This case involves the determination of antidumping duties for canned pineapple fruit ("CPF") from Thailand. Thai Pineapple Canning Industry Corporation ("TPC") raises two issues on appeal from the Court of International Trade. First, TPC challenges the methodology used in this case by the Department of Commerce ("Commerce") to determine cost of production and constructed value. The question is what is the proper methodology in computing antidumping margins during a period of rising costs when there is a significant lag time between the production of merchandise and its sale. Second, TPC contends that during an administrative review of an antidumping order, Commerce should not compute a single assessment rate for the entire period of review, but instead should compute two assessment rates, one for the period between Commerce's preliminary determination of sales at less than fair value and the International Trade Commission's affirmative injury determination, and a second rate for the remainder of the period of review after the Commission's injury determination. Because we conclude that Commerce's cost of production methodology is not reasonable in this case, we reverse the Court of International Trade's decision on that issue. We affirm the decision regarding the use of a single assessment rate for the entire period of review.

Commerce is required to impose an antidumping duty on imported merchandise that is being sold, or is likely to be sold, in the United States at less than fair value to the detriment of a domestic industry. 19 U.S.C. 1673 (1994). Commerce determines the duty by calculating the dumping margin, i.e., the amount by which the "normal value" (typically the price charged for similar merchandise sold for consumption in the exporting country) exceeds the "export price" or "constructed export price" (the price charged for the subject merchandise in the United States). Id. 1677(35)(A). Commerce uses the dumping margin as "the basis for the assessment of . . . antidumping duties on entries of merchandise covered by the [antidumping] determination and for deposits of estimated duties." Id. 1675(a)(2)(C).

Normal value preferably is based on sales of the foreign like product in the exporting country. Id. 1677b(a)(1)(B). When no foreign like product is sold for consumption in the exporting country, or the quantity sold is insufficient to permit a proper comparison with sales in the United States, normal value may be based on the price at which the foreign like product is sold in a third country. Id. 1677b(a)(1)(B), (C). Regardless of whether normal value is based on sales in the exporting country or in a third country, Commerce may undertake a below-cost sales investigation whenever it has reasonable grounds to believe that those sales have been made at prices less than the cost of production. Id. 1677b(b). If Commerce determines that sales below the cost of production have been made within an extended period of time in substantial quantities, it may disregard such sales when determining normal value, which then will be based on the remaining sales of the foreign like product in either the exporting country or third country. Id. If no sales remain, normal value will be based on the constructed value of the merchandise. Id. 1677b(b), (e).

Cost of production is calculated according to a statutory formula by adding together several costs and expenses, including the cost of materials, fabrication, containers, coverings, and other processing costs, and selling, general, and administrative expenses. Id. 1677b(b)(3). The constructed value of merchandise, which is the basis for normal value when there are insufficient sales in the exporting country or a third country, is the sum of the same costs and expenses used to calculate cost of production, plus realized profits. Id. 1677b(e). When normal value is based on sales in the exporting country or a third country, the effect of a higher cost of production is a higher normal value because more lower-priced sales will be disregarded. When normal value is based on constructed value, the effect of a higher cost of production is also a higher normal value because cost of production is the main component of constructed value. Thus, regardless of whether normal value is based on actual sales or constructed value, a higher cost of production brings about a higher normal value, which in turn creates a higher dumping margin. It is therefore advantageous to a foreign producer to demonstrate as low a cost of production as possible.

On July 18, 1995, Commerce published an antidumping duty order on CPF from Thailand. Antidumping Duty Order on Canned Pineapple Fruit from Thailand, 60 Fed. Reg. 36,775 (July 18, 1995). On August 15, 1996, Commerce initiated an administrative review of sales during the period from January 11, 1995, through June 30, 1996. Initiation of Antidumping and Countervailing Duty Administrative Reviews and Request for Revocation, 61 Fed. Reg. 42,416 (Aug. 15, 1996). As part of its review, Commerce initiated a below-cost sales investigation to determine whether TPC had sold foreign like product at prices below the cost of production.

During the administrative review, Commerce determined that TPC did not sell a sufficient amount of foreign like product in Thailand and therefore based normal value on sales prices in Germany, TPC's largest third country market. Canned Pineapple Fruit From Thailand; Preliminary Results and Partial Termination of Antidumping Duty Administrative Review, 62 Fed. Reg. 42,487, 42,488 (Aug. 7, 1997) ("Preliminary Results"). Commerce found that TPC made third country sales of some CPF products at prices below the cost of production and therefore excluded those sales from its normal value determination. Id., 62 Fed. Reg. at 42,491. For those products with no remaining third country sales, Commerce used constructed value as the basis for normal value. Id.

