AK Steel Corp. v. United States

21 Ct. Int'l Trade 1204

• Court: United States Court of International Trade
• Decided: November 14, 1997
• Docket: Court No. 96-05-01312
• Status: Published

Opinion

Opinion

Restani, Judge:

This matter is before the court on a Motion for Judgment on the Agency Record pursuant to USCIT R. 56.2 by plaintiffs, AK Steel Corporation, Bethlehem Steel Corporation, Inland Steel Company, Inc., LTV Steel Company, Inc., and U.S. Steel Group, a unit of USX Corporation (collectively “plaintiffs”); against defendant, United States Department of Commerce (“Commerce”), and defendant-intervenors, Continuous Colour Coat, Ltd. (“CCC”), Dofasco, Inc. (“Dofasco”), So-revco, Inc. (“Sorevco”), and Stelco, Inc. (“Stelco”). Plaintiffs are domestic producers of corrosion-resistant carbon steel flat products. Under review are the final results of an administrative review. Certain Corrosion-Resistant Carbon Steel Flat Products and Certain Cut-to-Length Carbon Steel Plate from Canada, 61 Fed. Reg. 13,815 (Dep’t Commerce 1996) (final results of admin, review) [hereinafter “Final Results”].

Plaintiffs contend that: (1) Commerce’s acceptance of Dofasco’s cost methodology is unsupported by substantial evidence on the agency record and is not in accordance with law; (2) Commerce improperly included a revision by Dofasco to an expenditure from a prior period as an element of cost in the current review; (3) Commerce erred in its treatment of excess prime merchandise in its model-match exercise for which Stelco failed to report complete product characteristics; (4) Commerce’s decision to accept Stelco’s adjustments to prices is unsupported by substantial evidence on the agency record and is not in accordance with law; (5) the court should remand for the correction of ministerial errors in Commerce’s final results for Stelco; and (6) Commerce erred in accepting CCC’s improperly allocated price adjustments.

In response, Commerce argues that all of its challenged determinations were supported by the agency record and in accordance with the law. Commerce also requests a remand to correct ministerial errors contained in the final margin calculation for Stelco. This request is granted.

Background

In August 1993, Commerce issued an antidumping duty order covering corrosion-resistant carbon steel flat products from Canada. Certain Corrosion-Resistant Carbon Steel Flat Products and Certain Cut-to-Length Carbon Steel Plate from Canada, 58 Fed. Reg. 44,162, 44,162 (Dep’t Commerce 1993)(antidumping duty orders). In August 1994, Canadian producers of the subject merchandise filed a request for administrative review of Commerce’s antidumping order. Certain Corrosion-Resistant Carbon Steel Flat Products and Certain Cut-to-Length Carbon Steel Plate from Canada, 60 Fed. Reg. 42,511, 42,511 (Dep’t Commerce 1995)(prelim, results). The period of review (“POR”) was from February 1993 through July 1994. Id. In August 1995, Commerce published the preliminary results of the administrative review.¹ Id. In March 1996, Commerce published its final results of the administrative review. Final Results, 61 Fed. Reg. at 13,815. Plaintiffs raised the following challenges to Commerce’s determination below and now again on appeal.

1. Dofasco:

Plaintiffs challenged Dofasco’s order-specific cost methodology for determining cost of production (“COP”) and constructed value (“CV”). Plaintiffs argued that Dofasco calculated COP on the basis of home market sales only and CV on the basis of U.S. sales only. Commerce rejected plaintiffs’ argument notingthat its acceptance of Dofasco’s COP and CV data is proper because Dofasco’s calculations are based upon a weighted-average production basis for the costs incurred to produce the subject merchandise. Furthermore, Commerce found that Dofasco’s cost methodology complies with both the law and Commerce’s questionnaire. For these reasons, Commerce determined that application of best information available (“BIA”) was inappropriate.

Plaintiffs also challenged Commerce’s acceptance of Dofasco’s partial reversal of restructuring charges taken in a prior period. Plaintiffs argued that Commerce has consistently held that costs relating to a prior period, and the credit associated with them, have no logical relation to production costs of the subject merchandise during a later period. Commerce argued that the inclusion of disputed cost figures was consistent with Commerce policy.

2. Stelco:

Plaintiffs argued that Commerce’s model-match methodology when applied to Stelco’s sales of excess prime merchandise reported with missing physical characteristics was not in accordance with law or supported by substantial evidence on the record. Plaintiffs made two arguments in support of their challenge: (1) Commerce must match sales based on actual physical characteristics of the products and may not match products with different missing physical characteristics; and (2) as a result of Stelco’s inability to provide Commerce with adequate information, Commerce is required to apply BIA.

Commerce rejected plaintiffs’ challenge assertingits brpad discretion to determine what merchandise is categorized as “such or similar.” Commerce also verified Stelco’s data and argues that the sales of the merchandise missing physical characteristics were minimal and therefore harmless. Finally, Commerce rejected the use of BIA arguing that where the respondent cooperated and provided all the information requested, the use of BIA is inappropriate. Final Results, 61 Fed. Reg. at 13,830-31.

Plaintiffs also challenged Stelco’s reported gross unit prices in Canadian and American markets. Plaintiffs argue that Stelco’s price adjustments were made to gross unit prices directly and were not separately reported as required. Furthermore, plaintiffs alleged the price adjustments were not transaction-specific, but instead are impermissibly spread across multiple sales referenced on a particular credit or debit memo.

Commerce rejected plaintiffs’ challenge arguing that it appropriately verified Stelco’s data, that separate reporting was not required, and that Stelco properly adjusted prices for debit and credit notes.

3. Continuous Colour Coat:

Finally, plaintiffs argued that Commerce erred by accepting CCC’s adjustments to prices of the subject merchandise through credit notes or debit notes that were issued after invoicing of the merchandise and that were allocated over multiple sales. They asserted that Commerce should have used BIA when the adjustments were not matched to a specific sale. Commerce, however, found that the debits and credits were transaction-specific and allowed them as a direct price adjustment.

Standard of Review

The court will uphold the final results of an antidumping duty administrative review unless 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). Substantial evidence is “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Universal Camera Corp. v. NLRB, 340 U.S. 474, 477 (1951). “ ‘ [T]he possibility of drawing two inconsistent conclusions from the evidence does not prevent an administrative agency’s finding from being supported by substantial evidence. ’ ” Magnesium Corp. of Am. v. United States, 938 F. Supp. 885, 889 (Ct. Int’l Trade 1996)(quoting Consolo v. Federal Maritime Comm’n, 383 U.S. 607, 620 (1966)).

Discussion

I. Dofasco:

A. Cost Methodology:

When calculating antidumping duties, Commerce must determine whether home market sales were made at less than the cost of production (“COP”). 19 U.S.C. § 1677b(b) (1988). Commerce is, in certain circumstances, also required to disregard such sales in determining foreign market value. Id. When Commerce disregards such below-cost sales and determines the remaining home market sales “to be inadequate as a basis for the determination of foreign market value, ” the statute requires Commerce to use constructed value (“CV”) of the merchandise to determine its foreign market value. Id.

Both COP and CV include a determination of the cost of manufacture (“COM”) for the product. COM is the total material and fabrication cost incurred in producing “such or similar merchandise” during the period of review. 19 U.S.C. § 1677b(e)(l) (calculation of CV); 19 C.F.R. § 353.51(c)(1994) (calculation of COP). The statute defines “such or similar merchandise” as:

[t]he merchandise which is the subject of an investigation and other merchandise which is identical in physical characteristics with, and was produced in the same country by the same person as, that merchandise.

19 U.S.C. § 1677(16)(A)(1988).

Plaintiffs challenge three aspects of Dofasco’s methodology used in calculating the cost of manufacture. First, plaintiffs claim that Dofasco did not calculate its weighted-average costs per product using its entire production volume, but rather used only production for home market sales to determine the COM for COR and only production for U.S. sales to determine the COM for CV Second, by using quarterly-averaged prices for raw materials the costs were distorted as they were dependent on when the merchandise was produced. Third, plaintiffs claim that Do-fasco misled Commerce in its responses concerning its methodology and thus Commerce should have applied BIA.

