NMB Singapore Ltd. v. United States

MICHEL, Chief Judge. JTEKT Corporation and Koyo Corporation of U.S.A. (collectively, “JTEKT”) and the Timken Company (“Timken”) have cross-appealed different aspects of the Court of International Trade’s judgment affirming a final decision of the Department of Commerce (“Commerce”). On appeal, JTEKT argues that Commerce ignored evidence and argument JTEKT presented, and that Commerce’s finding that dumping was likely to continue (and the resulting continuation of the anti-dumping duty) should be vacated. In a cross-appeal, Timken argues that Commerce lacked substantial evidence to justify recalculating (and reducing) dumping margins for the Japanese parties and also im-permissibly deviated in several respects from its established methodology when it decided to recalculate their dumping margins. Oral argument was held on December 5, 2008. We affirm the Court of International Trade’s judgment in JTEKT’s appeal, but vacate and remand in Timken’s cross-appeal. I. BACKGROUND NSK Corporation and NSK, Ltd. (collectively, “NSK”), NTN Corporation and NTN USA (collectively, “NTN”), JTEKT, and Timken all manufacture ball bearings. NTN, NSK, and JTEKT are Japanese manufacturers, while Timken is a domestic entity. 1 In 1988, Commerce began an investigation into whether Japanese-made antifriction bearings were being dumped in the United States. In 1989, Commerce determined that Japanese manufacturers were dumping bearings and issued an anti-dumping order. Antifriction Bearings (Other than Tapered Roller Bearings) & Parts Thereof from Japan, 54 Fed.Reg. 19,101 (Dep’t of Commerce May 3, 1989) (final determination). Commerce subjected the Japanese manufacturers to anti-dumping duties as follows: JTEKT, 73.55%; NSK, 42.99%; and NTN, 21.36%. Id. at 19,108. Commerce must periodically reconsider such anti-dumping orders, 19 U.S.C. § 1675(c)(1), in what are termed “sunset reviews.” Agro Dutch Indus. v. United States, 508 F.3d 1024, 1028 (Fed.Cir.2007). On appeal are the results of the second sunset review of the anti-dumping order against Japanese antifriction bearing imports. In late 2005, Commerce issued a preliminary second sunset review decision in which Commerce determined (1) that dumping was likely to continue, and (2) that the margins applicable to the Japanese companies should remain unchanged. Ball Bearings & Parts Thereof from Japan & Sing., 70 Fed.Reg. 76,754 (Dep’t of Commerce Dec. 28, 2005) (prelim.results); *1319 Ball Bearings & Parts Thereof from Japan & Sing., 70 ITADOC 76,754, available at http://ia.ita.doc.gov/frn/summary/ multiple/05-24510-1.pdf (Dep’t of Commerce Dec. 28, 2005) (prelim, results memo.) (“Preliminary Memo”). 2 Commerce based its preliminary decision to leave the anti-dumping order in place on its findings “that the total weight and value of complete Japanese ball bearings imported decreased substantially post-order and remain well below pre-order levels” and “that dumping continues at above de minimis levels.” Preliminary Memo, slip op. at 10. Commerce based its preliminary decision to not recalculate the applicable margins on its finding that “above de minimis ” dumping persisted with the order in place. Id. at 12. Commerce thereafter received arguments from interested parties, including JTEKT and Timken. In a final decision released in the spring of 2006, Commerce reaffirmed — although with an altered rationale — that dumping was likely to continue (a finding JTEKT contests on appeal). See Ball Bearings & Parts Thereof from Japan & Sing., 71 Fed.Reg. 26,321, 26,322 (Dep’t of Commerce May 4, 2006) (final results) (“Final Results”); Ball Bearings & Parts Thereof from Japan & Sing., http://ia.ita.doc.gov/frn/summary/ MULTIPLE/E6-6763-l.pdf, slip op. at 5 (Dep’t of Commerce May 4, 2006) (final results memo.) (“Final Memo”). 3 Commerce mentioned only “above de minim-is ” levels of dumping in support of its final decision to leave the anti-dumping order in place, and did not discuss import volumes. Final Memo, slip op. at 5. Reversing course, Commerce also determined that the margins applicable to the Japanese respondents should be lower than before the anti-dumping order (a finding Timken contests). See id. at 7-10, 12. Commerce based this decision on its findings that the Japanese respondents’ dumping margins had decreased during the order while their import volumes had increased or remained steady. Id. at 9-10. The Court of International Trade affirmed Commerce’s final decision in all respects. NMB Singapore Ltd. v. United States, 533 F.Supp.2d 1244 (Ct. Int’l Trade 2007). We have jurisdiction over appeals from the Court of International Trade pursuant to 28 U.S.C. § 1295(a)(5). II. DISCUSSION A. Standard of review The Federal Circuit reviews Commerce’s anti-dumping orders using the same standard of review used by the Court of International Trade. NSK Ltd. v. United States, 481 F.3d 1355, 1359 (Fed.Cir.2007); NTN Bearing Corp. of Am. v. United States, 368 F.3d 1369, 1372 (Fed.Cir.2004). This court will uphold Commerce’s decisions 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); see also Timken U.S. Corp. v. United States, 434 F.3d 1345, 1350 (Fed.Cir.2006). Commerce must explain the basis for its decisions; while its explanations do not have to be perfect, the path of Commerce’s decision must be reasonably dis-cernable to a reviewing court. Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (“State Farm ”) (“We will ... ‘uphold a decision of less than ideal clarity if the agency’s path may reasonably be discerned.’” (quoting Bowman Transp. v. Ark-Best Freight *1320 Sys., 419 U.S. 281, 286, 95 S.Ct. 438, 42 L.Ed.2d 447 (1974))). Specifically in the anti-dumping context, a final determination by Commerce must include “an explanation of the basis for its determination that addresses relevant arguments[ ] made by interested parties who are parties to the investigation or review.” 