Dorbest Ltd. v. United States

*1265 OPINION POGUE, Judge. This matter arises from an affirmative antidumping duty determination by the Depaxtaient of Commerce (“Commerce”) in its investigation of wooden bedroom furniture from the People’s Republic of China (“PRC”). Plaintiffs challenge numerous aspects of that determination here. Before the court are USCIT R. 56.2 Motions for Judgment on the Agency Record filed by the parties, specifically by Dorbest Limited et al (“Dorbest” also known as “Respondents”), the American Furniture Manufacturers Committee for Legal Trade et al. (“AFMC” also the “Petitioners” in the investigation), and Commerce. For the reasons set forth below, the court grants in part each of these motions and denies in part each of these motions; the coui’t also reserves decision on several issues pending the results on remand. BACKGROUND A. On December 17, 2003, Commerce commenced an antidumping investigation of wooden bedroom furniture from the PRC in response to a petition filed by the domestic industry. See Wooden Bedroom, Furniture from, the People’s Republic of China, 68 Fed.Reg. 70,228 (Dep’t Commerce Dec. 17, 2003) (initiation of anti-dumping duty investigation). 1 The investigation covered more than 211 Chinese exporters/producers of wooden bedroom furniture making this one of the largest investigations involving PRC companies. See Wooden Bedroom, Furniture from, the People’s Republic of China, 69 Fed.Reg. 35,312, 35,313 (Dep’t Commerce June 24, 2004) (notice of preliminary determination and postponement of final determination) (“Preliminary Determination ”). The period of investigation (“Period of Investigation” or “POI”) encompassed imports of the subject merchandise from April 1, 2003 through September 30, 2003. Commerce rendered an affirmative less than fair value determination for the subject merchandise and imposed the antidumping duty order and dumping margins that are at issue here. Wooden Bedroom Furniture from the People’s Republic of China, 69 Fed.Reg. 67,313, 67,317 (Dep’t Commerce Nov. 17, 2004)(notice of final determination of sales at less than fair value) (“Final Determination ”) amended by Wooden Bedroom Furniture from the *1266 People’s Republic of China, 70 Fed.Reg. 329, 330 (Dep’t Commerce Jan. 4, 2005) (notice of amended final determination of sales at less than fair market value and antidumping duty order) (“Amended Final Determination ”). More specifically, Commerce determined that the PRC is an NME country and that available information did not permit the foreign market value of the merchandise to be determined as it would in a market economy. See Preliminary Determination, 69 Fed.Reg. at 35,318. Consequently, Commerce derived the respondent’s normal value through aggregating the surrogate costs of the factors of production required to produce the product. See id. at 35,324. Because of the large number of companies under investigation, pursuant to 19 U.S.C. § 1677Í-1 (c)(2)(B), Commerce limited its investigation to the seven largest manufacturers of wooden bedroom furniture from the PRC. 2 Among these seven was Respondent Dorbest. See Preliminary Determination, 69 Fed.Reg. at 35,-318. In the investigation, Commerce chose India as the surrogate country and chiefly relied on a data set referred to as the Monthly Statistics of Foreign Trade in India (“MSFTI”) to value the factors of production (numbering over 500). Id.; Id. at 35,324; Memorandum from James H. Joe-hum to Jeffrey A. May, Issues and Decision Memorandum for the Less-Thark-Fair-Value Investigation of Wooden Bedroom Furniture from the People’s Republic of China, at 41 (Cmt.2), Dep’t of Commerce (November 8, 2004), P.R. Doc.1933, available at http://ia.i ta.doc.gov/frn/sum-mary/prc/04-25507-l.pdf (“Issues & Decision Mem.”). Likewise, Commerce used arne. financial statements from Indian companies to calculate profit, overhead, and general expenses. Id. at 23. For its calculation of the wage rate, Commerce ran a regression to determine the relationship between nations’ per capita Gross National Product and their wage rates; Commerce then multiplied the resulting coefficient by the PRC’s per capita gross national product to derive China’s wage rate. See Wooden Bedroom Furniture from the People’s Republic of China: Final Results of Redetermination Pursuant to Court Remand Orders (Dep’t Commerce Aug. 1, 2005) (“Remand Determination”). B. The court must sustain a final determination in an antidumping duty investigation if that determination is supported by substantial evidence on the record and is otherwise in accordance with law. 19 U.S.C. § 1516a(b)(1)(B)(i) (2000); Ta Chen Stainless Steel Pipe, Inc. v. United States, 298 F.3d 1330, 1335 (Fed.Cir.2002). The parties here have collectively alleged more than a score of issues requiring review, some with