1. Methodology:

Plaintiffs argue that Commerce’s acceptance of Dofasco’s methodology was not in accordance with law because the methodology limited the COM for cost of production and constructed value calculations according to the product’s destination. Commerce’s interpretation of the statutory requirements identifying the product for which the respondent must supply cost information (i.e., the product for COP is the product sold in the home market; the product for CV is the product sold in U.S.) are not contested. Instead, plaintiffs challenge the interpretation of the statutory provisions identifying how COM for a product must be determined once the product is identified. Plaintiffs assert that both the statute and regulations expressly provide that all costs to produce identical merchandise, irrespective of the destination of the completed product, must be included in the weighted-average COP and CV calculations.

To support their position, plaintiffs suggest that this rule was articulated by Commerce in Certain Circular Welded Carbon Steel Pipes and Tubes from the Republic of Korea, 49 Fed. Reg. 9926 (Dep’t Commerce 1984)(final determ.)[hereinafter “Steel Pipes from Korea”]. In Steel Pipes from Korea, the producer of coil used to produce the merchandise under investigation maintained a two-tier pricing system in which coil for eventual domestic consumption was sold at higher prices than the coil for export. Id. at 9927. Respondents in that case argued that the cost of producing the merchandise under investigation must be calculated only on the basis of the higher coil costs incurred in producing merchandise for domestic consumption. Id. Commerce disagreed with respondents stating that “where there are different costs associated with producing merchandise produced for export as compared with merchandise produced for domestic sale and the merchandise is physically identical, we have used the average cost of producing that merchandise in calculating cost of production or constructed value.” Id. at 9928. Commerce further noted that section 1677b(e) of Title 19 states that “the constructed value of imported merchandise” is “the cost of materials * * * and of fabrication or other processing of any kind employed in producing such or similar merchandise.” 19 U.S.C. § 1677b(e)(l). Thus, plaintiffs allege that the CV of the merchandise sold in the U.S. includes all costs to the company to produce that product, regardless of where it is ultimately sold.²

The court finds this argument unpersuasive for the following reasons. First, Steel Pipes from Korea does not explicitly stand for the rule suggested above. See 49 Fed. Reg. at 9928. Instead, it merely states that in a situation with a two-tiered price system for the raw materials used in the merchandise under investigation, Commerce will use the average cost of producing that merchandise. Id. This case does not involve a two-tiered pricing system for raw materials.

Second, the statute does not expressly resolve this issue as plaintiffs contend. It addresses only the issue of identifying the relevant product for CV and COP, but does not address the method by which Commerce must calculate the COM in either situation. Thus, neither plaintiffs nor Dofasco state a plain meaning of the statute which resolves the issue at hand.

The court is thus confronted with the classic situation where a statute does not expressly address the issue before the court. Under Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-43 (1984), the court must defer to an agency’s reasonable interpretation of the statute if the court finds Congress “has not directly addressed the precise question at issue.” Id. at 843. The question for the court now is whether Commerce’s application of 19 U.S.C. § 1677b reflects a permissible construction of the statute. See id. ■

Here, Commerce accepted Dofasco’s methodology that first calculated a per unit production cost based on total production irrespective of the intended market and then calculated a weighted-average of all products by control number, which reflected production for sale in a particular market. Final Results, 61 Fed. Reg. at 13,824. Commerce’s acceptance of Dofasco’s methodology essentially finds a middle ground. Total production costs are incorporated into the COM, but final COP and CV are determined based on a weighted-average reflecting production for a particular market. The statute does not dictate a different result.

Although Dofasco’s methodology, based on its particular method of record keeping, is not a usual method, substantial evidence on the record shows that Commerce considered and understood the mechanics of the methodology before accepting the methodology as a reasonable approach under the statute. Contrary to plaintiffs’ claim, Dofasco sufficiently disclosed its methodology to Commerce before verification thereby giving Commerce the opportunity to understand the mechanics of the methodology and conclude that it was reasonable before applying it. See Dofasco’s Section VI Response (Mar. 6,1995), at 26-27, ER. Doc. 312, Pis.’ App., Tab 10; Dofasco’s Section VI Supplemental Response (Mar. 31, 1995), at 2, C.R. Doc. 162, Dofasco’s App., Tab 10.

The record compiled by Commerce provides a clear description of Do-fasco’s methodology. When verifying Dofasco’s reported costs, Commerce reported that the accounting system calculates product cost per unit based on total production costs. Dofasco’s Section VI Response (Mar. 6,1995), at 1, 9-13, ER. Doc. 312, Def.’s App., Ex. 11. The system then identifies the specific cost centers through which each individual sales order was processed. Dofasco Cost Verification Exhibit No. 18 (May 12,1995), C.R. Doc. 200, Pis.’ App., Tab 25. Thus, to calculate cost for COP and CV Dofasco calculated the total cost for a specific product and then divided that amount by the total production of that product.³ See Cost Verification Exhibit No. 21 (Apr. 18,1995), C.R. Doc. 200, Do-fasco’s App., Tab 5; CostVerificationExhibitNo. 22 (Apr. 18,1995), C.R. Doc. 200, Dofasco’s App., Tab 6. Finally, Commerce verified that Dofas-co in its second step, appropriately calculated a weighted-average cost for each control number. See Dofasco’s Cost Verification Report (May 12,1995), at 4, RR. Doc. 417, Def.’s App., Ex. 13; Dofasco’s Cost Verification Ex. No. 16, C.R. Doc. 417, Def.’s Confid. App., Ex. 7; Dofasco’s Section VI Response, at 2, 30, RR. Doc. 312, Def.’s App., Ex. 11; Dofasco’s Section VI Supplemental Response, at 2, Dofasco’s App., Tab 10.

2. Distortion:

Plaintiffs argue that as the rate for each component was revised based on the costs incurred and the volume produced during a specific quarter, the production costs varied depending on when it was produced during the POR. Plaintiffs also argue that because the order specific costs would not necessarily be the same for identical products, it was important to include the costs for all production orders of a product in the weighted-average calculation. Thus, plaintiffs conclude that the omission of relevant production orders from the weighted-average calculation resulted in distorted costs, because the COM for identical products should have been the same. This distortion, plaintiffs argue, creates a loophole through which Dofasco is able to deflate its production costs and distort the margin calculation.

Specifically, plaintiffs argue that Dofasco’s quarterly averaging of prices results in differing COMs for merchandise found in the same CONNUM.⁴ Plaintiffs further argue that for those CONNUMs where merchandise was sold in both markets, the resulting COMs are incorrect because the costs of United States and third country production orders were omitted from COP and the costs of Canadian and third country production orders were omitted from CV⁵

Upon review of the record, the court finds two problems with plaintiffs’ argument. First, the fact that Dofasco weight-averaged sales within a CONNUM based upon the relative production volume of home market or U.S. sales does not mean, as plaintiffs argue, that Dofasco omitted relevant costs. As described above, the COM for each product within a CONNUM was based upon the entire production volume of that product. Furthermore, differing costs of manufacture for COP and CV indicate that the sales to each market have been appropriately matched with their costs. Plaintiffs do not argue that the methodology is specifically designed to mask artificially raised or lowered costs, as in a two-tiered pricing system. Instead, as plaintiffs note themselves, the “distortion” occurs in part because of the use of quarterly costs reports.