19 U.S.C. § 1677f(i)(3)(A). However, Congress did not intend for enactment of § 1677f(i) to alter the law; § 1677f(i) is instead a partial codification of the Supreme Court’s standard for judicial review of administrative decisions from State Farm. Timken U.S. Corp. v. United States, 421 F.3d 1350, 1355 (Fed.Cir.2005). B. JTEKT’s appeal JTEKT makes two arguments against Commerce’s decision to continue the anti-dumping order. First, JTEKT argues that Commerce did not consider — in violation of Commerce’s statutory mandate— evidence JTEKT submitted, specifically, (1) JTEKT’s evidence that its import levels did not decrease substantially and thus could be considered “steady” and (2) evidence that a U.S. recession caused decreases in JTEKT’s import levels around 2001. Second, JTEKT argues that even if Commerce may disregard certain portions of JTEKT’s evidence, Commerce’s decision to continue the anti-dumping order is not supported by substantial evidence. However, these two arguments overlap and we address them simultaneously. 4 Both the United States and Timken oppose JTEKT on appeal. In deciding whether lifting an anti-dumping order would be likely to lead to continued or resumed dumping, Commerce is required by statute to consider previous estimates of dumping margins and the volume of imports pre and post-order. 19 U.S.C. § 1675a(c)(l). The statute does not explain how this comparison is to indicate to Commerce whether an anti-dumping order should be lifted, but a report by the House of Representatives on the Uruguay Round Agreements Act, H.R.Rep. No. 103-316 (1994), as reprinted in 1994 U.S.C.C.A.N. 4040, commonly known as the Statement of Administrative Action (or “SAA”), is, by statute, an “authoritative expression” of the interpretation of that act, including § 1675a. 19 U.S.C. § 3512(d); see also Timken, 421 F.3d at 1354 n. 2. According to the SAA, “declining (or no) dumping margins accompanied by steady or increasing imports may indicate that foreign companies do not have to dump to maintain market share in the United States and that dumping is less likely to continue or recur if the order were revoked.” 1994 U.S.C.C.A.N. at 4213-14. A regulation provides that Commerce “will revoke” an anti-dumping order during a sunset review if Commerce “determines that revocation or termination is not likely to lead to continuation or recurrence of ... dumping.” 19 C.F.R. § 351.222(i)(l)(ii). The SAA provides: The Administration believes that existence of dumping margins after the order, or the cessation of imports after the order, is highly probative of the likelihood of continuation or recurrence of dumping. If companies continue to dump with the discipline of an order in place, it is reasonable to assume that *1321 dumping would continue if the discipline were removed. 1994 U.S.C.C.A.N. at 4214 (emphasis added). A similar standard is set forth in Commerce’s Sunset Policy Bulletin: 5 [T]he Department normally will determine that revocation of an antidumping order or termination of a suspended dumping investigation is likely to lead to continuation or recurrence of dumping where— (a) dumping continued at any level above de minimis after the issuance of the order or the suspension agreement, as applicable;.... 63 Fed.Reg. at 18,872 (emphasis added). JTEKT argues that it submitted comments and evidence regarding import levels to Commerce, but that Commerce ignored JTEKT’s import levels in violation of § 1675a(c). Specifically, JTEKT points to the weight and dollar value of all ball bearings imported from Japan into the United States, and compares the average values for the two years before the anti-dumping order (1987 and 1988) with average values for the five years of the sunset review period (2000 through 2004). By weight, average annual Japanese ball bearing imports for 2000 through 2004 were 90 percent of the levels for 1987 and 1988; by dollar value, 126 percent. 6 JTEKT does not argue that this particular method of comparison is required; JTEKT does argue that these data show Commerce could not have reasonably determined that import levels decreased after the antidump-ing order issued, and JTEKT states that Commerce should have explained why it did not agree with JTEKT’s analysis. 7 Commerce did discuss import levels in the Preliminary Memo, slip op. at 10, but omitted any such analysis from the Final Memo, see slip op. at 4-5. While in the Final Memo Commerce acknowledged that it is to consider import levels, Commerce also stated, “Normally, the Department will determine that revocation of an anti-dumping duty order is likely to lead to continuation or recurrence of dumping where dumping continued at any level above de minimis after the issuance of the order.” Final Memo, slip op. at 4. Commerce justified its finding of a likelihood of continued dumping with the following paragraph: The calculated margins in the investigation and fifteen subsequent antidumping *1322 administrative reviews for ball bearings from Japan have been above de minimis for NSK, NTN, and [JTEKT], Given that dumping continued with the discipline of the orders in place, it is reasonable to assume dumping would continue if the orders were removed. Therefore, we find that revocation of the orders would likely lead to continuation of dumping. Id. at 5. When upholding Commerce’s decision on this point, the Court of International Trade stated that Commerce had considered volume: “Here, Commerce determined that dumping continued above de minimis levels and identified