Full opinion text
OPINION POGUE, Judge. Plaintiffs Koenig & Bauer-Albert AG (“KBA”) and MAN Roland Druekmaschinen AG and MAN Roland Inc. (“MAN Roland”), respondents in the underlying investigation, and Plaintiff Goss Graphic Systems, Inc. (“Goss”), petitioner in the underlying investigation, filed separate motions challenging various aspects of the determination of the International Trade Administration of the United States Department of Commerce (“Commerce” or “ITA”) regarding imports of large newspaper printing presses (“LNPP”) from Germany. Large Newspaper Printing Presses and Components Thereof, Whether Assembled or Unassembled, from Germany, 61 Fed.Reg. 38,166 (Dep’t Commerce 1996) (final det.)(“Germany Final”). The motions were consolidated. The antidumping investigation of LNPPs from Germany was conducted simultaneously with Commerce’s investigation of sales of LNPPs from Japan. Issues common to both investigations were discussed in Germany Final. The Court affirmed Commerce’s determinations with respect to common issues of scope and standing. Mitsubishi Heavy Indus. v. United States, 986 F.Supp. 1428 (1997). Familiarity with the Court’s opinion on scope and standing issues is presumed. MAN Roland challenges Commerce’s refusal to accept amendments to the contract prices of two presses; Commerce’s circumstance of sale adjustment for imputed credit expenses; Commerce’s allocation of indirect selling expenses incurred by MAN Roland’s U.S. subsidiary, MRU; Commerce’s decision to rely on facts available in lieu of cost data supplied by MAN Roland; Commerce’s choice of facts available; Commerce’s choice of a variance to adjust MAN Roland’s estimated overhead expenses to approximate actual expenses; Commerce’s refusal to average MAN Roland’s costs with those of MAN Roland’s wholly owned subsidiary; Com-meree’s decision to include home-market sales with “abnormally high profits” in its profit calculation for constructed value; Commerce’s treatment of certain U.S. sales as constructed export price (“CEP”) sales; and Commerce’s decision to treat MAN Roland’s installation costs as further manufacturing costs. See Mem. Pis. MAN Roland Druekmaschinen AG and MAN Roland Inc. Support Mot. J. Agency Record Company- and Country-Specific Issues at 3-4 (“MAN Roland brief’). Goss challenges four aspects of Commerce’s final determination. Two of Goss’s objections, regarding Commerce’s decision to deduct imputed interest from normal value and Commerce’s allocation of indirect selling expenses incurred in Germany and Japan, are common to both the German and the Japan investigations. These are discussed in Mitsubishi Heavy Indus. v. United States, slip op. 98-82, 1998 WL 417423, 15 F.Supp. 807 (CIT June 23, 1998). With regard to the Germany investigation, Goss objects to Commerce’s allocation of costs for two of MAN Roland’s U.S. sales, and Commerce’s inclusion of Canadian warranty expenses in estimating MAN Roland’s U.S. warranty expenses. Brief Support Goss Graphic Sys., Inc. Rule 56.2 Mot. J. Agency Record Comp./Country Specific Issues (“Goss brief’). KBA challenges Commerce’s choice of facts available in calculating its dumping margin. See Brief of Pis. Koenig & BauerAl-bert AG and KBA-Motter Corp. Support Mot. J. Agency Record (KBA brief). DISCUSSION I. MAN Roland’s Price Ajwendments Commerce calculated MAN Roland’s dumping margin based on sales of two complete Geoman presses, to The Rochester Chronicle and Democrat (“Rochester”) and The Times Leader of Wilkes Barre (‘Wilkes Barre”) and two sales of German components for U.S.-made presses or additions. A fifth sale, of parts and subcomponents, was excluded from Commerce’s final analysis because Commerce determined that it was outside the scope of its investigation. See Germany Final at 38,172. Several months after Goss filed its anti-dumping petition, MAN Roland amended its contracts for the Rochester and Wilkes-Barre presses. In the final determination, Commerce refused to accept the price adjustments and calculated U.S. price based on the original contract prices of the two sales. MAN Roland challenges this decision. According to the statute, United States price is to be based on “the price at which the subject merchandise is first sold (or agreed to be sold)” to an unaffiliated purchaser in the United States. 19 U.S.C. § 1677a(a), (b) (1994). The statute does not discuss the treatment of price amendments made after the period of investigation. Therefore, the Court must defer to Commerce’s interpretation of the statute if that interpretation is reasonable. See Chevron, U.S.A., Inc. v. Natural Resources Defense Council, 467 U.S. 837; 843, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984) (“If, ... the court determines Congress has not directly addressed the precise question at issue, the court does not simply impose its own construction on the statute,.... Rather, if the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency’s answer is based on a permissible construction of the statute.”) (footnotes omitted); see also Koyo Seiko Co. v. United States, 36 F.3d 1565, 1570 (Fed.Cir.1994) (“[A] court must defer to an agency’s reasonable interpretation of a statute even if the court might have preferred another.”). MAN Roland argues that Commerce’s refusal to accept the price amendments was inappropriate because “[tjhere was no evidence in this case that the prices were being manipulated to avoid a dumping finding.” MAN Roland brief at 5. However, MAN Roland’s implicit statement of the relevant legal standard is inaccurate. Commerce was not looking for evidence of manipulation. Commerce’s rejection of the price amendments was based upon the potential for manipulation, not evidence of actual manipulation. In past cases, the Department [Commerce] has stated that its* standard practice is not to accept price adjustments instituted after the filing of a petition.... [W]e have held that we are cautious in accepting price increases ... so as to discourage potential manipulation of potential dumping margins, and have determined the original contract price which pre-dated the filing of the petition as the proper basis for U.S. price. Germany Final at 38,181. Commerce’s decision to reject price amendments that present the potential for price manipulation was a permissible interpretation of the statute. See Mitsubishi Elec. Corp. v. United States, 12 CIT 1025, 1046, 700 F.Supp. 538, 555 (1988) (“The ITA has been vested with authority to administer the antidumping laws in accordance with the legislative intent. To this end, the ITA has a certain amount of discretion [to act] ... with the purpose in mind of preventing the intentional evasion or circumvention of the anti-dumping duty law.”), aff'd 8 Fed. Cir. (T) 45, 898 F.2d 1577 (1990); see also Dastech Int'l, Inc. v. ITC, 21 CIT -, 963 F.Supp. 1220, 1229 (1997) (“Post-petition pricing changes may be suspect because of the possibility of posturing to promote the outcome a party desires.”). Commerce’s decision also was supported by substantial evidence. The price amendments themselves, made after the initiation of the investigation, constitute evidence that manipulation may occur, as does the fact that Commerce was unable to verify the amendments. See Def.’s Mem. Opp. Mot. MAN Roland Druckmasehinen AG and MAN Roland Inc. J. Agency Record Country and Company-Specific Issues at 7-8 (“Commerce brief’) (“ITA could not conclude that MRD’s amendment was bona fide, where the amendment