’s Appeal
Under the statutory scheme, “due allowance shall be made” by Commerce in its determination of FMV for “differences in circumstances of sale” between sales in the United States and sales in the home market of the exporter. 19 U.S.C. § 1677b(a)(4)(B) (1988). Koyo Seiko claimed a circumstances of sale adjustment to account for its warranty expenses in its home market. Koyo Seiko reported to Commerce a home market warranty expense factor that was calculated by dividing the total warranty expenses incurred on antifriction bearing products during the period of review by the sales value of antifriction bearing products during that period. This warranty factor was not calculated or reported for each model or customer, nor were warranty claims identified for particular transactions. Consequently, the reported warranty factor related, in part, to non-scope merchandise such as tapered roller bearings.
Koyo Seiko’s proposed home market warranty factor was nonetheless accepted by Commerce. Commerce explained:
Although Koyo calculated a warranty expense factor based on the ratio of total warranty claims to total bearing sales, there is no evidence on the record that [1331]*1331the calculated warranty expense factor would vary by class or kind of bearing or by customer. Therefore, as in [the Final Results of the third administrative review], where Koyo used the same allocation methodology, we find that Koyo reasonably allocated direct warranty expenses, and we have accepted them for the final results.
Final Results, 60 Fed.Reg. at 10,910.
Torrington appealed to the Court of International Trade, which ruled that “Commerce improperly included Koyo’s warranty expenses relating to non-scope merchandise in adjusting FMV.” NSK I, 969 F.Supp. at 46. The court reasoned that Commerce was prohibited from using a methodology for the calculation of adjustments to FMV that included warranty expenses on non-scope merchandise. See id. Consequently, the court remanded to Commerce “to review the record to determine whether it is possible to remove those portions of Koyo’s warranty expenses which relate to non-scope merchandise from the adjustments to FMV or to deny the adjustments if such removal cannot be made.” Id.
On remand, Commerce determined that it was not possible to separate Koyo Seiko’s warranty expense information relating to within-seope merchandise from that relating to non-scope merchandise. As a result, Commerce excluded all of Koyo’s reported home market warranty expenses from the pool of permitted adjustments to FMV. The Court of International Trade affirmed this aspect of the Remand Results in NSK II.
On appeal to this court, Koyo Seiko argues that Commerce’s acceptance of the proposed warranty factor was made pursuant to a reasonable interpretation of section 1677b and thus should have been upheld by the Court of International Trade. We agree. We have previously stated that this provision “does not expressly limit the exercise of the Secretary’s authority to determine adjustments, nor does it include precise standards or guidelines to govern the exercise of that authority.” Smith-Corona, 713 F.2d at 1575, 1 Fed. Cir. (T) at 137. Indeed, “the statute does not define the term ‘circumstances of sale’ nor does it prescribe any method for determining allowances. Congress has deferred to the Secretary’s expertise in this matter.” Id. Thus, because “[t]he statute provides no specific guidelines for the treatment of warranty expenses,” Zenith, 988 F.2d at 1583, 11 Fed. Cir. (T) at 67, Commerce’s interpretation of the statute with respect to circumstances of sale adjustments for warranty expenses cannot be disturbed if reasonable, see id., 988 F.2d at 1584, 11 Fed. Cir. (T) at 68 (citing Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 843-44, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984)); see also Torrington, 156 F.3d at 1363-64 (explaining that “substantial deference” is accorded to Commerce’s interpretation of 19 C.F.R. § 353.56 in which Commerce promulgated criteria for the allowance of circumstances of sale adjustments to FMV).