a steady decrease in import volumes and therefore reasonably concluded that dumping was likely to recur.” NMB Singapore, 533 F.Supp.2d at 1255. The United States likewise indicates on appeal that Commerce did consider JTEKT’s volume information when deciding whether to continue the anti-dumping order. Both the Court of International Trade and the United States cite to the Preliminary Memo to support their assertions that Commerce considered import levels; they do not address the absence of such analysis from the Final Memo. In contrast, JTEKT argues that the Preliminary Memo is only incorporated into the Final Memo where Commerce does so expressly, e.g., “For the reasons we discuss above in response to the comments of the Japanese respondents, we affirm our decision in the Preliminary Sunset Results.” Final Memo, slip op. at 6 (discussing likelihood of continued dumping by Singaporean companies). The parties have not made clear whether Commerce’s final results should be considered a replacement for the preliminary results, or instead an addition to them; 8 or, more narrowly, whether Commerce’s discussion of import volumes only in its Preliminary Memo comports with the mandate of § 1675a(c)(l)(B). Even if we assume that — as JTEKT implies' — only Commerce’s Final Results and Final Memo are the proper subjects of our review, we may, as explained below, nonetheless affirm Commerce’s decision to leave the anti-dumping order in place. Timken points out that it is consistent with the SAA for Commerce to leave the anti-dumping order in place based solely on Commerce’s finding that above de min-imis levels of dumping persisted in spite of the anti-dumping order. JTEKT does not address the fact that Commerce’s determination that dumping is likely to continue without an anti-dumping order based solely on above de minimis dumping during the review period is consistent with not only the SAA, but also the Sunset Policy Bulletin and Commerce’s past practice. 9 Because JTEKT does not challenge the validity of the SAA or the Sunset Policy Bulletin, we assume they are valid. Thus, we reject JTEKT’s contention that the *1323 absence of discussion of import levels from Commerce’s Final Memo by itself constitutes reversible error. Because we agree Commerce could permissibly decide to leave the antidump-ing order in place based solely on its finding of the persistence of dumping, we need not address JTEKT’s argument that the conflict between Commerce’s finding in the Preliminary Memo that imports had “decreased substantially” (in connection with its decision to leave the anti-dumping order in place) and Commerce’s finding in the Final Memo that imports remained at a “steady level” (in connection with its decision to recalculate the likely margin of dumping) renders Commerce’s decision to leave the anti-dumping order in place unsupported by substantial evidence. Compare Preliminary Memo, slip op. at 10, with Final Memo, slip op. at 10. Similarly, JTEKT’s argument that Commerce did not consider “relevant arguments ... made by interested parties” (including JTEKT’s arguments that import volumes were affected by a recession) as required by § 1677f(i)(3)(A) fails because we agree the persistence of dumping alone was sufficient to justify Commerce’s decision in these circumstances. We have previously held that § 1677f(i) does not require us to invalidate a decision of Commerce if Commerce failed to explicitly address a party’s non-dispositive argument. See Timken, 421 F.3d at 1357. While a more substantial explanation from Commerce might have been helpful to us or preferable to JTEKT, its absence here is not grounds for us not to affirm because we can nonetheless reasonably discern the path of Commerce’s decision. See State Farm, 463 U.S. at 43, 103 S.Ct. 2856. C. Timken’s cross-appeal Although defending Commerce’s determination to continue the anti-dumping order, Timken attacks a different facet of Commerce’s decision, namely, Commerce’s recalculation (and lowering) of the anti-dumping margins applicable to the Japanese parties. Timken argues that (1) substantial evidence does not support Commerce’s finding that imports by the three Japanese parties had remained steady or increased, (2) Commerce’s methodology for determining import levels was an unexplained — and therefore impermissible — departure from Commerce’s established past practice, and (3) Commerce’s method of determining market share was likewise an unexplained departure from established past practice. Timken’s cross-appeal is opposed by the United States, JTEKT, and NSK. 1. Evidence of increasing imports Commerce ordinarily assumes that if an anti-dumping order is lifted, exporters will resume dumping at the same margins that existed before the anti-dumping order was in place. SAA, 1994 U.S.C.C.A.N. at 4214 (“The Administration intends that Commerce normally will select the rate from the investigation, because that is the only calculated rate that reflects the behavior of exporters and foreign governments without the discipline of an order or suspension agreement in place.”). The parties agree, however, that if an exporter’s volume of goods has been steady or increasing and its margin of dumping has been steady or decreasing over the life of an anti-dumping order, Commerce may instead calculate a new margin based on more recent data and assume this would be the margin after an anti-dumping order was lifted. See id. (“For example, if dumping margins have declined over the life of an order and imports have remained steady or increased, Commerce may conclude that exporters are likely to continue dumping at the lower rates found in a more recent review.”). The Japanese respondents submitted