Second, plaintiffs fail to meet their burden of proving that Dofasco’s cost methodology distorts the antidumping margin.⁶ Plaintiffs argue for an alternative method that would prohibit the use of quarterly averaged prices for raw materials and require that all production orders be included when calculating COP and CV The existence of an alternative method may not in itself demonstrate the lack of reasonableness in Do-fasco’s accounting method. See Federal-Mogul Corp. v. United States, 63 F.3d 1572,1579-80 (Fed. Cir. 1995) (according considerable deference if application requires agency’s expertise); Hyster Co. v. United States, 18 CIT 119,122, 848 F. Supp. 178,182 (1994)(Commerce’s determination need not be the one the court views as the most reasonable). Ordinarily, quarterly averaging of prices would not manifest a substantial distortion in cost data requiring a reversal of Commerce’s decision. Quarterly averaging of prices may pose a concern if it were shown that Dofasco chose the quarter in which it would produce all merchandise exported to the United States for that year, another quarter specifically designated as the “home market quarter,” and yet another specifically designated as a “third country quarter. ” In such a situation Dofasco would be able to hide costs and circumvent antidumping law. There is, however, no indication of a deliberate scheme by Dofasco to manipulate manufacturing and take advantage of such a loophole.

The court will not reject Dofasco’s methodology as Commerce’s decision to accept Dofasco’s methodology was a reasonable means of effectuating the statutory purpose and is supported by substantial evidence in the record.

3. Best Information Available:

Finally, plaintiffs argue that Dofasco has impeded Commerce’s investigation and, therefore, Commerce must reject the reported costs and apply adverse BIA. See 19 U.S.C. § 1677e (c)(1988); 19 C.F.R. § 353.37(a)(l)(1994). Plaintiffs argue that Dofasco impeded the course of the investigation (1) by failing to timely disclose the nature of the cost methodology and deliberately mischaracterizing the methodology to Commerce and (2) by insuring that Commerce was unable to correct the subsequent distortions.

Specifically, plaintiffs find inadequate Dofasco’s March 1996 response to Commerce’s antidumping review questionnaire where Dofasco stated “[t]he costs were calculated by determining the weighted-average cost per cwt. of all the sales invoices with the product characteristics, as defined in Appendix V of the Department’s questionnaire.” Dofasco’s Section VI Response, at 27, PR. Doc. 312, Pis.’ App., Tab 10. Thus, plaintiffs argue that Dofasco gave Commerce no indication, prior to verification, that it had weight-averaged order-specific costs based on the destination of the merchandise.

The court, however, finds plaintiffs’ argument meritless. The record indicates that Dofasco sufficiently disclosed its cost methodology at the outset of the antidumping review. In response to Commerce’s anti-dumping review questionnaire, Dofasco stated that “the reported COP is the fully-absorbed cost per hundred weight incurred to produce corrosion resistant products sold in the home market” and further stated that the “reported CV is the fully-absorbed cost incurred to produce corrosion resistant products sold in the United States.” Dofasco’s Section VI Response, at 1-2, PR. Doc. 312, Def.’s App., Ex. 11. The court finds that juxtaposing these two phrases, Dofasco at least implied its use of a market specific methodology for weight-averaging in its COP and CV cost of manufacture calculations. Furthermore, there is no indication that Commerce investigators were surprised or were unprepared to analyze Dofasco’s costs at verification.

One of the purposes of BIA is to induce the recalcitrant participant into providing Commerce with complete and accurate information in the course of an antidumping investigation. NTN Bearing Corp. of Am. v. United States, 17 CIT 713, 719, 826 F. Supp. 1435,1440 (1993). Commerce has the discretion to resort to adverse BIA where it believes that the respondent, through refusal or mere inability, is not complying with the investigators. See id. at 720, 826 F. Supp. at 1441; see also 19 U.S.C. § 1677e(c). Here, however, the record plainly demonstrates both that Commerce was aware of Dofasco’s methodology and that Dofasco cooperated with Commerce’s review of the antidumping investigation. The court finds that Commerce’s determination that Dofasco timely responded to its inquiries and in sufficient detail was reasonable and thus the application of adverse BIA is inappropriate.

B. Dofasco’s Partial Reversal of Restructuring Charges Taken in a Prior Period:

As background, it should be observed that in calculating Dofasco’s COP and CV during the earlier less than fair value (“LTFV”) investigation, Commerce included as costs Dofasco’s estimated multi-year expenditures related to restructuring of the corporation. Certain Hot-Rolled Carbon Steel Flat Products, Certain Cold-Rolled Carbon Steel Flat Products, Certain Corrosion-Resistant Carbon Steel Flat Products, and Certain Cut-to-Length Carbon Steel Plate from Canada, 58 Fed. Reg. 37,099, 37,108 (Dep’t Commerce 1993) (final determ.). The estimated future expenditures were considered ordinary expenses that Dofasco charged against its 1992 income. See id. The LTFV determination, however, is not before the court and the agency record relating thereto is not part of the record for the purposes of this action.

In this first administrative review, Dofasco reported certain partial reversals of the earlier estimated restructuring expenditures.⁷ Dofas-co’s Section VI Response, at 7, 40-41, PR. Doc. 311, Def.’s App., Ex. 11; Dofasco’s Supplemental Response (Mar. 31, 1995), at Ex. 1, ER. Doc. 364, Def.’s App., Ex. 12. These reductions were included in Dofasco’s financial statements in 1993 and 1994 as a credit to costs. Dofasco’s Supplemental Section VI Response (Mar. 31, 1995), at Ex. 1, ER. Doc. 364, Els.’ App., Tab 11. In the final results of the first administrative review Commerce stated:

In the present review, Dofasco’s financial statements include certain partial reversals of those earlier restructuring estimates. In order for the Department to be consistent and abide by its long-standing policy, it must also include these partial reversals in its calculation of COE and CV for Dofasco.

Final Results, 61 Fed. Reg. at 13,825. Flaintiffs now argue that because the reversals concern costs from a prior fiscal year and do not relate to costs to produce the merchandise during the FOR they should not have been included in the reported costs. The inclusion of reversals, plaintiffs argue, is also contrary to Commerce’s past practice.

In defendant’s memorandum in partial opposition to plaintiffs’ motion for judgment on the agency record, the government requested a remand to the Department of Commerce to clarify its policy with respect to the reversal charges and to determine if the adjustments made for Do-fasco were consistent with that practice and policy. The court did not grant immediate remand, but ordered the government to explain and describe the policy and past practice of the Department of Commerce.

As articulated before the court, Commerce’s past practice regarding reversal of charges for a prior period has two components. As a first step, Commerce will rely upon a respondent’s books and records prepared in accordance with the home country’s Generally Accepted Accounting Frinciples (“GAAF”) unless those accounting principles do not reasonably reflect the costs of producing the merchandise. See Certain Cut-To-Length Carbon Steel Plate from Germany, 61 Fed. Reg. 13,834, 13,837 (Dep’t Commerce 1996) (final results of admin, review). This general principle is not in dispute and Commerce has found that reversals of prior period costs in a subsequent period distort costs. Id. Thus, even though the reversals of prior charges correctly appear on financial statements in accordance with GAAR Commerce has not relied on the respondent’s books in such a situation.

Commerce reached the same conclusion in Small Diameter Circular Seamless Carbon and Alloy Steel, Standard, Line and Pressure Pipe from Italy, 60 Fed. Reg. 31,981,31,991 (Dep’t Commerce 1995) (final de-term.) [hereinafter “Steel Pipe from Italy”] where petitioners challenged respondent’s decision to include in its reported costs reversals that were recorded during the period of investigation (“FOI”), but that related to operating expense accruals and write-downs of equipment and inventory from a period predating the POI.⁸ Commerce held that such reversals should not be included in the reported costs:

We do not consider it appropriate to reduce the current year production costs by the reversal of prior year operating expense accruals and write-downs of equipment and inventory. The subsequent year’s reversal of these estimated costs does not represent revenue or reduced operating costs in the year of reversal. Rather, they represent a correction of an estimate which was made in a prior year.

Steel Pipe from Italy, 60 Fed. Reg. at 31,991 (citation omitted) (emphasis added). Commerce noted that reducing a subsequent year’s costs because of the reversal in that year of a prior year’s estimate would mean “distorting the actual production costs incurred in a subsequent year.” Id. Commerce did not expressly conclude, as defendant appears to argue, that the distortion would only occur when the expense was incurred prior to any Commerce investigation and the reversal occurred during the POI. Instead, Commerce noted that the adjustment should occur in the same period as the accrual. Specifically, Commerce stated:

If the Department is able to verify that an operating expense accrual or an equipment or inventory write-down recorded during the POI is subsequently adjusted because the company overestimated the cost, we will use the corrected figure, but only for the same period in which the accrual or write-down occurred.