Commerce’s interpretation of the governing statute and regulation to permit an adjustment for home market warranty expenses based on a factor derived, in part, from out-of-scope merchandise was reasonable. Torrington and the government point to precedent from the Court of International Trade holding that “the Court cannot allow [Commerce] to use a methodology which allows the inclusion of ... warranty expenses on out-of-scope merchandise.” Federal Mogul Co. v. United States, 862 F.Supp. 384, 407 (Ct. Int’l Trade 1994); see also Torrington Co. v. United States, 818 F.Supp. 1563, 1578 (Ct. Int’l Trade 1993) (stating that “merchandise which is outside the scope of an antidumping order cannot be used in the calculation of antidumping duties”). However, aside from the fact that such [1332]*1332precedent is not binding on this court, its import simply is not pertinent to the methodology employed by Commerce in the Final Results. Commerce did not grant a circumstances of sale adjustment for the home market warranty expenses of out-of-scope merchandise. Rather, Commerce relied upon both within-scope and out-of-scope merchandise warranty data to calculate an expense factor which it then applied only to the within-scope merchandise. Thus, no adjustment was granted for expenses relating to out-of-scope merchandise.
That the factor was calculated using both within-scope and out-of-scope merchandise was entirely reasonable. As a general matter, warranty data, by its very nature, can be difficult to tie to precise transactions. Indeed, when examining a recent sales period, there is plainly no way of predicting precisely how many customers will exercise their warranty rights in the future. Moreover, internal corporate accounts recording warranty expenses might be presumed unlikely to be categorized according to the ex post facto scope classifications of Commerce’s antidumping duty orders. For these reasons, Commerce has been content, as a matter of policy, to use “warranty costs incurred during the period of investigation or historical warranty data incurred over a longer period as best information for warranty costs associated with sales made during the period of investigation.... ” Study of Antidumping Adjustments Methodology and Recommendations for Statutory Change 39 (Dep’t Commerce 1985). More recently, the Court of International Trade recognized the reasonableness of calculating a warranty factor using both within-scope and out-of-scope merchandise in Koenig & Bauer-Albert v. United States, 15 F.Supp.2d 834 (Ct. Int’l Trade 1998). In Koenig, the court rejected the respondent’s argument that Commerce improperly applied a United States sales warranty factor based upon both United States sales and out-of-scope Canadian sales. The court reasoned that it was within Commerce’s discretion to rely upon warranty data relating in part to out-of-scope merchandise, when that data could not be segregated, in order to calculate a factor to apply to within-scope merchandise that would be as accurate as possible. See id. at 46.
We agree with Koenig’s analysis and hold that it is within Commerce’s discretion to calculate a warranty expenses factor for purposes of determining a circumstances of sale adjustment, using data for that factor that relates in part to out-of-scope merchandise, provided that Commerce is satisfied that (i) the out-of-scope merchandise data cannot be segregated from the data relating to within-scope merchandise or that such segregation would be distortive; (ii) there is no evidence that the warranty expenses incurred on within-scope and out-of-scope merchandise are likely to differ; and (iii) the reliance on data relating to out-of-scope merchandise comports with “the basic purpose of the statute — determining current margins as accurately as possible,” Rhone Poulenc, Inc. v. United States, 899 F.2d 1185, 1191, 8 Fed. Cir. (T) 61, 67 (Fed.Cir.1990). See generally 19 C.F.R. §§ 353.56(a), 353.54 (1993) (requiring Commerce to “limit allowances to those circumstances which bear a direct relationship to the sales compared” and setting forth that the claimant “must establish the claim to the satisfaction of the Secretary”).
Commerce’s partial reliance here on data relating to out-of-scope merchandise comported with these requirements. Commerce expressly found in the Remand Results that the data that it used to calculate the warranty expenses factor could not be segregated into data relating only to within-scope or out-of-scope merchandise. In the Final Results, Commerce [1333]*1333expressed itself satisfied that “there is no evidence on the record that the calculated warranty expense factor would vary by class or kind of bearing or by customer.” Id., 60 Fed.Reg. at 10,910. Finally, Commerce’s very acceptance of Koyo Seiko’s proposed warranty expenses factor in its Final Results evidences Commerce’s determination that the goal of an accurate dumping margin calculation was furthered by its use. Consequently, we reverse the Court of International Trade with respect to its disallowance of Koyo Seiko’s warranty expenses factor permitted by Commerce in the Final Results.