pertained to a contract in connection with an unfinished press, the payment of which would not occur (and, thus, could not be verified) until after the investigation had ended.”). MAN Roland also argues that because “there was no factual basis for concluding that the price amendments at issue ... were in any way fictitious or manipulated ... the Department’s determination can only be sus-tamed if there is, in fact, a per se rule requiring the Department to automatically disregard any prices negotiated after the petition was filed.” MAN Roland brief at 7. MAN Roland concludes that Commerce does not operate under such a rule and therefore, Commerce erred in excluding the post-petition price amendments. MAN Roland is mistaken. The Court need not find that Commerce has a per se rule in order to uphold its decision. As explained above, Commerce’s rejection of MAN Roland’s price amendments was in accordance with law and supported by substantial evidence. Therefore, the Court will not disturb Commerce’s decision. See 19 U.S.C. § 1516a(b)(l) (“The court shall hold unlawful any determination, ... found ... to be unsupported by substantial evidence on the record, or otherwise not in accordance with law_”). II. Imputed Interest Expense In calculating both normal value and U.S. price, Commerce made a circumstance-of-sale adjustment for imputed interest expenses incurred during production of the subject merchandise. Commerce’s usual imputed interest calculation is based only on the cost of financing receivables between shipment date and payment date. See Germany Final at 38,187. MAN Roland contends that Commerce should have treated imputed interest expense incurred prior to shipment as a cost of production. Commerce chose to treat the imputed interest as a circumstance of sale, MAN Roland argues, only because the statute would not permit Commerce to deduct imputed interest from the cost of production. “[Ujnder both German and U.S. accounting principles, capitalized interest could not have been included in the cost of MAN Roland’s LNPPs. And, because the statute incorporates those accounting principles, capitalized interest could not have been included in the constructed value....” MAN Roland brief at 10-11. Commerce does not dispute MAN Roland’s assertion that German generally accepted accounting principles (“GAAP”) would not allow imputed interest to be included in cost of production. Furthermore, in some instances, Commerce is required to calculate costs “based on the records of the exporter or producer of the merchandise, if such records are kept in accordance with the generally accepted accounting principles of the exporting country ... and reasonably reflect the costs associated with the production and sale of the merchandise.” 19 U.S.C. § 1677b(f)(l)(A). However, that requirement does not apply to the circumstance of sale provision. See id. Thus, the issue here is whether Commerce had the discretion to treat the imputed interest expense as a circumstance of sale rather than part of the cost of production. The statute requires that constructed value be “increased or decreased by the amount of any difference ... between the [U.S. price] and [constructed value] ... that is established to the satisfaction of the administering authority to be wholly or partly due to ... differences in circumstances of sale.” 19 U.S.C. § 1677b(a)(6)(C). The statute does not define the term circumstance of sale. Thus, Commerce has broad discretion in determining what constitutes a circumstance of sale. See Sawhill Tubular Div. Cyclops Corp. v. United States, 11 CIT 491, 497, 666 F.Supp. 1550, 1555 (1987) (“It is ... clear that Congress had deferred to the expertise of the agency and vested broad discretion in it to make adjustments for the differences in the circumstances of sale.”). That discretion, however, is limited. “One recognized limitation upon this power is that adjustment be made for those factors and conditions which have a direct bearing on or relationship to the sales under consideration.” Id. In the final determination, Commerce explained, We believe that it is appropriate in this instance to recognize the comprehensive financing arrangement for each sale as a circumstance of sale adjustment. LNPPs require substantial capital expenditures over an extended time period because of their size and lengthy production process .... Moreover, the projects generally call for the purchaser to provide scheduled progress payments before completion of a project. Our normal imputed credit calculation ... does not measure the effect of progress payments made relative to production costs incurred. To adjust sales prices for the effect of the respondent incurring significant capital outlays at the beginning of a project ... or receiving large sums of money up front ... we calculated imputed credit for each home market and U.S. sale.... 61 Fed.Reg. at 38,187. Commerce calculated imputed interest by subtracting progress payments from capital outlays made during the construction period and then multiplying the balance by the applicable interest rate. See Mem. Re: Constructed Value, Further Manufacturing and Constructed Export Price Adjustments for Final Determination, N-P Doc. 107 at 4. Commerce’s treatment of the imputed interest as a circumstance of sale allowed it to fully recognize the complexity of the financing arrangements at issue here. Given that construction on a single sale in this case may take several years to complete, as well as the fact that financing arrangements may vary widely from sale to sale, the Court finds Commerce’s decision to treat the imputed interest expenses as a circumstance of sale was reasonable. MAN Roland argues that Commerce’s actions here were inconsistent with its actions in Zenith Elec. Corp. v. United States, 18 CIT 870 (1994). In Zenith, Commerce made a circumstance of sale adjustment for respondent’s imputed credit expense. However, Commerce refused to offset the imputed credit expense — based on “the average duration of accounts receivable,” — by “the average duration of accounts payable.” Id. at 875. Commerce explained, by allowing the customer a period of time to pay, the seller effectively reduces the sales price of the merchandise because of the time value of money.... To the extent that a manufacturer can delay paying its suppliers, the cost of materials is reduced by the time value of money, resulting in a saving in the cost of production. Since accounts payable are, thus, production costs, they cannot result in a cireum-stanee-of-sale adjustment. Id. at 875, n. 8. Commerce explained further, that it could not treat accounts payable as a circumstance of sale adjustment because doing so “would require us to adjust for factors relating to cost of production, which are unrelated to the sales at issue.” Id. In this ease, unlike Zenith, Commerce made the imputed interest adjustment to compensate for different payment schedules by customers to respondent. Thus, the adjustment was related directly to the sales at issue and appropriately treated as a circumstance of sale. III. Allocation of INDIRECT Selling Expenses In calculating United States indirect selling expenses for MAN Roland’s affiliated agent in the United States, MRU, Commerce divided total United States indirect selling expenses incurred on sales of LNPPs during the period of investigation, by the value of United States sales realized — i.e., shipped and invoiced — during the same