Id.

The past practice described above is consistent with Commerce’s matching principle for cost analysis in LTFV investigations and periodic reviews. The matching principle states that an expense should be recorded in the period in which the product makes its contribution to revenue with the intent to ensure that expenses and revenues are recorded in the proper period. Donald E. Kieso & Jerry J. Weygandt, Intermediate Accounting 46 (8th ed. 1995)(attached to Def.’s Response to Court Order, at Ex. 1). Applied to Commerce’s cost analysis, the matching principle implies that “any adjustment (such as reversals) recorded in an accounting period that relates to an expense incurred and accrued in a prior accounting period, in fact, represents a change in the cost from that prior accounting period, rather than a cost adjustment pertaining to the current accounting period.” Def.’s Response to Court Order, at 3.

Although Commerce stated unequivocally that there is “not justification for distorting actual production costs incurred in a subsequent year by reducing subsequent year costs by the overestimated amount,” Certain Cut-To-Length Steel Plate from Germany, 61 Fed. Reg. at 13,837, the government asserts that as a second step in the analysis Commerce may recognize an exception to its general rule in a case such as this one.

It articulates the exception as follows:

[t]he matching principle in accounting may be superseded, however, as a consequence of the conservative nature of accounting and its purpose in providing information to financial statement users. In these situations, the concept of conservatism dictates that certain expenses [such as restructuring] relating to liabilities for current and future periods be accrued in the first accounting period in which they can be reasonably estimated.

Def.’s Response to Court Order, at 4-5. Thus, the government argues that because in the LTFV investigation Commerce included, in its entirety, the amount of estimated expenditures in the COP/CV calculation and because implementation of the multi-year restructuring plan was still in progress during the review, it “was reasonable to allow Dofasco to include in its COP/CV calculation certain adjustments (reversals) to the estimated expenditures accrued in 1992.” Id. at 6.

Defendant’s explanation of why an exception to the matching principle exists and was applied here does not withstand analysis. First, conservatism does not supersede the matching principle, but rather is incorporated into it as a general quality found in all information used in financial statements. See 2 Financial Accounting Standards Board, Original Pronouncements, Accounting Standards 1041-42 (1993).

Second, accruing Dofasco’s entire estimate in the first available accounting period “because of conservatism” does not lead logically to the conclusion that including partial reversals in a subsequent period is appropriate because they would relate to the costs of that period. For accounting purposes, reversals should be credited as soon as possible. If they are credited before the relevant period of review is completed there is no problem. See Steel Pipe from Italy, 60 Fed. Reg. at 31,991. But excessive estimates are respondent’s responsibility and if they are not corrected promptly they may not be subject to correction in later antidumping proceedings.

Third, while it may not have been appropriate to include all costs for a multi-year restructuring in the LTFV investigation cost calculation, no matter what the proper result is for the LTFV investigation, it is not before the court and the cost adjustments at issue must stand on their own.

Finally, giving a credit against costs accounted for years earlier when they were estimated but not incurred may result in a double distortion. The costs may actually impact the company in the current period. As defendant noted, the restructuring is still in progress, but the costs were deducted instead of added creating the potential for a double distortion. The periodic investigation now before the court, however, sets the actual assessment rates and it must be accurate, whether or not the original estimates from the LTFV investigation were correct. As Commerce has recognized, whether or not financial accounting practices are generally acceptable, they must be rejected for antidumping purposes if they do not accurately reflect costs.

Commerce’s past practice with regard to reversals appears reasonable. The exception now proffered is not supported by any of the arguments made by defendant. In any case, the court may not accept counsel’s post hoc rationalizations for agency action. Burlington Truck Lines, 371 U.S. at 168-69. ‘“[A] simple but fundamental rule of administrative law * * * is * * * that a reviewing court, in dealing with a determination or judgment which an administrative agency alone is authorized to make, must judge the propriety of such action solely by the grounds invoked by the agency.’” Id. at 169 (quoting SEC v. Chenery Corp., 332 U.S. 194, 196 (1947)). Commerce in the final results stated only that it must abide by its long standing policy, but its practice was the opposite of what it did here. See Final Results, 61 Fed. Reg. at 13,825.

Moreover, the parties have not cited to substantial evidence on the record before Commerce that certain facts of this case warrant an exception to Commerce’s past reversals practice. There also is no evidence that Commerce determined that including the reversals in the period of review COP and CV calculations did not distort costs. Accordingly, this matter is remanded to Commerce to eliminate the credit for the reversals unless it can articulate a rational reason for abandoning its past practice. Reiterating the reasons proffered in this action will be insufficient.

II. Stelco:

A. Model-Match Methodology:

Section 1677(16) of Title 19 defines “such or similar merchandise” as used in determining foreign market value (“FMV”). Pursuant to that section, Commerce must first look for the “merchandise which is the subject of an investigation and other merchandise which is identical in physical characteristics with, and was produced in the same county by the same person as, that merchandise.” 19U.S.C. § 1677(16)(A);seeaZso Hussey Copper, Ltd. v. United States, 895 F. Supp. 311, 313 (Ct. Int’l Trade 1995). Only if “such” merchandise is unavailable may Commerce use the most physically similar product to the merchandise sold in the United States. 19 U.S.C. § 1677(16); Hussey Copper, 895 F. Supp. at 313. Finally, if Commerce uses similar merchandise in determining FMVJ it must adjust for any difference between the products if such difference is wholly or partly due to physical differences. 19 C.F.R. § 353.57 (1995); Hussey Copper, 895 F. Supp. at 313. Commerce, however, has discretion to define what constitutes “such or similar” merchandise in each case. See Hussey Copper, 895 F. Supp. at 314.

In its determination of FMV( Commerce applied the following model-match methodology to Stelco’s sales. Commerce identified and ranked in the order of importance eleven product characteristics.⁹ Commerce’s Questionnaire (Sept. 15, 1994), at App. V-6 to V-9, RR. Doc. 20, Def.’s App., Ex. 2. For reported U.S. and home market sales, the program assigned a numeric value to each designated product characteristic. See Corrosion Resistant Model Match Program (Feb. 7, 1996), lines 1005-1095 (home market sales), lines 1207-1300 (U.S. sales), Def.’s Conf. App., Ex. 6. The program compared each product characteristic for a particular U.S. model with the product characteristics of a particular home market model and assigned a “dif” variable to the comparison. The “dif” variable represents the difference between the numeric values assigned to each product characteristic. See id. at lines 1350-1424. Examining the “dif” variable for each product characteristic, see id. at lines 1426-29, the program found identical matches when the resulting “dif” variables for all product characteristic comparisons were zero. Id. at lines 1430-36. The program then identified the next most similar models based upon the examination of each resulting “dif” variable combination (i.e., the smallest “dif” variable for a comparison at a higher ranked product characteristic was considered the best match based upon physical criteria). See id. at lines 1438-46.

Commerce applied a variation of this methodology for Stelco’s sales of excess prime merchandise reported with missing characteristics. All missing product characteristics were assigned the same generic value. When comparing a particular U.S. model with a particular home market model, if the U.S. model had a missing characteristic, the program assigned a value of zero to the comparison for that product characteristic (i.e., the resulting “dif” variable would be zero). See generally id. at lines 1350-1425. If the particular home market model had a missing product characteristic that was not missing in the U.S. model, the particular home market model was not used for price comparison with that particular U.S. model. Id.

Plaintiffs argue that the departure from the original model-match methodology resulted in the following practical consequences. If a U.S. sale had missing characteristics, Commerce’s program simply assumed that the missing characteristics were the same as the characteristics on any potentially matching home market sale, regardless of whether the home market sale had missing characteristics or not. Where a U.S. sale had complete characteristics, however, Commerce never matched the sale to a home market product with missing characteristics, even if the home market product was identical in all respects except for the missing characteristics.