IV. Torrington’s Cross-Appeal
Torrington cross-appeals Commerce’s determination, affirmed by the Court of International Trade, to revoke the antidumping duty order with respect to Honda due to its finding that Honda had satisfied the requisite three-year period of no sales (or de minimis sales) at less than fair value. Specifically, Torrington argues that Commerce’s calculation of USP with respect to Honda’s imports was based on an impermissible construction of the governing statute.
Under 19 U.S.C. § 1673 (1988), Commerce is required to calculate the applicable antidumping duty as the amount, if any, by which FMV exceeds USP. The statutory scheme then sets forth the methodology for calculating USP:
(a) United States price
For purposes of this subtitle, the term “United States price” means the purchase price, or the exporter’s sale price, of the merchandise, whichever is appropriate.
(b) Purchase price
For purposes of this section, the term “purchase price” means the price at which merchandise is purchased, or agreed to be purchased, prior to the date of importation, from a reseller or the manufacturer or producer of the merchandise for exportation to the United States....
(c)Exporter’s sales price
For purposes of this section, the term “exporter’s sales price” means the price at which merchandise is sold or agreed to be sold in the United States, before or after the time of importation, by or for the account of the exporter, as adjusted under subsections (d) and (e) of this section.
19 U.S.C. § 1677a (1988) (emphasis added). Commerce has defined by regulation the statutory term “reseller” to mean “any person (other than the producer) whose sales [Commerce] uses to calculate foreign market value or U.S. price, including the foreign reseller or exporter.” 19 C.F.R. § 353.2(s) (1993).
In the Final Results, Commerce determined that Honda was a reseller. See Final Results, 60 Fed.Reg. at 10,951-52. Honda is a Japanese corporation engaged in the business of manufacturing and selling motorized products for worldwide sale. Honda also sells replacement and original parts, including antifriction bearings. Honda does not manufacture antifriction bearings, but rather purchases them from Japanese manufacturers and then utilizes them in its products or else resells them to customers in Japan, the United States, and elsewhere. Honda sells its replacement parts in the United States exclusively to a California affiliate, which then resells them to unrelated dealers. Honda also sells a number of original parts to affiliated Ohio and North Carolina corporations.
As a part of the fourth administrative review, Commerce sent Honda a questionnaire, which asked for, inter alia, (1) a list of “all firms that supplied you with bearings”; and (2) information on whether “your suppliers have reason to know that you purchased [antifriction bearings] for [1334]*1334export to the United States,” based on such identifiers as the “product numbers or codes,” the “physical characteristics of the merchandise,” different shipment locations, different packing, or “any other factors, such as the timing of shipments, the terms of sale, the timing of orders.” Honda responded with a list of suppliers and explained that its orders did not indicate by way of the above identifiers, or any other identifiers, the ultimate destination of the antifriction bearings. Commerce officials conducted a verification in Japan and reported that the ultimate destination of parts was not indicated in any way, but rather that the parts went into a general inventory. In addition, the report noted that Honda products are exported to over twenty countries and that all parts exported to the United States are also offered for sale in Japan.
In early 1994, Commerce filed its preliminary results, finding de minimis margins, and thus preliminarily determined to revoke the antidumping order with respect to Honda because Honda would have satisfied the three year period of no sales at less than fair value required to revoke such an order. See 19 C.F.R. § 353.25(a) (2)(i) (1993). In the briefing that followed, Torrington argued that Honda’s sales should be treated as purchase-price transactions, rather than treating Honda as a reseller. However, in its Final Results, Commerce rejected this argument, noting that “Torrington acknowledges that the Department found no evidence at verification that Honda suppliers were aware of the ultimate destination of their merchandise.... ” 60 Fed.Reg. at 10,951. Commerce concluded:
We agree with Honda that it should be treated as a reseller. This issue was examined extensively at verification. The standard for the “knowledge test” is high. Based on this standard, we concluded that Honda’s suppliers did not have reason to know that their sales to Honda would be exported to the United States. Therefore, we continue to classify Honda as a reseller.