period. This rate was applied to the subject sales (sales made but not necessarily realized during the period of investigation) to compute the indirect selling expenses attributable to each such sale. MAN Roland argues that Commerce’s methodology overstated MAN Roland’s indirect selling expenses. Just prior to the period of investigation, MAN Roland explains, the LNPP industry experienced a period of depressed sales. Because MAN Roland had a below-average number of sales made just before the period of investigation, it had a below-average number of sales realized during the period of investigation. MAN Roland brief at 15. MAN Roland’s period of sales depression ended at the end of 1993, so that during the period of investigation, which lasted from July 1993 to June 1995, MAN Roland’s sales made returned to normal levels. Id. The great difference between the value of the orders [sales made] MAN Roland received during the period and the value of the shipments [sales realized] ... during the same period resulted in a serious distortion in the Department’s calculation of the indirect selling expense adjustment. In simple terms, the Department took the indirect selling expenses incurred by MRU during the two-year period and divided them by the very low level of shipments during the period to calculate a very high indirect selling expense rate. It then applied this high expense rate to the high level of sales ordered during the period. Id. at 15-16. MAN Roland contends that in calculating the indirect selling expense rate, Commerce should have divided total indirect selling expenses by the value of sales made during the period of investigation, rather than by the value of sales realized. “As a matter of simple arithmetic, the denominator used to calculate an expense rate must be determined on the same basis as the figures to which the rate will be applied.” Id. at 16. The statute requires that Commerce deduct from CEP all selling expenses, including indirect selling expenses, defined as “any selling expenses” not deducted as commissions, direct selling expenses, or selling expenses that the seller pays upon behalf of the purchaser. 19 U.S.C. § 1677a(d)(l)(D). According to the Statement of Administrative Action to the URAA, indirect selling expenses are expenses that “would be incurred by the seller regardless of whether the particular sales in question are made, but reasonably may be attributed (at least in part) to such sales.” H.R. Doc. No. 103-316 at 824 (1994) (“SAA”). Neither the statute nor the SAA explains how indirect selling expenses are to be allocated. Therefore, the Court must accept Commerce’s methodology if that methodology is reasonable. In order to calculate an allocation ratio for indirect selling expenses, Commerce needed to divide MAN Roland’s total indirect selling expenses for the period of investigation, by the total value of the sales for which those expenses were incurred. However, indirect selling expenses cannot be tied to specific sales, so Commerce had to choose an appropriate pool of sales to serve as a denominator. Commerce normally calculates the indirect selling expense ratio by dividing indirect selling expenses by the sales revenue realized during the period of investigation. Preliminary Det. Concurrence Mem., N-P Doc. 70 at 10-12 (“Concurrence Mem.”). The decision to do so here was within Commerce’s discretion and in accordance with law. Sales revenue realized represents a verifiable pool of sales for which expenses have actually-been incurred. MAN Roland also argues that Commerce’s allocation methodology was inconsistent with Sweaters from Taiwan, 55 Fed.Reg. 34,585, 34,596 (Dep’t Commerce 1990)(final det.). In that case, Commerce declined to allocate indirect selling expenses over sales realized and instead allocated them over purchase orders received (sales made). However, in Sweaters from Taiwan, Commerce found a correlation between the indirect expenses incurred during the period of investigation and the sales ordered during the investigation. See id. at 34,596 (“The Department has determined that it is most appropriate to allocate selling expenses over the value of sales for which such expenses were incurred.”); see also Concurrence Mem. at 12 (“In Sweaters from Taiwan, there was a close correlation between POI sales orders and POI selling expenses.”). In this case, MAN Roland has provided no evidence to indicate that a correlation exists between the sales ordered during the period of investigation and the indirect selling expenses incurred during that period. Furthermore, Commerce found that for LNPPs, “sales efforts last for years and yield only large sales at irregular intervals_” Germany Final at 38,183; see also Concurrence Mem. at 12 (“In the LNPP industry due to the significant lead time involved among the different phases of the sales process (ie., sales negotiation, sales contract, production, shipment and the receipt of sales revenue for accounting purposes), such correlation does not exist.”). For the above reasons, the Court finds, Commerce’s allocation of MRU’s indirect selling expenses was consistent with its previous determinations, supported by substantial evidence, and in accordance with law. IV. Constructed Value Issues FOR MAN ROLAND 1. Commerce’s Rejection of MAN Roland’s Submitted Cost Data In calculating normal value for the subject sales, Commerce relied on constructed value, pursuant to 19 U.S.C. § 1677b(4). During the investigation, MAN Roland requested that, for purposes of the preliminary determination, Commerce accept projected costs of production for the Wilkes-Barre and Rochester presses because they were still in the production process, and actual costs were not yet available. Commerce normally accepts only actual costs of producing subject merchandise. Germany Final at 38,186. The cost accounting system used by MAN Roland to generate cost projections for the two presses, the ADK system, was described by MAN Roland as follows: [t]o produce a press, MAN Roland must produce a project-specific bill of materials that specifies the materials to be used in the production and provides detailed work instructions for the workers involved in the production. MAN Roland has a “project specific standard costing system” (the “ADK” system) which can calculate costs precisely based on those detailed work instructions. The ADK system calculates standard costs based on the project-specific bill of materials by (1) valuing materials based on actual prices for the materials and (2) valuing labor by multiplying the standard labor time for each production operation by the actual wage rate of the worker responsible for performing the operation. MAN Roland brief at 21-22. In its decision to include the two U.S. sales, despite the fact that they would not be completed by the date of the preliminary determination, Commerce “relied on MAN Roland’s representations that project-specific standard costs ... would be available for use in the preliminary determination, and that we would be able to compare these costs against substantial actual costs at verification and adjust them accordingly.” Sales Exclusion Mem., N-P Doc. 68 at 4. In the final determination, Commerce rejected MAN Roland’s projected costs and relied on information available to calculate CV for the incomplete presses. MAN Roland objects to Commerce’s rejection of its submitted CV information. MAN Roland argues that “[a]t bottom, the Department decided to reject the costs generated [by MAN Roland’s standard costing system] because those costs were not verified. But the fault in that regard was not MAN Roland’s. Those costs were accurate and fully verifiable.” MAN Roland Brief at 23. MAN Roland suggests two methods by which Commerce could have verified the costs generated by its cost accounting system. “[T]he Department could have compared the actual costs and the [projected] costs for similar presses produced for the home market.... Alternatively, the Department could have compared ‘the reported costs to the actual costs for the completed elements of the presses at verification.’ ” Id. at 25 (citing Feb. 23 Mem. from team to Richard More-land re: sales exclusion issues at 4, N-P Doc. 68). However, MAN Roland contends, “[t]he Department’s verifier simply refused to examine this information. Apparently believing (incorrectly) that the [projected] costs were subject to manipulation because they had been generated after the initiation of the case, she refused to look further.” Id. at 26. “Congress has afforded ITA a degree of latitude in implementing its verification procedures.” Floral Trade Council v. United States, 17 CIT 392, 399, 822 F.Supp. 766, 772 (1993). “The decision to select a particular method of verification rests solely within the agency’s sound discretion.... If a reasonable standard is applied and the verification is supported by substantial evidence, the court will sustain the methodology.” Id. (citing Hercules, Inc. v. United States, 11 CIT 710, 726, 673 F.Supp. 454, 469 (1987)). In this ease, Commerce verifiers conducted a physical inspection of MAN Roland’s production plant and reviewed source documents to verify MAN Roland’s questionnaire responses for numerous elements necessary to calculate constructed value. See Mem. Re: Verification of Constructed value (“CV”) and Further Manufacturing (“FM”) Data, May 16, 1996, N-P Doc. 95 (“Cost Verification Report”). The documents examined by Commerce included MAN Roland’s work-in-process (“WIP”) inventory, where MAN Roland records production costs during the manufacturing phase; MAN Roland’s work orders for individual presses; and MAN Roland’s audited financial statements, which Commerce reconciled with the WIP inventory. Id. Based on the description above, excerpted from Commerce’s verification memorandum, the Court finds, Commerce’s verification procedures were reasonable and adequate and therefore, in accordance with law. See Floral Trade Council, 17 CIT at 399, 822 F.Supp. at 772 (finding Commerce’s verification to be adequate where Commerce “conducted] a physical inspection of [respondent’s] production areas, and [reviewed] source documents_”). Furthermore, Commerce’s decision to reject MAN Roland’s projected costs was supported by substantial evidence. At verification, Commerce discovered that although MAN Roland prepares a project-specific bill of materials for each LNPP, in the normal course of business, MRD does not prepare project-specific standard workplans [i.e., apply the ADK system] for each LNPP that it manufactures. According to company officials, the detailed, project-specific workplans are only prepared for new press models or for presses with unique or special features. MRD officials also indicated that the workplans for the Rochester and Wilkes-Barre projects had been prepared ... solely for the purpose of providing CV information in this ease. CV Verification Report at 2. The fact that these costs were generated specifically for purposes of the investigation supports Commerce’s conclusion that the submitted costs were not independently reliable. See SAA at 834-35 (“In determining whether a company’s records reasonably reflect costs, ... Commerce ... will consider whether the producer historically used its submitted cost allocation methods to compute the cost of the subject merchandise prior to the investigation ... and in the normal course of its business operation.”). In addition, Commerce’s verifier found that MAN Roland does not reconcile actual costs recorded in its WIP inventory, to the estimated costs in its project-specific workplan. Thus, company officials could not identify for us those parts of the Rochester and Wilkes-Barre presses that had been completed as of December 31, 1995 (and for which actual cost data could be reported) and those that remained unfinished as of that date (and would therefore require costs to be reported based on the project-specific standard workplan).... Although MRD reported that the Rochester and Wilkes-Barre sales were 60 and 81 percent complete, respectively, ... these figures were based on a comparison of project costs recorded in the company’s WIP inventory account to total standard work-plans. Thus, ... MRD did not report actual costs incurred to date for the specific projects. Rather, the reported COP and CV figures for these projects were based solely on costs from the company’s standard work plans,.... Cost Verification Report at 2-3. Commerce’s determination, that it was unable to verify MAN Roland’s submitted cost data was in accordance with law and supported by substantial evidence. Therefore, the Court finds, Commerce’s decision to rely on information available was appropriate. See 19 U.S.C. § 1677e. 2. Commerce’s Choice of Facts Available Rather than relying upon MAN Roland’s submitted costs, Commerce used, as facts available, the original estimated costs submitted by MAN Roland, adjusted for the difference, or variance, between estimated and actual costs for a single home market sale designated as HM4. N-P Doc. 107 at worksheet 4. MAN Roland argues that Commerce’s decision to adjust MAN Roland’s estimated costs based on the variance of a single sale was arbitrary and impermissibly inflated MAN Roland’s costs. Commerce is authorized to use the facts available if “an interested party or any other person ... provides ... information but the information cannot be verified.... ” 19 U.S.C. § 1677e(a)(2)(D). The statute does not define facts available, thus Congress has left the choice of facts to Commerce’s discretion. According to the SAA, Commerce need not “prove that the facts available are the best alternative information. Rather, the facts available are information or inferences which are reasonable to use under the circumstances.” SAA at 869. The relevant case law, requires that the choice of facts available bear a rational relationship to the subject matter at issue. See Manifattura Emmepi S.p.A. v. United States, 16 CIT 619, 624, 799 F.Supp. 110, 115 (1992)(“[Com-merce’s] authority to select best information otherwise available is subject to a rational relationship between data chosen and the matter to which they are to apply.”). In this case, Commerce used cost estimates that MAN Roland prepares for every press at the time the customer signs the final agreement. Cost Verification Report at 8. The cost estimates submitted for the Rochester and Wilkes-Barre sales were prepared by MAN Roland in the ordinary course of business, prior to the initiation of this case. Germany Final at 38,186. These cost estimates were adjusted using the actual variance for a contemporaneous home market sale of the' same model press. Id. at 38,186-87. Thus, the Court finds, Commerce’s choice of facts available bore a rational relationship to the actual costs of the presses at issue. MAN Roland states that Commerce received information on all of MAN Roland’s completed LNPP projects since 1990 and that