multiple subparts. To issue a coherent opinion, the court has grouped the issues as follows: (1) Commerce’s selection of data sets, specifically, (a) Commerce’s use of Indian surrogate data to value the factors of production, (b)Commerce’s use of the MSFTI, (c) Commerce’s calculation of the wage rate, and (d) Commerce’s selection of surrogate companies for the calculation of profit, overhead, and selling, general and administrative expenses (collectively “financial ratios”); (2) Commerce’s valuation of certain specific factors of production; (3) other individual company-specific protests; and (4) the application (or lack thereof) of *1267 adverse inferences in Commerce’s selection of facts otherwise available. For ease of reference, the discussion is organized as follows: I. DATA SETS.............................................................1267 A. Selection of Surrogate Countries .......................................1270 (1) Evaluation of Indonesian Data......................................1272 (2) Commerce’s finding that India was a producer of the comparable merchandise....................................................1273 (3) Commerce’s finding that India was a significant producer of subject merchandise....................................................1274 (4) Weighing the choice between Indonesia and India.....................1274 B. Monthly Statistics of Foreign Trade in India.............................1276 (1) In general .......................................................1276 (2) MSFTI as a primary data set.......................................1276 (3) MSFTI as applied to individual factors...............................1278 (a) Mirrors......................................................1279 (i) Imported Mirrors As Inputs...............................1279 (ii) Glass Yug...............................................1280 (iii) Tarun Vadehra, Highland House and Goldfindo ..............1282 (iv) The MSFTI data is either non-inclusive or distortive of mirror inputs..........................................1283 (v) Infodrive India...........................................1284 (vi) Commerce’s evaluation....................................1286 (b) Paints........................................................1288 (e) Cardboard....................................................1290 C. Wage rate...........................................................1291 (1) Facial Challenge..................................................1292 (2) As Applied Invalidity..............................................1292 (a) Creation of the Regression Model ...............................1294 (i) Notice and Comment Rulemaking..........................1295 (ii) Deadlines ...............................................1296 (b) Distortion of Regression Model..................................1296 (3) Proper Data Set..................................................1299 D. Financial Ratios......................................................1300 (1) IFP.............................................................1302 (2) Jayaraja.........................................................1303 (3) Evergreen.......................................................1304 (4) Swaran, Nizamuddin, Fusion Design, and DnD........................1305 (5) Indonesian Firms.................................................1307 II. VALUING SPECIFIC FACTORS OF PRODUCTION........................1308 (A) Hooks and Connectors................................................1309 (B) Hinges..............................................................1309 (C) Resin...............................................................1310 (D) Styrofoam...........................................................1311 (E) Cardboard...........................................................1312 (F) Iron Components.....................................................1313 III. DISCRETE COMPANY-SPECIFIC ISSUES ...............................1314 (A) Voluntary Remand Issues.............................................1314 (B) Zeroing.............................................................1315 IV. FACTS OTHERWISE AVAILABLE/ADVERSE INFERENCES..............1317 (A) Factor Inputs........................................................1319 (B) Wood Scraps.........................................................1321 V. CONCLUSION...........................................................1321 *1268 I. DATASETS As noted above, because pricing information in NMEs is largely unreliable, section 223 of the Tariff Act of 1930, 19 U.S.C. § 1677b(c)(1) 3 authorizes Commerce to approximate the cost of production with pricing information from “surrogate” countries and companies. The court notes that the antidumping duty statute both authorizes and requires that “the valuation of the factors of production shall be based on the best available information regarding the values of such factors in a market economy country or countries considered to be appropriate by the administering authority.” 