confidential sales data to Commerce. *1324 Commerce summarized these data as follows: The Department also analyzed each respondent’s submission that reported its relative market share of imports of subject merchandise from Japan to the United States for the five-year sunset review period. NSK’s import volumes since the issuance of the order and market share of imports, as measured in complete bearings, have remained steady or increased during the sunset review period. Likewise, NTN has maintained the level of its market share of imports and import volumes since the issuance of the order over the past five years with negligible fluctuations, [JTEKT]’s market share of imports and level of imports since the issuance of the order also have remained steady over the same five-year period. The information provided by the parties reflects a steady level of exports of subject merchandise to the United States from Japan. Final Memo, slip op. at 9-10 (footnotes omitted; emphasis added). We note that Commerce specified it relied upon data on “complete bearings” for NSK, 10 but did not specify the type of data it relied upon for JTEKT or NTN. The data for NSK’s shipments of ball bearings to the United States, using 2000 as a baseline, are as follows: [[Image here]] Aside from a downward drop from 2002 to 2003, NSK’s share of the Japanese export market increased year to year during the sunset review period. 11 *1325 NSK argues that, although its levels of finished bearings dropped significantly from 2000 to 2001, its levels from 2001 through 2004 were steady, and since these four years were the ones “closest to the time in which the order would be revoked” it would be reasonable to disregard 2000. It is not at all apparent that Commerce analyzed NSK’s finished bearing data in this manner. Because, however, there are methodological flaws in Commerce’s analysis (as discussed below in parts II.C.2 and 3), it is irrelevant whether Commerce’s findings regarding NSK’s import levels are supported by substantial evidence. The data for JTEKT’s shipments of ball bearings to the United States, using 2000 as a baseline, are as follows:_ [[Image here]] Aside from a slight uptick from 2003 to 2004, JTEKT’s share of the Japanese export market declined for the entire sunset review period. JTEKT’s share in 2004 was 59 percent of what it was in 2000. Commerce’s finding that JTEKT’s “market share of imports ... have remained steady over the ... five-year period,” Final Memo, slip op. at 10, lacks substantial evidentiary support. Commerce did not specify what data it was using to determine JTEKT’s “level of imports,” but Commerce did find JTEKT’s imports “remained steady.” Id. JTEKT’s volume of finished bearings was not completely steady over the sunset review period (and drops noticeably from 2002 to 2003), but provides the most support for a finding that JTEKT’s import levels were “steady.” Other possible measures, such as the number of bearing parts or the total number of bearings and bearing parts sold, clearly do not support such a finding. As with NSK, it is irrelevant to the outcome whether we could affirm this particular finding in isolation for the reasons discussed below in parts II.C.2 and 3. The data for NTN’s shipments of ball bearings to the United States, using 2000 12 as a baseline, are as follows: [[Image here]] While NTN’s share of the Japanese export market was relatively stable from 2000 *1326 through 2003, NTN’s market share halved from 2003 to 2004. Commerce’s finding that “NTN has maintained the level of its market share of imports ... since the issuance of the order over the past five years with negligible fluctuations,” Final Memo, slip op. at 9-10, lacks substantial evidentiary support. As with JTEKT, Commerce did not specify what data it was using to determine NTN’s import volumes. NTN’s volume of finished bearings more than doubled from 2000 to 2002, and then fell back to 50 percent over the 2000 level — also hardly “negligible fluctuations.” At oral argument, the United States indicated that an arrow drawn in the margin of the table of data NTN provided (and pointing to the line for total bearings and parts numbers) showed that Commerce had considered total numbers of bearings and parts for NTN. The United States also suggested that Commerce had considered for NTN just the “stabilization” of the levels of total bearings from 2003 to 2004. The least fluctuating figures NTN provided were for dollar values of total bearings sold. Commerce, however, said it was considering NTN’s “import volumes,” implying number of units sold, rather than total dollar value sold. Furthermore, it would seem incongruous for Commerce to consider numbers of finished bearings sold for JTEKT and NSK, but without explanation, dollar value of total bearings sold for NTN. We thus must hold that Commerce’s finding on this point is not supported by substantial evidence. Looking at Commerce’s findings as a whole, it is difficult to square many of Commerce’s statements that the Japanese importers’ levels of imports were steady or increasing with the actual data before Commerce, even if we accepted the arguments that Commerce could permissibly consider different types of import data and different segments of the five-year review period for different importers while ignoring pre-order levels. This is a situation in which we cannot reasonably discern Commerce’s path because there does not appear to be a reasonable way to consider the data as a whole and arrive at Commerce’s conclusion that dumping margins should be recalculated. See State Farm, 463 U.S. at 43, 103 S.Ct. 2856. 