Plaintiffs also argue that Commerce’s departure from its chosen methodology is not in accordance with law because Commerce did not provide an explanation on the record for its decision. Reliance on Stel-co ’ s allegedly unexplained failure to provide the missing data is in plaintiffs’ view, an insufficient justification to depart from the chosen methodology. Moreover, even if the departure was explained on the record, an exercise of discretion cannot justify a violation of the statutory mandate. Here, plaintiffs argue that the methodology as applied to excess prime sales with missing product characteristics violates the statute by allowing Commerce to exclude potentially identical or the next most similar sales based on the assumption that the missing characteristics were the same. Similarly, plaintiffs contend that reliance on the same assumption prevents Commerce from considering appropriate adjustments to similar sales and also adversely affects the constructed value calculation.¹⁰

The court may uphold a determination by Commerce of “‘less than ideal clarity if the agency’s path may be reasonably discerned.’” Out-okumpu Copper Rolled Prods. AB v. United States, 17 CIT 848,855,829 F. Supp. 1371,1378 (1993) (quotingBowman Transp., Inc. v. Arkansas-Best Freight Sys., Inc., 419 U.S. 281, 286 (1974)). The agency’s determination, however, must still “‘disclose the basis of its order’ and * * * must articulate a rational connection between the facts found and the choice made. ” Id. (citation omitted). Commerce did not explicitly articulate the reason for its departure from its original methodology. It did, however, make an implicit decision based on more than Stelco’s allegedly unexplained inability to provide the relevant information, that under these circumstances, the methodology used was reasonable.

Commerce’s decision that the missing product characteristics were not commercially meaningful for the few sales of excess prime in question and thus not required for the model-match methodology' is implied in its position articulated in the Final Results. 61 Fed. Reg. at 13,831. Commerce stated:

[t]he few prime sales Stelco made that did not have complete physical characteristics were orders for which the mill order number had been lost. The Department verified that Stelco designates prime sales lacking complete characteristics as excess prime sales before the product is sold. Stelco then finds customers for this merchandise. Although the material in question is prime, Stelco reported and the Department verified that it is sold at a reduced price, and in the vast majority of cases to distributors. While this merchandise is not defective, full and complete physical characteristics were not needed to make the sale to the customer. The end uses of such material are applications for which knowledge of certain of the product’s characteristics was unimportant.

Therefore, because Stelco sold this merchandise in both markets, because the missing physical characteristics were not important to Stelco’s customers and because we verified that respondent reported all physical characteristics it could, the Department matched this merchandise based on the limited physical characteristics reported. Since these were the only physical characteristics relevant to the way the product was sold, we conclude that we may make appropriate matches on the basis of only these physical characteristics in this limited circumstance.

Id. at 13,830-31 (emphasis added). While Commerce could have articulated a more precise determination explaining why various characteristics previously ranked by order of importance were no longer important, this generalized determination is sufficient. The facts relied upon by Commerce are clearly articulated and are rationally connected to its choice to proceed with the model-match methodology even without complete product characteristics. Thus, Commerce did not abuse its discretion in departing from the original model-match methodology.

Also relevant to Commerce’s decision is the fact that the number of sales-with missing characteristics is relatively small. In the home market, Stelco’s misreporting affected less than 5%11 of total home market sales.¹² Stelco’s Final Computer Database for Corrosion-Resistant Sales (April 14,1995), C.R. Doc. 169, Stelco’s App., Tab 12; Stelco’s Chart, at 1. Moreover, the misreporting affected all characteristics in Commerce’s model-match hierarchy for less than .4% of prime merchandise out of the total tons sold. Stelco’s Final Computer Database for Corrosion-Resistant Sales, C.R. Doc. 37, Stelco’s App., Tab 12; Stelco’s Chart, at 1. In the U.S., the misreporting affected .4% of Stelco’s total U.S. sales.¹³ Id. In addition, while Stelco’s misreporting affected five different product characteristics, including quality and strength, the fourth and fifth most important characteristics in Commerce’s hierarchy, no sale was missing more than three characteristics. See Stelco’s Final Computer Database (U.S. Observation No. 705), C.R. Doc. 169, Pis.’ App., Tab 20.

Moreover, Commerce’s chosen methodology as applied to excess prime sales with missing product characteristics is in accordance with law. The generic value assigned to all missing product characteristics does not reflect an assumption that all missing characteristics were the same. The generic value assigned reflects Commerce’s implicit decision that in this limited situation, those characteristics were not important to the end user and thus were not needed for the model-match exercise. In addition, plaintiffs have not demonstrated through specific examples how the model-match as applied prevented Commerce from making appropriate adjustments or adversely affected the constructed value calculation.

Plaintiffs also argue that Commerce’s factual findings in the Final Results were not supported by substantial evidence on the record. Substantial evidence has been defined as “‘such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.’” Usinor Sacilor v. United States, 18 CIT 1155, 1156, 872 F. Supp. 1000, 1003 (1994) (quoting Universal Camera Corp. v. NLRB, 340 U.S. 474, 477 (1951)). Moreover, Commerce’s decisions are not constrained by the inferences the parties make from the factual record. Bówe Passat Rein-igungs-Und Waschereitechnik v. United States, 951 F. Supp. 231, 235 (Ct. Int’l Trade 1996). Rather, “Commerce may, based on its experience in administering the statute and regulations, make justifiable inferences on the record before it.” Id.; see also Matsushita Elec. Indust. Co. v. United States, 750 F.2d 927, 933 (Fed. Cir. 1984) (reviewing whether “the evidence and reasonable inferences from the record support the [ITA finding].”).

Plaintiffs argue that the Final Results which included the following statements are unsupported by substantial evidence on the record and are contradicted by Stelco’s statements in its questionnaire response regarding excess prime and secondary products: (1) “because of the way Stelco keeps its records * * * it could not report the full physical characteristics”; (2) “Stelco sold this merchandise in both markets”; (3) the sales in question constitute only a small portion of total sales; and (4) full and complete product characteristics were not needed to complete sales of excess prime merchandise. Final Results, 61 Fed. Reg. at 13,831.

After reviewing the record, the court finds that Commerce’s findings are supported by substantial evidence. First, Commerce’s finding that Stelco reported all of the information that it could is supported by Commerce’s verification of Stelco’s sales. Commerce’s verification report, specifically the section on secondary merchandise, noted that “[djuring the verification, we observed the tracking and recordkeeping that Stelco employs during the steel production process.” Sales Verification Report of Stelco (May 2,1995), at 7, PR. Doc. 181, Def.’s Conf. App., Ex. 4. Moreover, the record includes multiple references to the importance of the mill order number. See, e.g., Stelco’s Section IV Response (Nov. 14, 1994), at 3, PR. Doc. 116, Stelco’s App., Tab 13 (“Stelco can fully identify the product characteristics of a particular sale only through the prime order number.”); Stelco’s Supplemental Questionnaire Response (Jan. 10,1995), at 30-32, PR. Doc. 210, Stelco’s App., Tab 14. While these references generally are in the context of discussing secondary merchandise and not excess prime, this evidence when combined with Commerce’s verification report, supports a finding that the information was not withheld wrongfully.

Second, the fact that Stelco sold excess prime in both markets is evidenced by the computer database and in the chart presented by Stelco at oral argument. Stelco’s Final Computer Database for Corrosion-Resistant Sales, C.R. Doc. 169, Stelco’s App., Tab 12; see also Stelco’s Chart, at 1. Third, the same chart and computer database clearly indicate that only a small portion of Stelco’s total sales were reported without full product characteristics. Stelco’s Final Computer Database for Corrosion-Resistant Sales, C.R. Doc. 169, Stelco’s App., Tab 12; Stelco’s Chart, at 1.