19 U.S.C. § 1677b (c)(1) (emphasis added); see also Globe Metallurgical, Inc. v. United States, 28 CIT -, -, 350 F.Supp.2d 1148, 1156-57 (2004). The term “best available” is one of comparison, ie., the statute requires Commerce to select, from the information before it, the best data for calculating an accurate dumping margin. The term “best” means “excelling all others.” II Oxford English Dictionary 139(2d 1989); Webster’s II New Riverside University Dictionary 168 (1988) (“[exceeding all others in excellence, achievement, or quality”). This “best” choice is ascertained by examining and comparing the advantages and disadvantages of using certain data as opposed to other data. See Guangdong Chemicals Imp. & Exp. Corp. v. United States, 30 CIT ——, 460 F.Supp.2d 1365, 1370, 2006 WL 2679989, *5, Slip Op. 06-142, at 8 (Sept. 18, 2006). In calculating factors of production, Commerce typically employs data sets. Court review of whether Commerce’s data set selection is the “best available information” addresses whether the particular selection is supported by substantial evidence and whether it is in accordance with law. Whether a data set selection issue is factual or legal, ie., reviewed for substantial evidence or for its accordance with law, depends on the question presented. If the question is whether Commerce may use a particular piece of data, whether Commerce may use a factor in weighing the choice between two data sources, or what weight Commerce may attach to such a factor, the question is legal. Cf. Writing Instrument Mfrs. Ass’n v. U.S. Dep’t of Commerce, 21 CIT 1185, 1187-88, 984 F.Supp. 629, 634 (1997); Lasko Metal Prods., Inc. v. United States, 43 F.3d 1442, 1446 (Fed.Cir.1994); Nation Ford Chem. Co. v. United States, 166 F.3d 1373, 1377 (Fed.Cir.1999). If the question is whether Commerce should have used a particular piece of data, when viewed among alternative available data, or what weight Commerce should attach to a price or data, the question is factual. Cf. Shandong Haurong Machinery Co., v. United States, 29 CIT-,-, Slip Op 05-54 at 10 (2005); Yantai v. United States, 26 CIT 605, 607, 610-612 (2002). In reviewing legal issues presented here, the court applies the two-step inquiry of Chevron U.S.A. Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 842-843, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). Accordingly, Commerce has considerable discretion in selecting a data source. See Nation Ford Chem. Co. v. United States, 166 F.3d at 1377. Commerce’s exercise of its discretion, however, must still have fidelity to its statutory mandate. Statutorily, Commerce has a duty to ensure that dumping margins are calculated as accurately as possible. See, e.g., Allied-Signal Aerospace Co. v. United States, 996 F.2d 1185, 1190 (Fed.Cir.1993); Rhone Poulenc, Inc. v. United States, 899 F.2d 1185, 1191 (Fed.Cir.1990). This rule applies with equal force to imports from an NME. S.Rep. No. 100-71 at 106 (1987) (Committee on Finance)(“[t]he Committee is particularly concerned that imports from certain nonmarket economy countries, such as the Peoples Republic of China, not be unfairly disadvantaged by use of the new methodology where price differences can be accounted for in whole or in part by quality differences in the imported merchandise.”). *1269 On factual issues, the court’s role “is not to evaluate whether the information Commerce used was the best available, but rather whether a reasonable mind could conclude that Commerce chose the best available information.” Goldlink Indus. Co. v. United States, 30 CIT-, -, 431 F.Supp.2d, 1323, 1327 (2006); see also CITIC Trading Co. v. United States, 27 CIT -, -, 2003 WL 1587093, *6 (2003) (“while the standard of review precludes the court from determining whether Department’s [sic] choice of surrogate values was the best available on an absolute scale, the court may determine the reasonableness of Commerce’s selection of surrogate prices.”); cf. Klein v. Peterson, 866 F.2d 412, 414 (Fed.Cir.1989) (reconciling an agency’s duty to find clear and convincing evidence of misconduct with the court’s substantial evidence standard of review as “a reasonable mind could have found the evidence of misconduct clear and convincing”). For the court to conclude that a reasonable mind would support Commerce’s selection of the best available information, Commerce needs to justify its selection of data with a reasoned explanation. Cf. Lasko Metal Prods. v. United States, 43 F.3d 1442, 1446 (Fed.Cir.1994); see Olympia Indus., Inc. v. United States, 22 CIT 387, 390, 7 F.Supp.2d 997, 1001 (1998) (“Commerce has an obligation to review all data