Final Results, 60 Fed.Reg. at 10,951-52 (citations omitted). The Court of International Trade affirmed this determination, explaining that “[t]he legislative history to [section 1677a] clearly demonstrates that Commerce’s knowledge test was anticipated by Congress and is a reasonable interpretation of the statute.” NSK I, 969 F.Supp. at 60.
On appeal, Torrington argues that Commerce should have determined USP by relying upon the transactions between Honda and its Japanese suppliers, not Honda’s resales of antifriction bearings in the United States. Torrington contends that the “knowledge test” is only applicable where the sale involves a middleman, such as a trading company, not, as here, a manufacturer purchasing bearings primarily for its own uses. Moreover, Torrington maintains that Honda’s suppliers must have been constructively aware that their bearings would end up in the United States, given the significant proportion of Honda products manufactured in the United States or shipped to the United States.
We are not persuaded by Torrington’s arguments. While Torrington cites to cases in which trading companies have been found to be resellers under the statute, this does not imply that the term “reseller” is limited to such trading companies. In any event, given that Honda admittedly does not manufacture antifriction bearings, but rather purchases them from manufacturers and then uses them itself or else resells them, it is not apparent how Honda’s role in those transactions is any different than that of a trading company or middleman. The “knowledge test” applied by Commerce to determine whether Honda was a reseller also accords with the legislative intent. Indeed, the Statement of Administrative Action submitted in con[1335]*1335nection with the enactment of the Trade Agreements Act of 1979, Pub.L. No. 96-39, 93 Stat. 144 (1979), stated with respect to the definition of “purchase price” that:
The definition makes clear that if the producer knew or had reason to know the goods were for sale to an unrelated U.S. buyer, and the terms of the sales were fixed ... the producer’s price will be used as a ‘purchase price’ to be compared with that producer’s foreign market value.
H.R. Doc. No. 153, Pt. 2, 96th Cong., 1st Sess. at 411-12 (1979), reprinted in, 1979 U.S.C.C.A.N. 681-82. Similar statements appear in the Senate and House Committee reports. See S.Rep. No. 249, 96th Cong., 1st Sess. at 94 (1979); H.R.Rep. No. 317, 96th Cong., 1st Sess. at 75 (1979). Thus, we cannot say that Commerce’s definition of the term “reseller” was unreasonable or at variance with the legislative intent.
Finally, with respect to Commerce’s application of this test, substantial evidence supports Commerce’s finding that Honda’s suppliers had no reason to know whether their individual sales transactions would cause antifriction bearings to be exported to the United States. While the supplier might have reasonably known that a minority of all antifriction bearings purchased by Honda would end up in the United States, there was more than adequate evidence that the individual suppliers could not have known whether any of their particular bearings would be among those so exported. This evidence included that Honda consigned its purchased anti-friction bearings into a general inventory from where they were shipped, as required, to destinations in Japan and abroad. Commerce’s verification found no shipping markings or peculiarities with regard to the timing of orders that could have revealed that particular antifriction bearings were intended for shipment to the United States. Thus, we must affirm the Court of International Trade because substantial evidence supports Commerce’s reasonable interpretation of the statute.
CONCLUSION
We affirm the judgment of Court of International Trade with respect to NTN’s appeal because we detect no reversible legal or factual error with respect to Commerce’s calculation of FMV in its determination of NTN’s dumping margin. In Koyo Seiko’s appeal, however, we reverse the judgment of the court because Commerce’s acceptance in the Final Results of Koyo Seiko’s home market warranty expense factor was a reasonable interpretation of the statute providing for circumstance of sale adjustments to FMV. Finally, we affirm with respect to Torring-ton’s cross appeal because Commerce reasonably interpreted the term “reseller” in the antidumping statute and substantial evidence supports its determination that, in calculating Honda’s USP, Honda was a reseller with regard to its sales of subject antifriction bearings. The judgment of the Court of International Trade is therefore
AFFIRMED-IN-PART and REVERSED-IN-PART.
COSTS
Each party to bear its own costs.