Moreover, Commerce’s findings that full and complete characteristics were not necessary for the sale of excess prime merchandise is also supported by substantial evidence. The record indicates that “[ejxcess prime merchandise differs from prime merchandise in that it has not been produced pursuant to a given customer’s specifications. Rather, it exists in stock and is sold ‘as is.’” Stelco’s Supplemental Questionnaire Response (Nov. 21,1994), at 26, P.R. Doc. 137A, Def.’s App., Ex. 4. Excess prime merchandise is generally sold at a reduced price¹⁴ to customer service centers that have the ability to process material and apply it to applications in their own customer base. Stelco’s Secondary Material Supplemental Questionnaire Response (Feb. 22,1995), at 8-9, C.R. Doc. 112, Stelco’s App., Tab 16. Thus, excess prime is material “which is not suitable for the original order in terms of size or mechanical properties, but which would still be suitable for another end use. ” Id. at 2. From this evidence, it was reasonable for Commerce to infer that for each individual sale of excess prime merchandise knowledge of all eleven product characteristics was not needed.

Plaintiffs also argue that Stelco’s Response to Commerce’s Supplemental Questionnaire contradicts most of Commerce’s findings.¹⁵ Stel-co’s response only contradicts Commerce’s findings when taken out of context. Stelco’s statement is part of a response to Commerce in which Stelco made arguments concerning how Commerce should treat excess prime sales with complete characteristics in its model-matching exercise. Thus, Stelco’s response does not undercut Commerce’s findings of fact that support its conclusion of how to treat excess prime sales without complete characteristics.

Thus, plaintiffs’ cites to the record do not provide direct contradictions to the facts found by Commerce. At most, they suggest that alternative inferences could have been made from the facts. The court, however, does not function as a finder of fact in this instance and Commerce is allowed to draw inferences that contradict the inferences drawn by the parties. As Commerce’s findings were supported by substantial evidence on the record, the court affirms the final results.

Finally, plaintiffs also argue that Commerce failed to follow its statutory mandate of applying a reasonably adverse inference to Stelco’s sales when a respondent refuses, without explanation, to report complete product characteristics and accurate difference in merchandise (“difmer”) data for every sale. Plaintiffs contend that respondents may have benefitted by Commerce conducting the model-match with the available, yet incomplete, information provided by Stelco. Moreover, plaintiffs argue that the lack of complete product characteristics effected Commerce’s ability to determine actual costs and thus, the surrogate of the average cost of production for all products having the same characteristics applied by Stelco was unreasonable.

Commerce responds that application of best information available (“BIA”) was not appropriate because the model-match methodology properly matched sales based on the information reported. Moreover, Commerce noted that Stelco could not report full product characteristics for the few sales at issue, but did provide all of the product characteristics that it could. In addition, Commerce noted both Stelco’s cooperation and the relatively few number of sales in reaching its determination that BIA was inappropriate.¹⁶

Section 1677e(c) of Title 19 states that Commerce shall:

In making their determinations under this subtitle, the administering authority and the Commission shall, whenever a party or any other person refuses or is unable to produce information requested in a timely manner and in the form required, or otherwise significantly impedes an investigation, use the best information otherwise available.

19 U.S.C. § 1677e(c) (emphasis added). “It is well established * * * that Commerce has broad discretion with regard to when the use of BIA is appropriate,” Timken Co. v. United States, 18 CIT 486,489,852 F. Supp. 1122, 1125 (1994), and that the court must grant Commerce considerable deference in its choice to apply or not to apply BIA. Al Tech Specialty Steel Corp. v. United States, 947 F. Supp. 510, 523 (Ct. Int’l Trade 1996).

The power to use BIA against recalcitrant parties cannot, however, be wielded arbitrarily. Olympic Adhesives, Inc. v. United States, 899 F.2d 1565,1572 (Fed. Cir. 1990). Thus, Commerce may not, as plaintiffs argue, characterize a party’s failure to provide information that does not exist as a “refusal” to provide data. Id. at 1573. Moreover, an incomplete response does not automatically sanction the application of BIA. Commerce may consider both the degree of cooperation by the respondent and the size of the omission in reaching its decision to either apply BIA or accept the available information. See Al Tech Specialty Steel, 947 F. Supp. at 523 (affirming application of neutral BIA in form of overall weighted-average calculated margin to account for unreported sales based on minimal omission and substantial compliance by respondent); Usinor Sacilor v. United States, 907 F. Supp. 426, 429 (Ct. Int’l Trade 1995) (application of neutral BIA appropriate when respondent substantially complies with request for information and omission arises from factors beyond control).

The facts of this case support Commerce’s decision to not apply BIA and to rely instead upon both the slight variation of the original model-match methodology and a surrogate average cost of production for all products having the same characteristics. Commerce verified that Stel-co had reported all of the information available. See Olympic Adhesives, 899 F.2d at 1573 (failure to provide information that does not exist does not constitute a “refusal” to comply). Notwithstanding Commerce’s verification, plaintiffs argue that the information would have been kept in the normal course of business¹⁷ and thus its absence is a result of an intentional refusal to provide information. Plaintiffs, however, have not presented evidence which would undermine Commerce’s finding. While it is true that excess prime sales normally would not lose their mill order numbers, the lack of mill order numbers on a few¹⁸ excess prime sales by itself does not demonstrate intentional conduct by Stelco to withhold information. Stelco’s Final Computer Database for Corrosion-Resistant Sales, C.R. Doc. 169, Stelco’s App., Tab 12; Stelco’s Chart, at 1. Rather, it equally supports Commerce’s implicit acceptance of Stelco’s statement that the missing product characteristics were lost inadvertently as Stel-co was able to report the majority of the information requested.

In addition, Stelco cooperated with Commerce throughout the investigation. Plaintiffs argue, however, that with respect to sales of excess prime merchandise, Stelco was not cooperative because it did not notify Commerce of the sales with missing characteristics until its Rebuttal Brief. While it is true that all previous discussions of sales with missing characteristics involved secondary merchandise or non-prime sales, the discussions did put Commerce on notice of the significance of the mill order number. Stelco repeatedly stated that if a sale lost its mill order number, it would be unable to provide complete product characteristics for that sale. See Stelco’s Section IV Response, at 2, ER. Doc. 116, Stel-co’s App., Tab 13; Stelco’s Supplemental Questionnaire Response, at 30-32, ER. Doc. 210, Stelco’s App., Tab 14. Logically, the same result would occur when an excess prime sale loses its mill order number.

Moreover, it is difficult to understand how such a small omission could have impeded the investigation and justified an application of adverse BIA. For U.S. sales of excess prime, only a few sales were missing product characteristics. See Stelco’s Chart, at 1. For home market sales, less than .4% by volume were missing information on all eleven characteristics. See Stelco’s Final Computer Database, C.R. Doc. 169, Els.’ App., Tab 20; Stelco’s Chart, at 1. Thus, the size of the omission combined with Stelco’s substantial compliance justify Commerce’s decision to not apply adverse BIA.¹⁹

B. Clerical or Billing Error Adjustments to Price:

There is no dispute between the parties as to the basic principles applicable to price adjustments in this case. Erice adjustments must be transaction specific. See SKF USA Inc. v. United States, Slip Op. 95-85, at 16-17,1995 WL 283855, at *7-8 (Ct. Int’l Trade May 8,1995); Anti-friction Bearings (Other Than Tapered Roller Bearings), and Parts Thereof from France, etal, 61 Fed. Reg. 66,472,66,498 (Dep’t Commerce 1996)(final results of admin, review)[hereinafter “AFB’s 1996”]. Adjustments are sufficiently transaction specific when they are directly attributable to individual sales or when they are made across several transactions but are allocable on a fixed percentage of sales basis or some other constant rate. SKF USA, Inc., Slip Op. 95-85, at 16-17,1995 WL 283855, at *7-8; AFB’s 1996, 61 Fed. Reg. at 66,498.²⁰

The adjustments at issue are not for widely available discounts, rebates or similar items which may or may not be determinable on a fixed or constant basis across numerous sales. Rather they are for generally variable adjustments for clerical or other billing errors. Such adjustments must be related on the record to specific transactions, either directly or through proper allocation. Eresumably because the adjustments were not across the board rebates or discounts they were not placed in a separate computer field, but instead were included within the gross unit price. See Antidumping Questionnaire (Sept. 15,1995), at App. I.B-8 to B-9,1.C-8 to C-10, App. II — 2, ER. Doc. 20; Stelco’s App., Tab 3. An indication was given, however, for each sales to which an adjustment was made, i.e., “C” for credit, “D” for debit or “B” for both. Stelco’s Supplemental Questionnaire Response (Jan. 10, 1995), at 46-47, ER. Doc. 210, Els.’ App., Tab 6; Stelco’s Fost-Verification Corrections to the Corrosion-Resistant Database (Apr. 13, 1995), at Exs. 1-3, ER. Doc. 381, Els.’ App., Tab 12. Less than 8% of sales had such notations. See Stelco’s Final Computer Database, ER. Doc. 169, Stelco’s App., Tab 12.