and then determine what constitutes the best information available or, alternatively, to explain why a particular data set is not methodologically reliable”). In doing so, Commerce must “conduct a fair comparison of the data sets on the record” with regard to its announced method or criteria. Allied Pac. Food (Dalian) Co. v. United States, 30 CIT-,-, 435 F.Supp.2d 1295, 1313-14 (2006). The court must also recognize that Commerce has limited resources and is under time constraints and, therefore, a certain level of imprecision is not unreasonable. See Geum Poong Corp. v. United States, 26 CIT 991, 995 (2002). In addition, choosing the best data requires expertise and, on occasion, judgment. See Lasko Metal Prods., 43 F.3d at 1446; cf. Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Aut. Ins. Co., 463 U.S. 29, 52, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (“It is not infrequent that the available data do not settle a regulatory issue, and the agency must then exercise its judgment in moving from the facts and probabilities on the record to a policy conclusion.”). Accordingly, parties and courts may aspire to, but cannot demand, perfection. Moreover, “the court may not substitute its judgment for that of the [agency] when the choice ‘is between two fairly conflicting views, even though the court would justifiably have made a different choice had the matter been before it de novo.’ ” Goldlink Ind. Co. v. United States, 30 CIT at-, 431 F.Supp.2d at 1326 (quoting Am. Spring. Wire Corp. v. United States, 590 F.Supp. 1273, 8 CIT 20, 22 (1984)). Another related principle that may seem obvious, but is relevant here, is that Commerce’s findings or conclusions rendered in equations or numeric form are not beyond scrutiny. Id.; accord Geum Poong, 26 CIT at 995. Scrutiny of surrogate values is important because they are proxies — they are not actual costs but estimates based on the best available information. If the proxy values selected prove unrepresentative, reliance on them defeats their purpose, namely, to derive a dumping margin that is as accurate as possible. See, e.g., Goldlink Indus., 30 CIT at -, 431 F.Supp.2d at 1334. Hence, if Commerce selects a particular data set that is demonstrably unrepresentative or distortional, a reasonable mind may rightly, question how such a selection could be the “best.” It may in fact be the best available information, but affirming *1270 Commerce’s choice requires a reasoned explanation from Commerce that is supported by the administrative record. In considering Commerce’s announced criteria, an additional consideration is that merely enlarging the size of an unrepresentative data set does not necessarily mean that a reasonable mind can conclude that the data set is the best available on the administrative record. Data sets need not be large (proportional to all possible data points) if the means of selecting the data points is statistically sound, e.g., statistical sampling. Similarly, a large but biased data set is of limited (if any) probative value. Cf. Bazemore v. Friday, 478 U.S. 385, 400, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986). For example, a researcher attempting to assess the literacy rate in the United States who surveys every college graduate in the country would create a large data set that has absolutely no probative value; however, a researcher who randomly samples a few thousand people, chosen from the population at large, would create a valuable data set. With these considerations in mind, the court’s review of Commerce’s compilation of data sets will focus on the soundness of Commerce’s announced method or criteria in selecting data points, and its adherence thereto. Cf. Shanghai Foreign Trade Enters. Co. v. United States, 28 CIT-, 318 F.Supp.2d 1339, 1351-52 (2004). A. Selection of Surrogate Countries As noted above, “the valuation of the factors of production shall be based on the best available information regarding the values of such factors in a market economy country or countries considered to be appropriate.” 19 U.S.C. § 1677b(c)(1) (emphases added). The statute further specifies that an “appropriate” market economy country is one “(A) at a level of economic development comparable to that of the nonmarket economy country, and (B) [a] significant producer[ ] of comparable merchandise.” 19 U.S.C. § 1677b(c)(4). Pursuant to this statutory direction, Commerce’s regulations specify that, other than for valuing labor costs, it “normally will use publicly available information to value factors” and that it “normally will value all factors in a single surrogate country.” 