In determining the cost of production, Commerce calculated a single average cost for the entire period of review. During the comment period following the Preliminary Results, TPC argued that a single average cost of production distorted the price comparison between normal value and the United States sales price in two ways. Notice of Final Results of Antidumping Duty Administrative Review: Canned Pineapple Fruit From Thailand, 63 Fed. Reg. 7392, 7399 (Feb. 13, 1998) ("Final Results"). First, because the cost of fresh pineapple increased substantially during the period of review,1 sales early in the period appeared to be below cost, while sales late in the period appeared to have high profit margins. Thus, TPC argued, Commerce should calculate separate costs of production for different fiscal years. Second, merchandise is held in inventory before sale,2 so TPC argued that the assignment of fiscal year costs to sales should take into account the average inventory period. To enable that assignment, TPC had submitted cost data for 1994, which was prior to the period of review. Commerce responded that it departs from its normal methodology of calculating a single weighted average cost of production only "in unusual cases where there are substantial changes in cost, e.g., cases involving high-inflation economies." Id. Commerce found that the fluctuations in the cost of pineapple did not warrant special treatment. Id., 63 Fed. Reg. at 7400.

TPC appealed to the Court of International Trade, challenging Commerce's calculation of a single average cost of production for the entire period of review. Thai Pineapple Canning Indus. Corp. v. United States, No. 98-03-00487, 1999 WL 288772 (Ct. Int'l Trade May 5, 1999) ("TPC I"). The Court of International Trade noted a number of cases in which Commerce had adjusted its methodology for changes in costs over the period of review or had matched costs to sales more accurately than in this case. Given the distortions in cost calculations caused by the price increase of fresh pineapple and the delay between production and sale of goods, the court remanded to Commerce with instructions to revisit the issue. Id. at *4. Specifically, the court directed Commerce to

reanalyze the data to determine whether TPC has provided sufficient data to match costs to appropriate fiscal year sales. If it has, in the absence of any proper antidumping policy reason for not doing this seemingly minimally burdensome and substantially less distortive comparison, Commerce must proceed as it has in the past and match fiscal year costs with sales.

Id.

On remand, Commerce recalculated separate costs for the 1995 and 1996 fiscal years and then matched 1995 costs to 1995 sales and 1996 costs to 1996 sales. Final Results of Redetermination Pursuant to Court Remand: Thai Pineapple Canning Industry Corp. v. United States, No. 98-03-00487, slip op. at *11 (Sept. 2, 1999) (unpublished, filed with the Court of International Trade) ("Remand Results"). In response to comments from TPC that matching sales to costs in the same time period continued to produce distorted results, Commerce reiterated its general policy of using the cost of producing merchandise during the period of review rather than the cost of producing the merchandise sold during the period of review. Id. at *27. Commerce conceded that the statute, which provides that costs are determined "during a period which would ordinarily permit the production of [the merchandise] in the ordinary course of business," 19 U.S.C. 1677b(b)(3)(A), 1677b(e)(1), is broad enough to allow for either approach. Nevertheless, Commerce stated that it will not stray from its normal practice except in cases with unique circumstances, which it maintained are not present in this case. Remand Results at *27-28.

TPC appealed again to the Court of International Trade, which this time agreed with Commerce. Thai Pineapple Canning Indus. Corp. v. United States, No. 98-03-00487, 2000 WL 174986, at *3-5 (Ct. Int'l Trade Feb. 10, 2000) ("TPC II"). The court concluded that the use of separate 1995 and 1996 costs sufficiently accounted for the rising pineapple costs and complied with its remand instruction, and the court determined that the facts of this case did not warrant further modification of Commerce's methodology. Id. at *5. TPC appeals that decision to this court.