Where an adjustment references one invoice covering one sale plaintiffs raise no objection. The objection is to adjustments which are related to multiple invoices or to invoices covering several sales. Stelco and plaintiffs agree that Stelco’s records enabled it to relate adjustments not just to individual invoices but to specific transactions within invoices. They disagree as to whether this relationship was demonstrated to Commerce.

The problem is that Commerce necessarily examined a limited number of sales. It examined just twenty corrosion resistant sales. Sales Verification Report for Stelco (May 2,1995), at 17-18, ER. Doc. 403, Stelco’s App., Tab 6. Of those, Commerce examined three billing error adjustments that affected a total of five transactions. See Sales Trace Verification Ex. 37 (May 2,1995), C.R. Doc. 37, Stelco’s App., Tab 11. Flaintiffs have not demonstrated that any of the three contained unallowable allocations or were not transaction specific. Flaintiffs point to one adjustment for cut-to-length plate steel that was improperly cross referred to sales. See Stelco Sales Verification Ex. 37 (May 5, 1995)(Home Market Observation 14220), C.R. Doc. 180, Els.’ App., Tab 21. Apparently, Commerce did not consider the plate adjustment indicative of Stelco’s normal methodology with respect to corrosion resistant steel. This conclusion is reasonable. There is no evidence that the plate allocation error represented Stelco’s normal practice for billing error adjustments for corrosive resistant steel, which had a separate billing department.²¹ See Sales Verification Report (May 2, 1995), at 2, C.R. Doc. 181, Def.’s Conf. App., Ex. 4.

While there is some confusion as to whether Commerce found the adjustments directly transaction specific or merely found an acceptable allocation to specific sales, it is clear that Commerce accepted the adjustments. Of the three corrosion adjustments that were verified, two involved multiple invoices, but it appears that the single debit or credit adjustment involved in each was properly related to each referenced invoice. See Sales Trace Verification Exhibits (Observation Nos. 642, 1850, & 1860), C.R. Doc. 37, Stelco’s App., Tab 11. While Commerce seems to find these examples not transaction specific, the adjustments at least do appear to be proper allocations. When dealing with limited adjustments applicable to a very few invoices the distinction between “directly transaction specific” and “properly allocated” seems to blur.

Finally, it should be noted that 15% of Commerce verified sales samples contained the adjustments at issue. See Sales Verification Report for Stelco, at 17-18, ER. Doc. 403, Stelco’s App., Tab 6; Stelco’s Sales Verification Exs., Stelco’s App., Tab 7. As noted previously, less than 8% of the total contained these adjustments. Thus, the court concludes that Commerce conducted a reasonable investigation with respect to this issue and no error has been demonstrated, other than some less than clear language in the Commerce determination.

III. Continuous Colour Coat:

A. Debit and Credit Adjustments to Prices:

While the Commerce determination may have been less than clear as to Stelco’s adjustments to price, the court was able to discern from the record that the adjustments were proper. This is not the case with CCC’s post-invoice debits and credits covering multiple invoices. Commerce is simply wrong in its implicit conclusion that the specificity at issue is the ability of the respondent to identify a particular debit or credit. See Final Results, 61 Fed. Reg. at 13,822. Obviously, the specificity required is with respect to sales transactions and the ability to relate or properly allocate a debit or credit to such transactions. Otherwise a respondent could apply any debit or credit it could identify to any sale in its entire sales base.

None of the parties have pointed to evidence in the record that makes clear that Commerce either did or did not act correctly in this regard. Defendant states that the adjustments should be upheld because plaintiffs had the burden of demonstrating error. While this may ordinarily be the case, when Commerce so clearly misstates the law, remand is the better option. Commerce shall indicate where in the record the debits and credits noted are shown to be properly related either directly or through allocation to specific sales transactions. If Commerce determines an acceptable level of specificity has been achieved or may otherwise be overlooked because a few such adjustments were made to the “same customer” within a “limited period,” id., Commerce shall indicate where this is factually supported in the record and why these facts are relevant.

Conclusion

The matter is remanded to Commerce for 45 days to correct ministerial errors in Stelco’s margin calculation and for reconsideration of Dofas-co’s partial reversal of restructuring charges and CCC’s error adjustments to price. Objections may be filed 11 days thereafter and any response within 5 days.

1 Pre-Uruguay Round Agreements Act (“URAA") law applies to this administrative review. Amendments to United States antidumping statute and regulation pursuant to the URAA are inapplicable to administrative reviews initiated prior to January 1,1995. Torrington Co. v. United States, 68 F.3d 1347, 1352 (Fed. Cir. 1995).

2 Dofasco offers an alternative statutory interpretation. Dofasco relies upon section 1677b(b) of Title 19 which states:

[if Commercel has reasonable grounds to believe or suspect thatsa/es in the home market of the country of exportation + K * have been made at prices which represent less than the cost of producing the merchandise in question, it shall determine whether, in fact, such sales were made at less than the cost of producing the merchandise.

19 U.S.C. § 1677b(b)(emphasis added). Dofasco argues that the “merchandise in question” relates to the “sales in the home market. ” Thus, when referring to the cost of producing “the merchandise, ” the statute plainly means the merchandise in question sold in the home market. Dofasco raises a similar argument for calculating CV The “merchandise” in 19 U.S.C. § 1677b(a)(2) is said to refer to the “imported merchandise” or the merchandise sold in the U.S. and thus CV is allegedly based on the COM of the merchandise sold in the United States.

3 Commerce verified Dofasco’s reported COP and CV figures. Commerce verified that the information contained in Dofasco’s sales orders used in conjunction with its quarterly grade cost table, allowed Dofasco to calculate precise production costs for any product. Dofasco’s Cost Verification Report (May 12,1995), at3,ER. Doc.417, Def.’s App., Ex. 13. Commerce noted that the production costs were based upon the known physical and chemical properties of the material produced, the line time each production stage required on a unit basis, the yield factor of each production process, and the beginning and ending weight of the material at each stage of the production process. Id. Commerce also noted that the grade cost table for slabs was a collection of costs which Dofasco updated on a quarterly basis to reflect the costs of producing slabs based upon raw material costs, processing cost, and overhead. Id.

4 “CONNUM” is the computer field name assigned to each unique product or each grouping of products sold that is considered identical for product matching purposes.

5 Plaintiffs cite to examples in tlie record to support their argument that costs for merchandise considered identical should not have differing COMs. In CONNUM 0101, plaintiffs argue that Dofasco’s methodology resulted in one COM for COP C$n/cwt. and a different COM for CV, C$r 1/cwt., a difference off] percent. Petitioners’ Case Brief for Dofasco (Sept. 15,1995), at 5, C.R. Doc. 228, Pis.’App., Tab 26 (citing Dofasco’s May 5,1995 cost computer tape). In CONNUM 0275, plaintiffs also note that Dofasco reported a COM for COR C$[ 1/cwt. and a different COM for CV, C$r 1/cwt., a difference of [ 1 percent. Id.

6 Plaintiffs incorrectly argue that any rationale which legitimizes Dofasco’s methodology because it is “more precise” is an impermissible post hoc rationalization and should be disregarded. Where Commerce’s rationalization is described in or supported by the record, as is the case here, it is not a post hoc rationalization presented for the first time in briefs before the court. See Burlington Truck Lines, Inc. v. United States, 371 U.S. 156,168-69 (1962).