19 C.F.R. § 351.408(c)(1) & (2) 4 (emphases added). As emphasized, Commerce “normally” will use pricing data from the surrogate country selected. The use of the word “normally” means that Commerce may select other data as warranted under the circumstances. Though Commerce, as is true of all agencies, is due deference for its reasonable interpretation of its own regulations, NSK Ltd. v. Koyo Seiko Co., 190 F.3d 1321, 1326 (Fed.Cir. 1999) (“we accord substantial deference to Commerce’s interpretations of its own regulations.”), the use of “normally” indicates that Commerce intends to give itself leeway in the selection of surrogate countries. However, Commerce cannot use this language to allow itself flexibility while simultaneously using the language to disallow consideration of data from other countries without regard for its own stated criteria. Cf. Shanghai Foreign Trade Enters. Co. v. United States, 28 CIT -, --, 318 F.Supp.2d 1339, 1351-52 (2004) (“Commerce’s decision to use Indian Import Statistics suffers from the same flaw that Commerce alleges as a basis for rejecting plaintiffs alternatives.”). For example, Commerce’s regulations acknowledge that “where a factor is purchased from a market economy supplier and paid for in a market economy currency, [Commerce] normally will use the price paid to the market economy supplier.” 19 C.F.R. § 351.408(c)(1). As such, the selection of a surrogate country does not re *1271 quire that the particular data that is used to value a specific factor of production also come from that surrogate country; rather, it appears that Commerce’s selection of a surrogate country creates, at most, a re-buttable presumption in favor of data from that surrogate country. In selecting the surrogate country, Commerce employs a four step process. First, Commerce compiles a list of countries that are at a level of economic development comparable to the country being investigated. Department of Commerce, Import Administration Policy Bulletin 04.1: Non-Market Economy Surrogate Country Selection Process at 2 (March 1, 2004), available at http://ia.ita.doc.gov/policy/bull 04-l.html. (“Policy Bulletin”). Commerce then ascertains which, if any, of those cited countries produce comparable merchandise. Id. Next, from the resulting list of countries, Commerce determines, which, if any, of the countries are significant producers of said comparable merchandise. Finally, Commerce evaluates the quality, e.g., the reliability and availability, of the data from those countries. Id. at 3. Upon review of these criteria, Commerce chooses the country most appropriate for use as a surrogate for the investigation. Here, Commerce listed five countries as economically comparable to the PRC: India, Indonesia, Sri Lanka, Pakistan and the Philippines. Memorandum from Jon Freed, Case Analyst, to File, through Edward C. Yang, Office Director & Robert Bolling, Program Manager, Re: Anti-dumping Investigation of Wooden Bedroom Furniture from the People’s Republic of China: Selection of a Surrogate Country, Dep’t of Commerce (March 8, 2004), P.R. Doc. 619. at 1 (“Surrogate Country Selection Mem.”). No party contests this finding. Id. at 5. Next, Commerce found that both India and Indonesia produced comparable merchandise. Commerce found that both countries are significant producers of comparable merchandise. Issues & Decision Mem., P.R. Doc. 1933 at 38, 40 (Cmt.2); but see id. at 41. Last, Commerce evaluated and compared the quality of Indian and Indonesian data. Although Respondents placed domestic Indonesian pricing data on the record, Commerce rejected this data in favor of Indonesian import statistics. Focusing then on Indonesian import statistics, Commerce found that Indian import statistics were preferable to its Indonesian counterpart because: (1) Indonesian data had been unsatisfactory in other investigations; (2) Indonesian “information was either unreliable or the Indonesian import statistics were reported in units for which the Department was unable to obtain a comparable value ... and (3) there existed gaps in the Indonesian data which required the use of gap-filler data that Commerce “prefers not to use unless there are clear distortions in the surrogate price import statistics .... ” Issues & Decision Mem., P.R. Doc.1933 at 42 (Cmt.2). Therefore, Commerce selected India as its surrogate country for this investigation. Respondents challenge Commerce’s approach in several respects: (1) that Commerce erred in rejecting Indonesian data^ — both its domestic and import statistics; (2) that India is not a producer of products identical to the subject merchandise; (3) that Indian production of the subject merchandise is not significant; and (4) Commerce erred in weighing the evidence, in the aggregate, that India was preferable to Indonesia. Though the weighing of the data is considered to be the fourth step of Commerce’s process of selecting a surrogate country, the court will address the issues regarding Indian and Indonesian data first, as the data issues are