Once Commerce makes a preliminary determination that merchandise is being sold in the United States at less than fair value, Commerce orders the posting of a cash deposit or bond for each entry of merchandise at a rate based on the preliminary estimated dumping margin. 19 U.S.C. 1673b(d)(1)(B) (1994). After Commerce makes a final determination that the subject merchandise is being sold at less than fair value, id. 1673d(a), and the International Trade Commission makes a final determination that an industry is materially injured, id. 1673d(b), Commerce publishes an antidumping order with a new assessment rate based on the dumping margin determined during Commerce's investigation. Id. 1673e(a). For entries of merchandise after Commerce's affirmative preliminary determination and before the Commission's affirmative injury determination, if the deposit of estimated duty under 1673b(d)(1)(B) is higher than the duty under the antidumping order, the difference is refunded. 19 U.S.C. 1673f(a) (Supp. V 1999). On the other hand, if the deposit is lower than the duty under the order, the difference is disregarded. Id. That is, the preliminary estimated duty acts as a "cap" on the duty that can be collected for entries made between the date of Commerce's preliminary determination and the date of the Commission's injury determination, often referred to as the "cap period."

In the case at hand, the cap period ran from January 11, 1995 to July 19, 1995. The period of review for Commerce's administrative review included the cap period and continued until June 30, 1996. In its administrative review, Commerce considered all sales made during the period of review and calculated a single assessment rate for the entire period of review. TPC challenged that determination, arguing that Commerce was required to compute two separate assessment rates, one based on sales during the cap period, and the other based on sales during the remainder of the period of review. TPC I, 1999 WL 288772, at *9-10. The Court of International Trade affirmed Commerce's methodology, noting that the cap is a limitation on collection during the cap period, unrelated to the determination of an antidumping duty in an administrative review pursuant to 19 U.S.C. 1675(a). Id. at *10. TPC appeals that decision to this court.

We have jurisdiction over this appeal pursuant to 28 U.S.C. 1295(a)(5) (1994). We review antidumping determinations made by Commerce by applying anew the standard of review applied by the Court of International Trade. Inland Steel Indus. v. United States, 188 F.3d 1349, 1359 (Fed. Cir. 1999). Accordingly, we will uphold Commerce's determination unless it is "unsupported by substantial evidence on the record, or otherwise not in accordance with law." 19 U.S.C. 1516a(b)(1)(B)(i) (1994).

The Government argues that the methodology used by Commerce should receive deference under Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984). In United States v. Mead Corp., 121 S. Ct. 2164 (2001), a decision issued after briefing and oral argument in this case, the Supreme Court made it clear that an agency's statutory interpretations are due Chevron deference when articulated in "a relatively formal administrative procedure tending to foster the fairness and deliberation that should underlie a pronouncement" having the force of law. 121 S. Ct. at 2172. We have subsequently held that statutory interpretations articulated by Commerce in antidumping determinations qualify for Chevron deference. Pesquera Mares Australes v. United States, 266 F.3d 1372, 1379-82 (Fed. Cir. 2001). Whether Commerce is entitled to Chevron deference or review under the less deferential regime set forth in Skidmore v. Swift & Co., 323 U.S. 134 (1944), is of no moment to this case, however. Under either, if the Government's position is unreasonable, deference does the agency no good.

The first issue before us is whether, on the facts of this case, Commerce is required to modify its methodology for matching costs to sales in its determination of dumping margins when the cost of the main raw ingredient, fresh pineapple, increased by a substantial amount before and during the period of review, and when the merchandise was held in inventory for a period of time before sale. Commerce has already revised its methodology once as instructed by the Court of International Trade in TPC I, dividing the period of review into two periods for purposes of calculating costs of production instead of computing a single average cost for the entire period of review. This change addressed the first problem presented by TPC-the increase in the cost of fresh pineapple during the period of review. TPC alleges that its dumping margins will continue to be distorted unless Commerce addresses the second problem-the delay between production and sale while merchandise remains in inventory.

TPC urges that the statutory language itself requires Commerce to match sales with costs calculated for a period during which the merchandise sold was actually produced, thus taking into account the inventory time. Cost of production is defined by statute as "the cost of materials and of fabrication or other processing of any kind employed in producing the foreign like product, during a period which would ordinarily permit the production of that foreign like product in the ordinary course of business," plus additional overhead and incidental expenses. 19 U.S.C. 1677b(b)(3) (emphasis added). The definition of constructed value contains similar language, the word "merchandise" replacing "foreign like product." Id. 1677b(e). In TPC's view, the only period of time that would "ordinarily permit" production is the period during which the imported goods under consideration were actually produced.

We do not read the statutory language as specifying the period to be used when determining costs, and no other statute or regulation provides further guidance. Moreover, the statute does not dictate the methodology for calculating cost of production and constructed value or for matching those costs against sales. Because we conclude that the statute does not clearly decide the matter, our next task is to determine whether Commerce's methodology is a reasonable interpretation of the statute. (No one questions Commerce's delegated authority to interpret the statute in the first instance.)