2. Consideration of pre-order levels Timken also argues that Commerce should have considered not just 2000 through 2004 data in isolation, but should have compared data for these years to data from before the anti-dumping order to determine whether import volumes from Japan were steady or increasing. Or, to state Timken’s argument more directly, just because imports were steady or increasing from 2000 to 2004 is not sufficient for Commerce to declare them steady or increasing for purposes of the second sunset review — imports must also have been steady or increasing during the review period compared to 1987 or 1988. According to Timken, Commerce has established a clear practice of comparing imports during a sunset review period to pre-order levels and has not explained why it did not in the instant matter. 13 As a threshold matter, NSK argues that Commerce indeed considered pre-order levels because pre-order data appear in the attachments to Commerce’s memo and Commerce even rearranged some of the data submitted by the parties so that 2000 through 2004 data appeared on the same page as pre-order data. However, the actual discussion of import levels (the paragraph quoted above on page 13) is limited *1327 solely to 2000 to 2004. 14 In any event, as discussed below, merely looking at such data but not relying on them when determining how import levels are changing would be contrary to Commerce’s established practice. 15 The parties did not each provide exactly the same type of pre-order data. JTEKT and NSK provided data from 1987, while NTN provided data from 1988. 16 JTEKT’s pre-order data included units and dollar figures for finished bearings, bearing parts, and total bearings. NSK’s pre-order data were only for finished bearings, and NTN’s was only for total bearings. Using each company’s pre-order data as a baseline, the Japanese respondents’ import figures generally show declines in numbers of bearings imported but have mixed results for changes in dollar value of bearings imported: [[Image here]] [[Image here]] [[Image here]] Notably, JTEKT’s imports of total bearings fell by over two-thirds from 1987 to 2000, and fell by over half from 2000 to 2004. The trends for NSK and NTN are less pronounced; both had increased numbers of units sold early in the sunset review period compared to before the anti-dumping order, but both finished the sun *1328 set review period importing fewer units than before the anti-dumping order. Pursuant to 19 U.S.C. § 1675a(c)(l)(B), when determining whether dumping is likely to recur if an anti-dumping order is terminated, Commerce is required to consider, inter alia, “the volume of imports of the subject merchandise for the period before and the period after the issuance of the antidumping duty order or acceptance of the suspension agreement.” Under 19 U.S.C. § 1675a(c)(3), Commerce is to determine what the margin of dumping is likely to be if an anti-dumping order is lifted. The margin is “normally” the one determined in the original, pre-order investigation, but Commerce is not directed by statute to consider any specific factors in determining the likely margin. Id. The SAA explains that, “[f|or example, if dumping margins have declined over the life of an order and imports have remained steady or increased, Commerce may conclude that exporters are likely to continue dumping at the lower rates found in a more recent review.” 1994 U.S.C.C.A.N. at 4214. The plain language of this report — “over the life of an order” — refers only to the period of time during which an antidumping order is in effect, not the period of time before the order. Because, however, Commerce is not acting in the first instance, we need not decide whether Commerce would be required by statute or the SAA to consider pre-order volumes when determining whether exports have remained steady or increased. Commerce, like any agency, is due deference from the courts in certain matters entrusted to it by Congress. See Br. Steel PLC v. United States, 127 F.3d 1471, 1475 (Fed.Cir.1997). Once Commerce establishes a course of action, however, Commerce is obliged to follow it until Commerce provides a sufficient, reasoned analysis explaining why a change is necessary. See id.; see also Save Domestic Oil, Inc. v. United States, 357 F.3d 1278, 1283-84 (Fed.Cir.2004) (“[I]f Commerce has a routine practice for addressing like situations, it must either apply that practice or provide a reasonable explanation as to why it departs therefrom.”). Timken points to a number of Commerce decisions that Timken claims support its course-of-action theory. Although many of the cases Timken cites do not address the issue clearly, those that do indicate that Timken is correct that Commerce ordinarily compares pre-order volumes to volumes during the life of an anti-dumping order, rather than just considers volumes during the life of an anti-dumping order in isolation. For example, in Antifriction Bearings from Romania, Commerce issued an anti-dumping order in 1989. 64 Fed.Reg. 60,-313, 60,314 (Nov. 4, 1999) (final results). Commerce noted that imports of ball bearings from Romania “declined significantly since 1988, dropping from 13.5 million units to 0.7 million units. Although imports increased to 2.5 million units in 1997, they remain significantly below pre-order volumes.” Id. at 60,316-17. Thus, Commerce was faced with a situation where levels of imports were increasing during the life of the anti-dumping order, yet were still below pre-order levels. During the order, dumping margins were consistently “below de minimis levels.” Id. at 60,317. Commerce nonetheless concluded that this did not present a situation warranting recalculation of the dumping margins because import levels were not steady or increasing. Id. Similarly, in Corrosiorir-Resistant Carbon Steel Flat Products from Japan, Commerce rejected a party’s contention that an increase just during the life of an order was sufficient to recalculate dumping margins; Commerce rejected this argument because the import levels were less during *1329 the order than they were pre-order. 