Commerce should have based its analysis on evidence provided for all of these sales rather than a single one. MAN Roland Brief at 26-27. However, as Commerce explained, it chose a single sale, HM4 because it was the only “fully-completed Geoman sale for which ITA had both detailed cost estimate sheets and actual costs from MRD’s cost accounting system.” N-P Doc. 39, at App. 17. The detailed cost estimate sheets were necessary, Commerce explains, because they “provided the information necessary to subtract costs included in the total estimate from actual costs incurred for completed projects.” Commerce brief at 29 n. 17. Furthermore, it was reasonable for Commerce to rely only on completed press sales. MAN Roland estimated costs for complete projects. Commerce needed to compare MAN Roland’s estimates with actual costs in order to establish the variance between estimated costs and actual costs. Commerce would not make this comparison using incomplete projects because the actual cost figures for such projects also would be incomplete. Therefore, Commerce would not know which part of the estimate to compare to the available actual costs. Thus, Commerce’s decision to rely on variance information from a single sale was reasonable. Commerce’s choice also was supported by substantial evidence. MAN Roland argues that HM6 was scheduled to be completed during the POI, and that Commerce’s failure to use information related to this sale as well as the HM4 information was the result of “a simple misreading of the percentage completion schedule submitted on March 13, 1996 by MAN Roland.” MAN Roland brief at 28. According to that schedule, which lists both the WIP balance and the percentage complete for each press as of September 30,1995 and as of December 31, 1995, the figures given for HM6 are based on actual rather than estimated costs. The fact that actual costs were available for HM6, MAN Roland argues, indicates that the press must have been complete at the time MAN Roland submitted the schedule. However, the schedule also indicates that as of December 31, 1995, HM6 was only 85.21 percent complete. Thus, Commerce had substantial evidence indicating that HM4 was the only home-market Geoman press for which it had complete cost information. 3. Overhead Variance Adjustment In its normal cost accounting system, MAN Roland calculates estimated overhead costs for each project by multiplying the direct labor costs for the project by fixed percentages. These percentages are not adjusted for projected or actual levels of activity in the factory. As a result, during periods when the factory is operating at high levels and direct labor costs increase, the estimated overhead costs also increase. However, actual overhead costs do not increase. Thus, when MAN Roland operates its production facilities at high levels, the estimated overhead costs overstate the actual overhead costs. MAN Roland brief at 29. MAN Roland’s Rochester and Wilkes-Barre presses were not completely manufactured as of December 1, 1995, the cut-off date for MAN Roland’s submissions to Commerce. In reporting the estimated cost of production for the incomplete presses, MAN Roland based its overhead variance on its first quarter results for fiscal year 1996 (July 1, 1995 — June 30,1996) and its budgeted fiscal year 1996 variance. Commerce rejected MAN Roland’s adjustment and applied the actual variance from MAN Roland’s most recently completed fiscal year, July 1, 1994—June 30, 1995. Germany Final at 38,187. MAN Roland contends that, “[t]he actual variances from past years did not provide an accurate measure of the variance for the 1995-96 period, ...” Brief of Pis. MAN Roland Druckmasehinen AG and MAN Roland Inc. Reply to Def.’s and Goss Graphics Systems, Inc.’s Resp. Briefs on Comp.- and Country-specific issues at 14 (“MAN Roland Reply brief’), and that “[t]he rationale offered by the department is ... completely without merit and totally unsupported by evidence on the record.” MAN Roland brief at 31. Constructed value is defined in relevant part, as, “an amount equal to the sum of—(1) the cost of materials and fabrication or other processing of any kind employed in producing the merchandise ... (2)(A) the aetual amounts incurred and realized by the specific exporter or producer ... for selling, general and administrative expenses, and for profits,....” 19 U.S.C. § 1677b(e). Thus Congress has allowed Commerce a certain degree of discretion in choosing specific methodologies for calculating elements of CV such as overhead. Here, Commerce explained: MRD’s budgeted variances do not accurately predict full-year operating results and rely on unrealistic capacity utilization levels. In addition, year-end adjustments or one-time annual costs may not be reflected in the partyear actual variance. Therefore, we rejected MRD’s reported part-year actual variance and budgeted fiscal year variance calculation for fiscal 1996. Germany Final at 38,187. Commerce’s decision to reject MAN Roland’s submitted variance was a reasonable application of the constructed value provision. The purpose of the antidumping scheme is to allow Commerce to calculate dumping margins as accurately as possible. See Rhone Poulenc v. United States, 8 Fed. Cir. (T) 61, 67, 899 F.2d 1185, 1191 (1990) (stating that “the basic purpose of the statute,” is “determining current margins as accurately as possible.”) Thus, it is within Commerce’s discretion to reject submitted information that may not accurately reflect actual costs. Furthermore, Commerce’s decision was consistent with Commerce’s longstanding preference for using actual rather than estimated costs. See supra, pp. 843-844. Commerce has the discretion to weigh the relative value of record evidence. “It is not within the Court’s domain either to weigh the adequate quality or quantity of the evidence for sufficiency or to reject a finding on grounds of a differing interpretation of the record.” Timken Co. v. United States, 12 CIT 955, 962, 699 F.Supp. 300, 306 (1988), aff'd, 894 F.2d 385 (Fed.Cir.1990). MAN Roland also argues that, the Department’s alleged concerns about using partial-year variance are simply inconsistent with its past practice. The Department’s normal practice has always been to calculate labor and overhead costs based on the costs incurred during the six-month investigation period—and not to use costs for longer ‘complete’ accounting periods. MAN Roland brief at 32. However, the cases cited by MAN Roland in support of this argument had six-month POIs. In those cases, therefore, Commerce reasonably decided that a six-month variance most accurately approximated the respondents’ actual POI experience. In this case, where the POI was two years, Commerce’s decision that a six-month variance was not the most accurate approximation of respondents’ actual POI experience was reasonable. MAN Roland also argues that Commerce’s decision was not supported by substantial evidence. Specifically, MAN Roland argues, “there was simply no evidence on the record to suggest that MAN Roland’s budgets for its 1995-1996 ... projected ‘unrealistic capacity utilization levels.’ ” MAN Roland brief at 31. However, Commerce found that MAN Roland’s capacity utilization estimates did not accurately predict actual utilization levels. See Cost Verification Report at 3 (“At verification, company officials indicated that MRD’s plant capacity is 2.5 million labor hours per year,_ Based on actual utilization data reported for the period July through December 1995, MRD exceeded its plant capacity by approximately 4.4% during these months.”). MAN Roland argues that the fact that “there was simply no evidence on the record to suggest that MAN Roland’s budgets for its 1995-1996 — which were prepared in the normal course of business ... projected ‘unrealistic capacity utilization levels,”’ because MAN Roland’s capacity utilization was “even higher — and its variances even more favorable — than budgeted.” MAN Roland brief at 31. Whether MAN Roland’s projection understated or overstated its capacity utilization is irrelevant. The fact that the projection was inaccurate supports Commerce’s conclusion that MAN Roland’s projections are not as reliable as actual historical data. Commerce appropriately rejected MAN Roland’s submitted variance. Therefore, the Court finds, Commerce’s decision to adjust MAN Roland’s 1996 fiscal year overhead using the actual variance from fiscal year 1995, the most recent fiscal year for which complete information was available, was reasonable. 4. Combining MAN Plamag and MAN Roland Production Costs In calculating its cost of manufacturing, MAN Roland argues, Commerce should have averaged the labor and overhead rates of MAN Roland’s production facility with those of MAN Plamag, MAN Roland’s wholly owned subsidiary. MAN Roland contends that, [u]nder the Department’s established practice, where a respondent has the ability to produce the subject merchandise at two plants, the reported costs should reflect the weighted-average cost of manufacture at both plants — and not just the costs at the plant where the merchandise was actually produced. The Department has applied this approach in cases where the two facilities are owned by separate subsidiaries of the respondent — as long as the specific merchandise in question can be produced at both plants. MAN Roland brief at 33. In the final determination, Commerce said it did not average costs for MAN Roland and MAN Plamag because “MAN Plamag is a separate corporate entity from MRD. Specifically, MAN Plamag is an affiliated party to MRD (not a division or factory within MRD)-” Germany Final at 38,188. The Court finds Commerce’s explanation to be insufficient. In Certain Fresh Cut Flowers from Colombia, 55 Fed.Reg. 20,491, 20,497 (Dep’t Commerce 1990) (final results admin, review), Commerce agreed to average the costs of production for two related companies. Its position here appears to be at odds with the position it took in that case. “Commerce has the flexibility to change its position providing that it explains the basis for its change and providing that the explanation is in accordance with law and supported by substantial evidence.” Cultivos Miramonte S.A. v. United States, 21 CIT -, 980 F.Supp. 1268, 1271 (1997) (footnotes omitted); see also Hussey Copper, Ltd. v. United States, 17 CIT 993, 997, 834 F.Supp. 413, 418 (1993)(“‘This rule is not designed to restrict an agency’s consideration of the facts from one ease to the next, but rather it is to insure consistency in an agency’s administration of the statute.’ ”)(quoting Citrosuco Paulista, S.A. v. United States, 12 CIT 1196, 1209, 704 F.Supp. 1075, 1088 (1988)). Therefore, the Court is remanding this issue for Commerce to reconsider its decision not to average MAN Roland’s costs with MAN Plamag’s. 5. Highly Profitable Sales In calculating constructed value, Commerce is required to calculate an amount for profit based on the profit margins of sales of a foreign like product made “in the ordinary course of trade.” 19 U.S.C. § 1677b(e)(2)(A). MAN Roland argues that in calculating its profit, Commerce should have disregarded certain highly profitable sales. MAN Roland brief at 35-37. Commerce explained its refusal to exclude such sales from its profit calculation as follows: [w]e disagree with respondents that simply because certain home market sales had profits higher than those of numerous other sales, the profits are automatically abnormally high and outside the ordinary course of trade for purposes of computing CV profit. In order to determine that profits are abnormally high, there must be certain unique or unusual characteristics related to the sales in question. However, the respondents have provided no credible information other than the numerical profit amounts to support their contention that certain home market sales had abnormally high profits. Germany Final at 38,178. In response, MAN Roland maintains that because Commerce’s interpretation confuses the concept “sales with abnormally high profits,” with the concept “abnormal sales with high profits,” “[Commerce’s] interpretation cannot be correct.” MAN Roland brief at 36. Although the Court agrees that Commerce’s explanation is not a model of clarity, the Court finds Commerce acted within its discretion in including all profitable sales in its calculation of CV profit. According to the statute, Commerce is to base SG & A and profit on sales made in the ordinary course of trade. The SAA explains that “... examples of sales that Commerce could consider to be outside the ordinary course of trade include sales of off-quality merchandise, sales to related parties at non-arm’s length prices, and sales with abnormally high profits.” SAA at 839-40 (emphasis added). Thus, Commerce has the discretion to decide under what circumstances highly profitable sales would be considered to be outside of the ordinary course of trade. In order for Commerce to exclude sales found to have been made below the cost of production, the statute requires that Commerce provide additional evidence demonstrating that such sales were actually outside the ordinary course of trade. See 19 U.S.C. § 1677b(b)(l). Similarly, Commerce’s decision to require additional evidence demonstrating that sales with higher profits were outside of the ordinary course of trade was consistent with the statutory scheme and a reasonable construction of the provision at issue. V. United States PRICE 1. Use of Constructed Export Price In calculating a dumping margin, Commerce compares United States price to the normal value of the subject merchandise. United States price is calculated using either an export price (“EP”) methodology or a constructed export price (“CEP”) methodology. Typically, Commerce relies on EP when the foreign exporter sells directly to an unrelated U.S. purchaser. CEP is used when the foreign exporter makes sales through a related party in the United States. See Sharp Corp. v. United States, 63 F.3d 1092, 1093-94 (Fed.Cir.1995) (“The statute defines [U.S. price], ... as either the United States purchase price [now EP] or the exporter’s sales price [now CEP], whichever is appropriate.... Commerce uses the [CEP] if the foreign manufacturer imports through a related company in the United States.”) (citations omitted). For each of the relevant LNPP sales by MAN Roland to the United States, Commerce calculated U.S. price based on a CEP methodology. MAN Roland argues that two of its sales should have been treated as EP sales. The antidumping statute defines EP as follows: the price at which the subject merchandise is first sold (or agreed to be sold) before the date of importation by the producer or exporter of the subject merchandise outside of the United States to an unaffiliated purchaser in the United States or to an unaffiliated purchaser for exportation to the United States.... 