7 In 1993, Dofasco reduced its 1992 costs by C$r 1; in 1994, it reduced its costs by C$n. Dofasco’s Supplemental Section VI Response (Mar. 31,1995), at Ex. 1, RR. Doc. 364, Pis.’ App., Tab 11.

8 Commerce reached a similar conclusion in Certain Cut-To-Length Steel Plate from Germany, 61 Fed. Reg. at 13,837. There, the original accrual occurred before the POI and respondent sought to include the subsequent reversal in the COM calculation. Id. Commerce noted that:

ftlhere is not justification for distorting actual production costs incurred in a subsequent year by reducing subsequent year costs by the overestimated amount.

9 These characteristics were; (1) type, (2) metallic coating process, (3) clad matenal/coating metal, (4) quality, (5) yield strength, (6) metallic coating weight, (7) minimum thickness, (8) width, (9) form, (10) temper rolling, and (11) leveling. Commerce’s Questionnaire (Sept. 15, 1994), at V-6 to V-10, PR. Doc. 20, Def.’s App., Ex. 2.

10 Plaintiffs also argue that Commerce violated the statute by ceding to Stelco its responsibility of determining which sales were identical or next most similar. See Timken Co. v. United States, 10 CIT 86,98,630 F. Supp. 1327,1338 (1986) (Commerce criticized for allowing respondent to report home market sales based upon respondent’s determination of such or similar merchandise rather than collecting all home market sales data and making own determination). Specifically, Commerce allegedly ceded its responsibility by altering its methodology based solely on Stelco’s unexplained failure to provide complete product information and thus Stelco was able to control the outcome of the model-match exercise The court finds, however, that by selecting the matching characteristics and requesting the information from Stelco, Commerce controlled the exercise. Moreover, plaintiffs’ argument relies upon a finding that Stelco’s lack of information was unexplained and intentional. As demonstrated below, this finding has not been made.

11 Stelco reported f 1 excess prime sales with missing characteristics out of a total of H prime sales. This information is from Stelco’s Final Computer Database (Apr. 13,1995), C.R. Doc. 169, and was summarized in a chart presented by Stelco at oral argument before the court. The parties stipulated to the facts contained in the chart.

12 Stelco asserts that 90 percent ((" 1 out of r 1) of the home market prime sales without full characteristics were sold at prices less than the cost of production. Thus, Stelco argues that “virtually none of these home market sales would have been used in the dumping calculation even if complete characteristics had been reported, and therefore, these home market sales essentially were irrelevant to the Department’s antidumping analysis.” Stelco’s Response Brief, at 26.

Plaintiffs persuasively demonstrate the flaw in Stelco’s logic. Commerce must base foreign market value on constructed value when above-cost sales of such or similar merchandise are inadequate. See 19 C.F.R. § 353.51(b)(l994). If as Stelco argues, most of the home market sales with missing characteristics were below cost and thus would require constructed value comparisons, the failure to include these sales in the dumping calculation would make the distortion of the dumping margin greater.

13 Stelco reported sales with missingproduct characteristics for f 1 out of T1 excess prime sales and [ ] out off 1 prime sales. See Stelco’s Final Computer Database (Apr 13,1995), C.R. Doc. 169, Pis.’ App., Tab 20; Stelco’s Chart, at 1.

14 Stelco stated on the record that selling excess prime at a reduced price has nothing to do with the physical characteristics of the merchandise Stelco’s Secondary Merchandise Supplemental Response (Feb. 22, 1995), at 3,C.R. Doc. 112, Stelco’s App., Tab 16. Plaintiffs argue that Stelco’s statement contradicts the factual finding of Commerce. While plaintiffs may be correct that the reduced price is unrelated to the physical characteristics of the merchandise and thus may not provide a sound reason for departing from the original methodology, Commerce did not base its decision to use the available information on this fact alone.

15 Plaintiffs rely on the following statement in Stelco's response to Commerce’s supplemental questionnaire to undercut the factual findings:

The model-match exercise seeks to compare various physical characteristics to determine which home market products are identical or most similar to the U.S. products. The focus and analysis of the model-match exercise is based solely on the physical characteristics. To the extent the customer type and conditions of sale are relevant considerations, they are i*elevant to circumstances of sale only and not to product classification.

There are no physical differences between merchandise identified as “excess prime” and prime merchandise. Stel-co’s designation “excess prime” simply means that the particular product was not applied (for whatever reason) to the original order. The fact that merchandise designated “excess prime” is sometimes sold at a lower price than identical prime merchandise has absolutely nothing to do with the physical characteristics of the merchandise. Consequently, no justification exists for making the distinction between excess prime and prime for model-match purposes.

Stelco’s Secondary Merchandise Supplemental Response (Feb. 22,1995), at 3, C.R. Doc. 112, Stelco’s App., Tab 16.

16 Commerce articulated its position as.

The use of BIA is not appropriate m this case, because Department methodology properly matches sales based on the information Stelco reported. The Department verified that because of the way that Stelco keeps its records Stelco could not report the full physical characteristics of the small number of sales in question. Petitioners’ reference to AFB’s from France is not precisely relevant, because in that case, the Department used the BIA cited by petitioners as total BIA for companies that either failed to respond to the Department’s questionnaire or were unable to complete verification. In this case, Stelco cooperated with the Department and provided all the product matchingphysical characteristics that it could report. In addition, the Department could use the information that Stelco provided for matching purposes. Consequently, the use of total BIA in this circumstance is unwarranted.

Final Results, 61 Fed. Reg. at 13,831.

17 Stelco maintains two sales processing systems: Cl) a “mill order” system which is used to both manufacture and sell prime merchandise, and (2) a “stock” order entry system which is used solely to ship and sell merchandise, usually the merchandise that does not meet the specifications for prime merchandise. Stelco’s Supplemental Questionnaire Response, at 30-32, ER. Doc. 210, Stelco’s App., Tab 14. Stelco can fully identify the product characteristics of a particular sale only through the prime order number (or mill order number). Stelco’s Section IV Response, at 3, ER. Doc. 116, Stelco’s App., Tab 13. Whether the mill order number is retained or not is usually determined by when the steel becomes non-prime. Stelco’s Supplemental Questionnaire Response, at 31-32, ER. Doc. 210, Stelco’s App., Tab 14. It remains if the decision to treat the steel as non-prime occurs after the steel has left the production line. Id. Generally, because excess prime is never downgraded to non-prime, it would normally retain its mill order number and Stelco would be able to report full product characteristics.

18 For example, f 1 out of [ 1 U.S. excess prime sales were missing their mill order numbers. Stelco’s Chart, at 1.

19 Plaintiffs rely Gray Portland Cement and Clinker from Mexico, for the proposition that Commerce must apply an adverse inference where a respondent fails to supply certain information such as complete product characteristics and accurate difmer data. Final Results of Redeterm motion Pursuant to Court Remand, Cemex, S.A. v. United States, Slip Op. 95-72 (April 24,1995), remand determination remanded on other grounds, Cemex, S.A. v. United States, Slip Op. 96-132 (Aug. 13,1996). Although Commerce did apply adverse BIA to sales for which respondent did not report difmer data in Gray Portland Cement, plaintiffs fail to note that Commerce applied BIA in part because the respondent withheld requested data and presented conflicting evidence concerning the cost differences of products. Here, however, Commerce verified that Stelco was not withholding information and determined that the information available was useable.

20 The parties quibble as to whether these are recognized departures from the fixed percentage or constant rate principle of allocation in certain fact situations, e.g., when the departures are insignificant, but they do not argue about the general rule.

21 Furthermore, the problem encountered in the plate example may not occur with regard to corrosion-resistant steel. The parties agree there were either none or very few single invoices covering multiple transactions. Thus, the plate adjustment fact scenario was unlikely to occur with respect to corrosion-resistant steel.