17 In other decisions Timken cites, that import levels were lower during the anti-dumping order than before the order was a factor in Commerce’s decision, although this was not the sole reason Commerce gave for leaving the anti-dumping orders in place. 18 Other decisions Timken cites were made after the decision Timken is appealing, so these cases do not prove that Commerce went against an established policy in the matter on appeal; however, in each of these cases, Commerce decided against recalculating dumping margins because import levels during the order were less than import levels pre-order. 19 No party cites any case in which Commerce has rejected the idea that import levels during an anti-dumping order should be compared to pre-order import levels. Timken is therefore correct that Commerce’s focus in this case on only import volumes during the life of the anti-dumping order — instead of a comparison of such volumes to pre-order volumes — -is a deviation from an otherwise consistent practice of Commerce. The decision of the Court of International Trade affirming Com-raerce must be vacated so that Commerce can reconsider its analysis of this issue accordingly. 3. Share of Japanese exports Timken also argues that Commerce, for the purpose of determining whether to recalculate dumping margins, has a clearly established policy of using measurements other than a company’s share of exports to the United States from its home country as the measure of import levels. 20 In connection with Commerce’s determination whether to rely on an old margin or calculate a new one, 19 U.S.C. § 1675a(c)(3) does not direct Commerce to consider any particular factors. However, the SAA notes that “declining (or no) dumping margins accompanied by steady or increasing imports may indicate that foreign companies do not have to dump to maintain market share in the United States.” 1994 U.S.C.C.A.N. at 4213-14 (emphasis added). The parties do not point to any controlling document that further explains the meaning of “market share” in this context. Timken, however, *1330 argues that “market share in the United States” makes it inappropriate for Commerce to use a measure from some foreign market, and furthermore that Commerce’s prior decisions preclude it from using respondents’ shares of Japanese ball bearing exports as a proxy for their shares of the U.S. ball bearing market. The relative data for each Japanese respondent’s share of the U.S. import market for the subject goods are — using 2000 as a baseline for each company — as follows: [[Image here]] (As noted above, even assuming that such data are properly considered instead of the U.S. market share, the declines experienced by JTEKT and NTN during the sunset review period cannot reasonably be considered “steady or increasing.”) Commerce does recognize there is a difference between an importer’s share of exports from its country and U.S. market share: “[Ajlthough there would be a correlation between a company’s share of total subject exports and its share of the U.S. market, we do not agree that a company’s share of exports is the same as its relative market share.” 21 Commerce has explained that in analyzing whether imports [sic] volumes remained steady or increased, the Department normally will consider the company’s relative market share. The emphasis on the market share of an importer derives from the SAA, which states that declining (or no) dumping margins accompanied by steady or increasing imports may indicate that foreign companies do not have to dump to maintain market share in the United States. 22 The other parties cite to only one case in which Commerce has clearly relied on relative export data as a proxy for U.S. market share, and this case post-dates the decision under appeal. 23 *1331 Thus, if it is the case that the U.S. market share of the Japanese companies could be determined from the data before Commerce, then Commerce’s determination went against the practice it had established as of the time of the decision under review. 24 Commerce, however, has noted that its analysis is driven by the information parties submit: “[I]n determining whether a more recently calculated margin is probative of an exporters’s [sic] behavior absent the discipline of an order, we will normally consider the company’s relative market share, with such information to be provided by the parties.” 25 Commerce is, in some circumstances, able to determine U.S. market share from other data. 26 Commerce did not state in this instance whether it was able to determine the U.S. market share of each of the Japanese respondents. However, the record contains a table entitled “BBs — Apparent Consumption and Imports from Japan as a Share Thereof,” which lists dollar values of total imports of ball bearings from Japan for 1987 through 1990 and 2000 through 2004. While it appears to us that it may be possible to estimate each Japanese company’s share of the U.S. market by combining these data with each company’s percentage share of Japanese exports to the United States, Commerce did not discuss this, and we may not make such a factual determination in the first instance. See Mittal Steel Point Lisas Ltd. v. United States, 542 F.3d 867, 875 (Fed.Cir.2008) (“Appellate courts do not make factual findings; they review them.”). Because we cannot discern whether Commerce’s finding on this issue is contrary to its past practice or supported by substantial