Under its standard methodology, Commerce determines cost of production by calculating a single weighted-average cost for the period of review. Final Results, 63 Fed. Reg. at 7399. Believing that the cost structure in place during the period of review is usually adequate for calculating the cost of production in the ordinary course of business, Commerce purports to depart from its standard methodology only in "certain rare situations where cost and price averages calculated over the entire period [do] not permit an appropriate comparison." Id. Nevertheless, TPC and the Court of International Trade have identified numerous cases in which Commerce adjusted its methodology in order to better match costs with prices.

One adjustment made by Commerce is the use of shorter cost reporting periods when, for example, prices have moved significantly during the period of review. See, e.g., Notice of Preliminary Determination of Sales at Less Than Fair Value and Postponement of Final Determination: Static Random Access Memory Semiconductors From Taiwan, 62 Fed. Reg. 51,422, 51,424 (Oct. 1, 1997) (using quarterly cost periods when sales prices declined significantly). TPC received the benefit of such an adjustment when Commerce calculated costs for separate fiscal years after remand from the Court of International Trade. Another adjustment is the lagging of costs when there is a delay between production and sales of goods in order to capture the actual costs of reported sales. See, e.g., Final Determination of Sales at Less Than Fair Value: Dynamic Random Access Memory Semiconductors of One Megabit and Above From the Republic of Korea, 58 Fed. Reg. 15,467, 15,473 (Mar. 23, 1993) ("DRAMs from Korea") (lagging costs for the length of time it takes for assembly and test and for average inventory holding periods when prices were declining); Fresh and Chilled Atlantic Salmon From Norway; Final Results of Antidumping Review, 58 Fed. Reg. 37,912, 37,912-13 (July 14, 1993) ("Salmon from Norway") (matching sales to actual costs incurred before the period of review because growth cycle of salmon was 18-24 months); Final Determination of Sales at Less Than Fair Value; Sweaters Wholly or in Chief Weight of Man-Made Fiber From Taiwan, 55 Fed. Reg. 34,585, 34,596 (Aug. 23, 1990) ("Sweaters from Taiwan") (using actual costs incurred for each production run).

Commerce asserts that the language of 19 U.S.C. 1677b(b)(3) and 1677b(e) is broad enough to allow its standard methodology of computing a single weighted-average cost for the period of review or any other methodology adjusted to better match costs with sales, including the use of the cost of the actual merchandise sold during the period of review. While we agree that the statutory language leaves room for some discretion by Commerce in determining the cost period, the standard methodology may not be permissible in all scenarios because Commerce has recognized that certain circumstances warrant exceptions. The question we must answer is whether Commerce's application of its standard methodology to this case, modified by the use of two cost periods but not matching actual costs to sales, falls within the range of permissible construction of the statute.

This case includes two elements not present in most antidumping determinations: a dramatic change in the cost of the product's main raw input, and a significant delay between production and sale of the product. These two factors together operate to distort the calculation of dumping margins. Splitting the period of review into two cost periods has eliminated some of the distortion caused by the increase in pineapple costs, but the problem can be remedied only by using the cost of actual merchandise sold during the period of review, even though it requires the use of costs outside the period of review. While factual distinctions between cases can almost always be found, this case is no different in principle from cases in which Commerce has modified its approach. Salmon and sweaters may have their peculiar characteristics, yet one of the reasons given for using actual costs in Salmon from Norway is similar to this case-the fluctuation of costs during the period of review. 58 Fed. Reg. at 37,913. Most compelling is DRAMs from Taiwan, in which Commerce lagged costs to account for inventory periods-essentially what TPC has requested in this case. See 58 Fed. Reg. at 15,473. The decrease in sales price of the subject merchandise in DRAMs has an effect similar to the increase in cost in this case, and we see no basis for Commerce's disparate treatment of the two cases.

The statutory language regarding cost of production and constructed value gives Commerce some latitude in determining costs and using those costs to derive dumping margins. While various methodologies are permitted by the statute, it is possible for the application of a particular methodology to be unreasonable in a given case when a more accurate methodology is available and has been used in similar cases. That is the situation presented by this case. Because we conclude that Commerce's failure to account for inventory holding time during a period of rising costs is unreasonable in this case, we reverse the decision of the Court of International Trade sustaining the methodology used by Commerce to determine TPC's cost of production and constructed value. We remand to the Court of International Trade with instructions to require Commerce to match sales of goods to costs based on the period in which those goods were manufactured, taking into account the inventory period.