19 U.S.C. § 1677a(a). Constructed export price is defined as follows: the price at which the subject merchandise is first sold (or agreed to be sold) in the United States before or after the date of importation by or for the account of the producer or exporter of such merchandise or by a seller affiliated with the producer or exporter, to a purchaser not affiliated with the exporter or producer.... 19 U.S.C. § 1677a(b). When U.S. price is based on CEP, Commerce bases its calculations on the price charged to the first unaffiliated purchaser. This is the starting price. Commerce then makes certain adjustments including several that are not required for EP sales. The CEP adjustments “are made for certain amounts associated with the sale of merchandise in the United States, typically, commissions for selling the merchandise ... and sales expenses generally incurred in selling the same type of merchandise in the United States.” PQ Corp. v. United States, 11 CIT 53, 59, 652 F.Supp. 724, 730 (1987). The purpose of these adjustments is to prevent foreign producers from competing unfairly in the United States market “by spending amounts on marketing and selling their products that are in excess of what they spend in their home markets.” Id. According to the SAA, “constructed export price is ... calculated to be, as closely as possible, a price corresponding to an export price between non-affiliated exporters and importers.” SAA at 823. The statute does not specify the circumstances under which Commerce is to choose EP or CEP. However, according to the SAA, [i]f the first sale to an unaffiliated person in the United States, or to an unaffiliated purchaser for export to the United States, is made by the producer or exporter in the home market prior to the date of importation, then Commerce will base its calculation on export price. If, before or after the time of importation, the first sale to an unaffiliated person is made by (or for the account of) the producer or exporter or by a seller in the United States who is affiliated with the producer or exporter, then Commerce will base its calculation on constructed export price.... Id. at 822-23. Thus, a sale to an unaffiliated party made prior to importation and involving a U.S. party affiliated with the producer or exporter could be an EP or a CEP sale. Therefore, in such a situation, “the determination of whether [EP] or [CEP] applies must be based upon additional circumstances.” PQ Corp. at 60, 652 F.Supp. at 731. In Certain Stainless Steel Wire Rods from France, 58 Fed.Reg. 68,865 (Dep’t Commerce 1993) (final det.), Commerce described the additional criteria it examines in deciding whether to use an EP or CEP methodology as follows: The first criterion is that the merchandise in question is shipped directly from the manufacturer to the unrelated buyer, without being introduced into the inventory of the related selling agent. The second criterion is that this arrangement is the customary commercial channel for sales of this merchandise between the parties involved. ... The third criterion is that the related selling agent located in the United States acts only as a processor of sales-related documentation and a communication link with the unrelated U.S. buyer. Id. at 68,868-69. Commerce will apply the EP methodology only if all three criteria apply to the sales at issue. Id. at 68,868. This test has been approved by this court. See Independent Radionic Workers v. United States, 19 CIT 375, 375 (1995)(describing Commerce’s three criteria as “the judicially approved test”). Commerce applied the same criteria in this ease, relying on the third criterion in its decision to base U.S. price on CEP. MRU’s role with respect to the sales at issue is beyond that of a mere “processor of sales documentation” and “communications link.” MRU played a major role in the negotiations between MRD and the U.S. customer for the Rochester and Wilkes-Barre sales, ... and incurred significant SG & A expenses in the process. The contractual documentation and sales-related correspondence viewed at verification attests to this fact. Furthermore, we verified that MRU supports MRD’s activities in the shipment and installation process relevant to these sales. This is evidenced by the fact that MRU is responsible for the post-sale warehousing of the merchandise shipped from Germany ... as well as the contracting of rigging companies and the sourcing of auxiliary parts essential to the installation process in the United States. Germany Final at 38,176. MAN Roland argues that, the standard for distinguishing between export price and constructed export price sales is whether the performance of functions by the U.S. subsidiary changes “the substance of the transaction or the functions themselves.” In other words, if the functions performed by the U.S. subsidiary could have been performed by the foreign parent, without any participation by the U.S. subsidiary, then the sale should be classified as an export price sale. MAN Roland brief at 38 (citing Stainless Steel Hollow Products from Sweden, 52 Fed. Reg. 37,810, 37,811 (Dep’t Commerce 1987) (final det.); Certain Internal-Combustion, Industrial Forklift Trucks from Japan, 53 Fed.Reg. 12,552, 12,553 (Dep’t Commerce 1988) (final det.)). “The evidence in this case demonstrates that all of the functions performed by MRU (MAN Roland’s U.S. subsidiary) on the sales to Rochester and Wilkes Barre were routinely performed by its German parent (using unaffiliated local contractors where necessary) for its sales in many of the third-country markets it supplies_ Thus, MAN Roland’s sales to Rochester and Wilkes Barre should properly have been treated as ‘export price’ sales.” MAN Roland brief at 39. However, as Commerce explained, it is only in cases where the three EP elements are met, that, the Department has regarded the routine selling functions of the exporter as “merely having been relocated geographically from the country of exportation to the United States,”.... [W]here the functions are performed “does not change the substance of the transactions or the functions themselves.” Germany Final at 38,175 (citing Forklift Trucks from Japan, 53 Fed.Reg. at 12,553). Only in such cases will Commerce apply the EP methodology. In this case, Commerce provided substantial evidence demonstrating that the third criterion had not been met. Specifically, Commerce examined contractual documentation and sales-related correspondence indicating that MRU “played a major role in the negotiations between MRD and the U.S. customer ... from the bidding stage through to the final contracts_” Germany Final at 38,176. Furthermore, the verification report demonstrates that MRU is responsible for the post-sale warehousing of the merchandise shipped from Germany, the contracting of rigging companies and the sourcing of auxiliary parts necessary for installation. See Verification QR. Responses of MRU, N-P Doe. 93 at 17 (post-sale warehousing expense); id at 27 (installation costs). Thus, Commerce’s decision to rely on CEP rather than EP in calculating U.S. price was appropriate. 2. Further Manufacturing In calculating MAN Roland’s U.S. price, Commerce treated installation of the subject merchandise as further manufacture. According to the statute, CEP is to be reduced by, “the cost of any further manufacture or assembly (including additional material and labor) ...” 