evidence, we must vacate and remand. At-chison, Topeka & Santa Fe Ry. Co. v. Wichita Bd. of Trade, 412 U.S. 800, 808, 93 S.Ct. 2367, 37 L.Ed.2d 350 (1973) (“Whatever the ground for the departure from prior norms, however, it must be clearly set forth so that the reviewing court may understand the basis of the agency’s action and so may judge the consistency of that action with the agency’s mandate.”). CONCLUSION For the reasons provided above, we affirm the Court of International Trade’s decision affirming Commerce’s determination to leave the anti-dumping order in place, but vacate and remand the Court of International Trade’s decision affirming Commerce’s recalculation of the anti-dumping duties for to JTEKT, NTN, and NSK. AFFIRMED-IN-PART and VACATED-AND-REMANDED-IN-PART Footnotes [1] NMB Singapore Ltd. is a Singaporean ball-bearing manufacturer that was involved in these proceedings before Commerce and the Court of International Trade, but is not a party to this appeal. Although Timken’s cross-appeal challenging Commerce's recalculation of the anti-dumping duty applies to NTN, NTN did not participate in the appeal, either. [2] Commerce released a version of the Preliminary Memo to the parties on December 19, 2005. [3] Commerce released a version of the Final Memo to the parties on April 27, 2006. [4] At oral argument, JTEKT asserted that Commerce’s reasoning was "opaque” and clarified that JTEKT was not asking this court to hold outright that Commerce should have lifted the anti-dumping order, but that JTEKT was asking this court to vacate Commerce's decision to leave the anti-dumping order in place and remand the matter to Commerce with instructions to provide a sufficient explanation of whatever new decision Commerce reached. [5] Policies Regarding the Conduct of Five-year ("Sunset") Reviews of Antidumping & Countervailing Duty Orders, 63 Fed.Reg. 18,871 (Imp. Admin., Int'l Trade Admin., Dep’t of Commerce Apr. 16, 1998). [6] Using 1988 as a baseline, the relative values for each year, considered individually, are as follows: [[Image here]] Although the parties have not requested confidential treatment of this particular set of data, the issues on appeal focus on relative trends rather than absolute numbers. For ease of comparison and to protect the confidentiality of data elsewhere in the opinion, we will use relative figures throughout. At oral argument, when asked by the court if using percentages in an opinion would be acceptable, counsel for JTEKT stated “I think that would be okay but I would want to take a look at the record and — and discuss internally.” The court has not heard further from the parties regarding this issue in the more than two months since oral argument. [7] JTEKT appears to recognize that for it to ultimately prevail, Commerce must find, inter alia, that Japanese imports of ball bearings have been “steady or increasing” since the anti-dumping order took effect. See SAA, 1994 U.S.C.C.A.N. at 4213-14. [8] Commerce's regulations generally require Commerce to issue preliminary and final results when reviewing anti-dumping orders, 19 C.F.R. § 351.221(b)(4)-(5), although in certain circumstances, Commerce may issue only final results when conducting a sunset review, 19 C.F.R. § 351.221 (c)(5)(ii). The content of these results, however, does not appear to be specified by regulation. [9] See, e.g., Ball Bearings from Sing., 64 Fed. Reg. 60,287, 60,295 (Dep't of Commerce Nov. 4, 1999) (final results) (“Regardless of the level of imports, dumping margins above de minimis levels continue as do imports of the subject merchandise; dumping continues to exist.’’); Certain Iron Constr. Castings from Braz., Can. & P.R.C., 64 Fed.Reg. 30,310, 30,-311 (Dep’t of Commerce Jun. 7, 1999) (final results) (Commerce “indicated that normally it will determine that revocation of an anti-dumping order is likely to lead to continuation or recurrence of dumping when ... dumping continued at any level above de min-imis after the issuance of the order.”). [10] NSK also provided Commerce with data for the sunset review period on import volumes of bearing parts, as well as dollar values for finished bearing and bearing part imports, but Commerce did not discuss this data in the Final Memo. Since the anti-dumping order applies to ball bearings and ball bearing parts, Antifriction Bearings (Other than Tapered Roller Bearings) & Parts Thereof from Japan, 54 Fed.Reg. at 19,101, it seems unusual that Commerce would, without explanation, not consider total levels of finished bearings and bearing parts when determining whether imports of the subject merchandise have increased or remained steady. At oral argument, the United States indicated that data on imports of finished bearings and bearing parts was not available for all three Japanese respondents. NTN and JTEKT did provide such data. While NSK provided separate tables for finished bearings and bearing parts for the sunset review period, it is a matter of simple addition to calculate total levels from these two tables. What NSK did not provide was complete pre-order data; NSK stated that it only had available data from 1987 on finished ball bearings. [11] As noted in footnote 7, above, throughout this opinion we use relative figures. In the case of the above table and the seven other tables in the opinion, we have arrived at relative figures by extrapolating from the absolute numbers in the record. Those absolute numbers, the parties have informed us, are confidential. We present the relative figures in the opinion in an effort to demonstrate what the confidential absolute numbers clearly reflect. Our selection of baseline years used to calculate the percentages in the charts was