During the administrative review, Commerce determined TPC's assessment rate by first determining the normal value and export price (or constructed export price) of each entry of the subject merchandise during the period of review and calculating the dumping margin for each entry in accordance with 19 U.S.C. 1675(a)(2)(A). Commerce then calculated an average assessment rate for the entire period of review by dividing the dumping margin on the subject merchandise by the entered value of the merchandise, in accordance with Commerce's established practice at the time and a more recent regulation, 19 C.F.R. 351.212 (2001). Final Results, 63 Fed. Reg. at 7393. TPC contends that the use of a single assessment rate for the entire period of review violates relevant statutory provisions and that Commerce is required to compute separate assessment rates for the cap period and the remainder of the period of review.

TPC's concern stems from the particular facts of this case. According to TPC, its highest dumping margins, as determined by the administrative review, occurred during the cap period, approximately the first third of the period of review. Final Results, 63 Fed. Reg. at 6393. Dumping margins of entries during the remainder of the period of review were lower. Thus, under Commerce's method of calculating a single assessment rate based on all dumping margins during the period of review, entries during the cap period contribute to a higher average assessment rate for the period of review. If Commerce calculated separate assessment rates based on entries during the cap period and those during the remainder of the period of review, as TPC advocates, the cap period would have an assessment rate higher than the average assessment rate calculated by Commerce, and the remainder of the period of review would have a lower assessment rate. But because the collected duty is capped in accordance with 19 U.S.C. 1673f(a) at a very low rate for entries during the cap period, TPC's overall assessment would decrease.

TPC contends that 19 U.S.C. 1675(a)(2) and 1673f(a) read together require Commerce to compute two separate assessment rates. Under TPC's interpretation of those statutes, because 1675(a)(2)(A) requires Commerce to determine the dumping margin for each entry during the cap period, it follows that Commerce is required to calculate the duty to be assessed on an entry-by-entry basis. Furthermore, 1673f(a) instructs Commerce to disregard the difference between the duty for entries during the cap period and the estimated duties deposited on those entries. According to TPC, the only way to comply with both provisions is to determine two assessment rates-one for the cap period and one for the remainder of the period of review.

We disagree with TPC's reading of the statutes. While 1675(a)(2) requires Commerce to determine dumping margins for each entry, there is no requirement that assessment rates or duties be determined for each individual entry. Commerce's method of calculating an average assessment rate for the entire period of review by dividing the dumping margins of all entries by the entered value of all entries complies with 19 U.S.C. 1675(a)(2)(C), which requires only that the calculated dumping margins be the "basis for the assessment" of antidumping duties.

Furthermore, 1673f(a) does not impact the review and determination of antidumping duties during an administrative review under 1675(a)(2). Section 1673f(a) relates to the deposit of an estimated dumping duty, required under 19 U.S.C. 1673b(d) as security after an affirmative preliminary determination by Commerce. When an exporter deposits an estimated duty for entries during an investigation, the cap provision prohibits the collection of the difference between the duty determined by the investigation and the deposited amount. See Koyo Seiko Co. v. United States, 95 F.3d 1094, 1098-99 (Fed. Cir. 1996) (discussing compliance with GATT limitation on collection of difference between posted security and final duty). Section 1673f(a) does not affect the duty for entries during the cap period; it simply limits the amount of that duty that can be collected. Thus, when Commerce determines a new duty as the result of an administrative review that is higher than the deposit of the estimated duty, the difference cannot be collected, and the duty for entries during the cap period is still capped in compliance with 1673f(a).

Accordingly, we affirm the Court of International Trade's decision that Commerce's use of one assessment rate comports with the relevant statutes. We also reject, as did the Court of International Trade, TPC's alternative argument that Commerce should calculate different assessment rates for its two affiliates. Because most of one affiliate's sales occurred during the cap period and most of the other's sales occurred during the post-cap period, TPC's argument is essentially another attempt to obtain different assessment rates for the cap period and the post-cap period.

We affirm the judgment of the Court of International Trade upholding Commerce's determination of a single assessment rate for the entire period of review. Because we conclude that Commerce's failure to modify its methodology for matching costs to sales in this case is unreasonable, and thus not within the range of permissible interpretation of the statute, we reverse the judgment of the Court of International Trade upholding Commerce's methodology and remand with instructions to remand to Commerce for further proceedings consistent with this opinion.

AFFIRMED-IN-PART, REVERSED-IN-PART, and REMANDED.

Each party shall bear its own costs.