19 U.S.C. § 1677a(d)(2). MAN Roland maintains that installation expenses for its Wilkes-Barre and Rochester sales should have been treated as movement-related expenses, pursuant to 19 U.S.C. § 1677a(c)(2)(A), which requires Commerce to reduce EP and CEP by “the amount, if any, included in such price, attributable to any additional costs, charges, or expenses ... which are incident to bringing the subject merchandise from the original place of shipment in the exporting country to the place of delivery in the United States.” The distinction is significant because Commerce calculates movement-related expenses without imputed profit. Further manufacturing costs, on the other hand, include an imputed profit attributable to the value added by the further manufacturing activities. See 19 U.S.C. § 1677a(d) (“For purposes of this section, the price used to establish constructed export price shall also be reduced by ... (2) the cost of any further manufacture or assembly ... and (3) the profit allocated to the expenses described in [paragraph] (2)).” MAN Roland argues that Commerce’s decision would lead to absurd results because “MAN Roland is able to install LNPPs in distant foreign countries using independent contractors, without any substantial investment in or support from a local subsid-iary_” MAN Roland brief at 42. For example, MAN Roland argues, if MAN Roland had decided to install the U.S. sales from Germany using only independent local contractors, without any involvement by its U.S. subsidiary, the Department would have had to classify the sales as export sales. And ... no adjustment could then have been made for further manufacturing costs.... The Department then would have been faced with two equally absurd options: It could continue to classify installation costs as further manufacturing costs, and make no adjustment for them at all [ (the statute does not include a further manufacturing deduction for EP sales) ]. Or, alternatively, the Department could hold that the cost of performing installation for such sales should be classified as movement charges ... even though the costs of performing the identical operations would be classified as further manufacturing when supervised by a U.S. subsidiary. MAN Roland brief at 42-43. MAN Roland’s hypothetical has no persuasive force, The Court must uphold Commerce’s determination if it is in accordance with law and supported by substantial evidence. The statute does not define further manufacture. Therefore, the Court must defer to Commerce’s interpretation if that interpretation is reasonable. Here, Commerce explained, it decided to treat installation costs as further manufacturing “because the U.S. installation process involves extensive technical activities on the part of engineers and installation supervisors and the integration of subject and non-subject merchandise necessary for the operation of LNPPs.” Germany Final at 38,177. This decision was a reasonable interpretation of the statute. The statute does not require Commerce to analyze costs deducted as CEP costs to determine whether the activities generating such costs would be consistent with an EP transaction. The statute does require that Commerce deduct further manufacturing costs from CEP. Here, Commerce appropriately classified the sales at issue as CEP sales. Thus, Commerce was obligated to deduct the cost of further manufacture. Commerce’s decision that the activities at issue here constituted further manufacture was consistent with Commerce’s usual practice, which is to classify installation costs as further manufacture where the activities at issue include, “the incorporation of integral, non-subject components during installation or complex installation operations that are more than mere reassembly....” Germany Final at 38,177. Therefore, the Court finds, Commerce’s deduction of MAN Roland’s installation costs as further manufacture was in accordance with law. VI. Goss’s Issues I. Commerce’s Cost Allocation for MAN Roland As explained above, after rejecting MAN Roland’s submitted cost figures, based upon the AJDK system, Commerce decided to rely on cost estimates prepared by MAN Roland prior to the initiation of the instant investigation. Goss agrees with Commerce’s decision to reject the cost information submitted by MAN Roland. However, Goss argues, “Commerce made an error that incorrectly reduced the estimated cost.” Goss brief at 11. Commerce made certain adjustments to the cost estimates by breaking them down into categories (i.e., cost of manufacturing, SG & A, and profit). “The error arose,” explains Goss, “because Commerce had to create a constructed value from two sets of data....” Id. at 12. During the investigation, MAN Roland reported actual SG & A costs that Commerce verified. Commerce decided to subtract the estimated SG & A from MAN Roland’s estimates, and add in the actual SG & A. The cost estimates did not break out SG & A, so Commerce tried to back the estimated SG & A out of the cost estimates by using the reported SG & A ratio. The SG & A ratio is used to allocate SG & A costs to individual projects. The ratio is calculated by dividing the total SG & A costs for all projects within a pool by an allocation base, here, total cost of production for all the projects in the pool. The ratio is multiplied by the cost of production of individual projects to calculate the SG & A costs that should be allocated to them. However, Goss explains, in the cost estimates submitted by MAN Roland to Commerce, the SG & A ratio had been calculated based on a denominator that excluded the cost of purchased parts worth more than DM 500. Goss Brief at 14 (citing MAN Roland Section D response at 27). Commerce multiplied this ratio by the full cost of production for individual projects. Therefore, the subtracted SG & A amounts were inflated. Id. Goss requests that the Court remand this issue so that Commerce may apply the appropriate allocation ratios to MAN Roland’s estimated costs. Id. at 15. Commerce agrees that it erred in allocating SG & A expenses and also requests a remand so that it may correct its error. Def.’s Mem. Opp. Motion Goss Graphic Systems, Inc., J. Agency Ree. Country/Company-Speeific Issues at 14 (“ITA understated the COM because the estimated cost data submitted by MRD did not include the cost of certain parts.... Therefore, ITA agrees that the case should be remanded to allow ITA to determine a more accurate cost allocation ... using the information contained in the administrative record.”). MAN Roland opposes the remand, arguing, “[i]f the Department really had been authorized to rely on ‘facts available’ in this case, then the choice it made was well within its discretion.” Brief of Pis. MAN Roland Druekmaschinen AG and MAN Roland Inc. Opp. Mot. J. Agency Rec. Goss Graphic Sys., Inc. Comp.- and Country-Specific Issues at 10 (“MAN Roland Response Brief’). MAN Roland’s argument is flawed, however. Commerce did not make a choice here. Commerce’s use of an estimated SG & A ratio based on less than the cost of production was inadvertent. As MAN Roland itself acknowledges, “[a]s a matter of simple arithmetic, the denominator used to calculate an expense rate must be determined on the same basis as the figures to which the rate will be applied.” MAN Roland brief at 16. In Koyo Seiko Co. v. United States, the Court observed that “[t]here can be no doubt that Congress intended final determinations to be precisely that.” 14 CIT 680, 682, 746 F.Supp. 1108, 1110 (1990). However, the court added, “failure to reopen a determination which is known to be