done for ease of comparison and may not necessarily reflect the methodology Commerce used in drawing its conclusions. Unless explicitly addressed in this opinion, our selection of baseline years should be considered immaterial. For example, we do not mean to suggest *1325 merely by using a certain year as a baseline that Commerce should consider that year the starting point for its analysis. [12] NTN’s data is reported for a fiscal year beginning with the fourth quarter of the listed year and running through the third quarter of the following year. [13] Although Timken argued in its opening brief that Commerce lacks the discretion not to consider pre-order export volumes when determining whether export volumes have increased or remained steady, Timken appears to have abandoned this argument in its reply. [14] Additionally, for each of the Japanese respondents, Commerce stated it considered import volumes “since the issuance of the order.” The language "since the issuance of the order" refers to the period of time during which an anti-dumping order is in effect, not the period before the order. [15] The United States argues — incorrectly— that the Court of International Trade addressed Timken's argument that Commerce had to consider pre-order volumes and rejected Timken’s position, citing NMB Singapore, 533 F.Supp.2d at 1258-59. The cited portion of the Court of International Trade’s opinion, however, does not contain any analysis of this issue. [16] NTN and NSK indicated that their data were for fiscal years, while JTEKT indicated its data were for a calendar year. [17] http://ia.ita.doc.gov/frn/summary/japan/00-195 62-1 .txt (Dep’t of Commerce Aug. 2, 2000) (final results memo.). [18] Certain Stainless Steel Sheet & Strip in Coils from Mex., http://ia.ita.doc.gov/frn/ summary/mexico/E5-514-1.pdf, at 4 (Dep't of Commerce Feb. 8, 2005) (final results memo.); Fresh & Chilled Atlantic Salmon from Nor., http://ia.ita.doc.gov/frn/summary/ norway/E5-8136-l.pdf, at 8 (Dep't of Commerce Dec. 30, 2005) (final results memo.). [19] Canned Pineapple Fruit from Thail., http:// ia.ita.doc.gov/frn/summary/THAILAND/E7-3891-l.pdf, at 4 (Dep't of Commerce Mar. 6, 2007) (final results memo.); Honey from Arg. & P.R.C., http://ia.ita.doc.gov/frn/summary/ MULTIPLE/E7-4052-1 .pdf, at 11 (Dep’t of Commerce Mar. 7, 2007) (final results memo.); Carbon & Certain Alloy Steel Wire Rod from Braz., Can., Indon., Mex., Mold., Trin. & Tobago, & Ukr., http://ia.ita.doc.gov/ frn/summaxy/MULTIPLE/E8-l 14-1.pdf, at 14 (Dep't of Commerce Dec. 31, 2007) (final results memo.). [20] Timken also argues that using market share of Japanese exports to the United States as a proxy for U.S. market share “makes no sense” and points out that it would be possible for total Japanese exports to the United States to decline significantly but for an individual manufacturer’s share of Japanese exports to remain steady or even increase significantly. The other parties do not dispute that changes in a manufacturer's share of Japanese exports to the United States do not necessarily correspond to changes in its U.S. market share. It appears that export market share may be a poor proxy. However, because the path of Commerce’s decision is unclear, we need not decide if or under what circumstances share of exports to the United States may be considered a reasonable proxy for share of the U.S. market. [21] Tapered Roller Bearings from P.R.C., 65 ITADOC 11,550 (Dep't of Commerce Mar. 3, 2000) (final results memo.). [22] Corrosion-Resistant Carbon Steel Flat Products from Can., 65 ITADOC 47,379 (Dep’t of Commerce Aug. 2, 2000) (final results memo.) (emphasis added). [23] Oil Country Tubular Goods from Mex., http://ia.ita.doc.gOv/frn/summary/MEXICO/E 6-22076-1.pdf (Dep’t of Commerce Dec. 26, 2006) (prelim, results memo.). There, Commerce based its analysis on a manufacturer’s relative share of its country’s exports to the United Slates rather than the manufacturer’s share of the U.S. market: In determining whether import volumes remained steady or increased, pursuant to die SAA, the Department will also normally consider a company’s relative market share. During the investigation, Hylsa accounted for less than 40 percent of all shipments of OCTG from Mexico to the United States. In 2005, Hylsa accounted for more than 50 percent of all shipments of OCTG from Mexico to the United States. The dumping margins for sales by Hylsa during the Second Sunset Review period are substantially below those of the "all others” rate in the investigation, the rate applied to exports by Hylsa after the publication of the order. Thus, the lower dumping margins for Hyl-sa’s sales in the Second Sunset Review period occurred while Hylsa increased shipments of OCTG from Mexico to the United States and captured larger market share of the Mexican OCTG market in the United States. Therefore, we find that a more recent rate obtained in an administrative review is a more appropriate rate for Hylsa in this sunset review. Id. at 15-16 (citations and footnote omitted). [24] Additionally, if the Japanese respondents' shares of the U.S. market could be estimated from the available data and were shown to have decreased during the sunset review period, we would have concerns about whether Commerce's findings that their market shares were steady or increasing were supported by substantial evidence. [25] Certain Circular Welded Non-Alloy Steel Pipe From Braz., S. Korea, Mex., Taiwan, & Venez., 64 Fed.Reg. 67,854, 67,858 (Dep't of Commerce Dec. 3, 1999) (final results) (emphasis added). [26] See Aramid Fiber Formed of Poly Para-Phenylene Terephthdlamide from Neth., 65 ITADOC 65,294 (Dep't of Commerce Nov. 1, 2000) (final results memo.).