intertwined with all aspects of selecting a surrogate country. *1272 (1) Evaluation of Indonesian Data Commerce has indicated that it does not have sufficient data from Indonesia, that the data it does have is suspect, and that it has had data problems with Indonesia in the past. Issues & Decision Mem., P.R. Doc.1933 at 42 (Cmt.2). These comments seem to be directed primarily towards the use of Indonesian import data, rather than Indonesian domestic data. However, a closer examination of the record indicates that Commerce also considered the domestic Indonesian data in its evaluation. Under established case-law, “[t]he decision on which price to use — domestic or import — is based on which value will result in a more accurate normal value.” Rhodia, Inc. v. United States, 25 CIT 1278, 1286, 185 F.Supp.2d 1343, 1352 (2001); Hebei Metals & Minerals Imp. & Exp. Corp. v. United States, 28 CIT - — -,-, 2004 WL 1615597, *8 (2004)(“Hebei I”). Commerce may prefer one to the other so long as a reasonable mind could find that the one preferred is more accurate than the other. See Goldlink Indus., 30 CIT at -, 431 F.Supp.2d at 1334; cf. Yantai Oriental Juice Co. v. United States, 26 CIT 605, 617 (2002). Here, Commerce found that respondent-selected domestic Indonesian data price lists are not more reliable than import statistics because “single prices from a narrow source are not necessarily representative of an entire industry during the entire period of investigation.” Def.’s Resp. Pls.’ Surrogate Value Related Rule 56.2 Mots. J. Agency R. 26 (“Def.’s Br.”); Issues & Decision Mem., P.R. Doc.1933 at 42 (Cmt.2); Polyethylene Retail Carrier Bags from the People’s Republic of China, 69 Fed.Reg. 34,125 (Dep’t Commerce June 18, 2004) (notice of final determination of sales at less than fair value) and accompanying Issues and Decision Memorandum at 47 (Cmt.9) available at http://ia.ita.doc. gov/frn/summary/prc/04-13815-l.pdf (“the experience of a single producer is less representative of the cost of an input in a surrogate country.”). In contrast, import statistics encompass a broader range of pricing data that are more representative of an entire industry during the entire period of investigation. Commerce was also “unable to find substantial information or directories for Indonesian furniture manufactures [sic].” Surrogate Country Selection Mem., P.R. Doc. 619 at 5. In addition to finding that Indonesian domestic data or price lists were wanting, Commerce found that Indonesian import data (1) had been unsatisfactory in other investigations; (2) that “information was either unreliable or the Indonesian import statistics were reported in units for which the Department was unable to obtain a comparable value ... and (3) “because of the inadequacies of the Indonesian import statistics [Respondents] submitted gap-filler data from various sources that the Department prefers not to use unless there are clear distortions in the surrogate price import statistics.... ” Issues & Decision Mem., P.R. Doc.1933 at 42 (Cmt.2). Admittedly, there is some circularity to Commerce’s explanation. Commerce rejected Indonesian domestic data due to its lack of completeness, and then rejected Indonesian import data due to the need to use “gap-filler” data which it will not use without finding distortion in import data. However, Commerce’s finding on the whole that Indonesian data is unreliable and insufficient is supported by substantial evidence. Respondents do admit that some of the Indonesian data is unreliable or unusable. Id. Though respondents do point to gap-filler data (such as “Indonesian domestic prices for woods and processed woods for the period of investigation published by the International Tropical Timber Organization *1273 (TTTO’))/’ Mem. P. & A. Support Pis.’ & Pl.-Intervenors’ Rule 56.2 Mot. J. Agency R. 10 (“Pis.’ Br.”), this represents but one set of various inputs required for valuing the factors of production. Assuming without deciding, that the ITTO data is reliable, accurate, and contemporaneous, it is still reasonable for Commerce to find, on the record here, that in toto the combination of domestic and import data from Indonesia was not of a sufficient quality or amount to allow for the valuation of all the necessary factors of production, and for the calculation of financial ratios. Therefore, Commerce’s determination that the Indonesian data was unreliable is supported by substantial evidence. (2) Commerce’s finding that India was a producer of the comparable merchandise Commerce’s determination that India is a producer of comparable merchandise is also supported by substantial evidence. Commerce found evidence on the record to indicate that India produces comparable furniture. Surrogate Country Selection Mem., P.R. Doc. 619 at 5; Issues & Decision Mem., P.R. Doc.1933 at 39 (Cmt.2) (“Respondents do not argue that India does not produce wooden bedroom furniture.”). In particular, Commerce found that International Furniture Producers, an Indian company, as well as other companies were significant producers of wooden bedroom furniture. Surrogate Country Selection Mem,., P.R. Doc. 619 at 5. Respondents, contend, however, that the finding that India produces comparable merchandise is irrelevant in the face of the fact that Indonesia produces identical merchandise. Pis.’ Br. 13-14, 16. As Respondents noted in their brief, Commerce automatically considers a producer of identical merchandise to be a producer of comparable merchandise. Pis.’ Br. 13. Respondents argue that because Indonesia produces identical merchandise, and they contend India does not, Indonesia should automatically be selected as the surrogate country. Commerce, however, does not always choose the producer of identical merchandise, if there is one. Commerce explains in its Policy Bulletin that “[if] considering a producer of identical merchandise leads to data difficulties, the operations team may consider countries that produce a broader category of reasonably comparable merchandise.” Policy Bulletin at n. 6. It appears to the court that this analysis is in accordance with the legislative history of the governing statute: Because the Commerce Department may have difficulties in getting detailed data from countries not subject to investigation, the bill gives the Commerce Department authority to use “comparable merchandise” as the basis for foreign market value. Comparable merchandise is a broader category than the “such or similar” merchandise comparison which is usually used in antidumping investigations. S.Rep. No. 100-71 at 106 (1987) (Committee on Finance). Given that Commerce chooses the surrogate country and identifies comparable merchandise on a case-by-case basis, it is reasonable for Commerce to decline to make the producer of identical merchandise the automatic choice. The process of constructing an export price is a necessarily laborious and data-intensive process, and it is reasonable, and necessary, for Commerce to feel assured that the data it is employing is sufficient and reliable. Therefore, assuming that Indonesia is a producer of identical merchandise, that identity does not upset Commerce’s determination that India produces comparable merchandise. The statute does not require Commerce to find a producer of *1274 identical merchandise, but rather a producer of comparable merchandise. Though Commerce has indicated a preference for identical goods, that preference must sometimes take second-seat if the use of identical goods leads to data selection problems. Given Commerce’s finding that Indonesian data was unreliable, Commerce’s decision to look to comparable producers of subject merchandise was supported by substantial evidence, as was its determination that India is a producer of comparable merchandise. (3) Commerce’s finding that India was a significant producer of subject merchandise In addition to determining that a country produces comparable merchandise, the statute requires that the surrogate country be a “significant producer” of the comparable subject merchandise. Here, Commerce found that both India and Indonesia were significant producers of comparable merchandise. Issues & Decision Mem., P.R. Doc.1933 at 39 (Cmt.2) (“it would be illogical to conclude, as argued by Respondents ..., that Indonesia is a significant producer of furniture while India is not.”). Respondents contest this finding asserting that India is not a significant producer of wooden bedroom furniture. Commerce identifies a significant producer based on a totality of the circumstances, and makes its decision concerning significance on a case-by-case basis. Policy Bulletin at 3. Commerce ascertained that “fixed standards such as ‘one of the top five producers’ ” are not helpful as the meaning can differ significantly from case to case. 5 Id. Commerce further explained that the selection of a surrogate country is, of course, highly dependent on the available data. Here,. Commerce determined that India was a significant producer of comparable merchandise. In particular, Commerce determined that there was at least one major Indian manufacturer of wooden bedroom furniture, IFP, and that there were other producers of comparable merchandise. Surrogate Country Selection Mem., P.R. Doc. 619 at 5-6. Commerce has also indicated that it was able to find additional companies that produce comparable merchandise, in addition to further information and directories with regard to 416 Indian furniture manufacturers. Id. at Attachs. II & III, fr. 11, 14. Commerce found that there are upwards of 11,000 producers of furniture in India, and a furniture industry output of $1.7 billion (compared with Indonesia’s output of $1.9 billion). Issues & Decision Mem., P.R. Doc.1933 at 39 (Cmt.2). Commerce also indicated that it was able to find nine Indian surrogate financial statements. Id. at 67 (Cmt.3). Consequently, Commerce’s finding that India is a significant producer of comparable merchandise is supported by substantial record evidence. (4) Weighing the choice between Indonesia and India As the discussion above indicates, Commerce’s decision to opt for India as the surrogate country in this investigation, as opposed to Indonesia (or any of the other economically comparable countries) was driven in large part by data concerns and considerations. Commerce had access to *1275 and was apprised of numerous sources of information from India. Commerce was also able to locate Indian company directories, Surrogate Country Selection Mem., P.R. Doc. 619 at 5 & Attach. III fr. 14, and Commerce had experience utilizing the Indian import statistic data base, the Monthly Statistics of Foreign Trade in India (“MSFTI”). Respondents point to various criteria, including the fact that Indonesia is a net exporter of wooden bedroom furniture while India is not, Pis.’ Br. 17, and the size of the industry, to indicate that Commerce should have chosen Indonesia instead of India as the surrogate country. Id. at 14-15 (claiming that the portion of India’s furniture production that is wooden furniture is 60 percent and the proportion that is bedroom furniture is 20 percent); Id. at 17 (“the values of [India’s] exports were 38 million and 30 million rupees (about $875 thousand and $690 thousand)” in 2002 and 2003 respectively). However, just as previous case law has established that Commerce need not pick the “most comparable economy”, Tehnoimportexport v. United States, 15 CIT 250, 256, 766 F.Supp. 1169, 1175 (1991) (emphasis in original), Commerce need not pick the most significant producer. Despite the fact that the Indian data may or may not be a perfect fit for every surrogate value, this court has noted time and again that the estimation of a normal value using surrogate values is an inexact science. See, e.g., Nation Ford Chem. Co. v. United States, 166 F.3d 1373, 1377 (Fed. Cir.1999) (“the process of constructing foreign market value for a producer in a nonmarket economy country is difficult and necessarily imprecise”) (citing Sigma Corp. v. United States, 117 F.3d 1401, 1408 (Fed.Cir.1997)). While accuracy is a touchstone, Commerce often finds that it has to choose between two (or more) suboptimal data sources. In such a scenario, where Commerce is choosing between two less-than-ideal data sources, its decision must be upheld provided it is supported by substantial evidence. Here, Commerce has demonstrated, through substantial evidence, that it was able to locate data from India that it knew to be reliable having previously worked with the data. Surrogate Country Selection Mem., P.R. Doc. 619 at 3. Therefore, based on the preceding analysis, Commerce’s determination to use India as the surrogate country for this investigation was appropriate; as a factual matter that decision was also supported by substantial evidence. The legislative history of the governing statute does indicate that “if possible,” Commerce should utilize data based on production of “the same general class or kind of merchandise using similar levels of technology and at similar levels of volume as the producers subject to investigation.” Conf. Rep. at 591. Though the desire for comparability of technology is clear, and one could argue, optimal, this desire is qualified by the phrase “if possible.” Here, Commerce reasonably determined that the quality of Indonesian data would be a hindrance in calculating surrogate values; therefore it would not be possible to utilize Indonesian data, whether or not the Indonesian furniture industry proved to be a closer match in production process to the Chinese furniture industry. Issues & Decision Mem., P.R. Doc.1933 at 42. The statute requires Commerce to look to the comparability of merchandise, and Commerce acted in accordance with law in determining that comparability of merchandise does not necessarily require comparability of industry. 19 U.S.C. § 1677b(c)(4). Ultimately Commerce can, and does, mix and match the data that it chooses in its surrogate value selection, for instance through the use of gap-filling data. See *1276 Lasko Metal Prods. v. United States, 43 F.3d 1442, 1445-46 (Fed.Cir.1994); Nation Ford, 166 F.3d. at 1378. Because Commerce can mix and match data when it finds data from its primary source to be wanting, the surrogate country selection amounts to a nod or a presumption as to what will be Commerce’s “go-to” country. In this investigation, the choice of India had a direct bearing on the choice of the main data set employed by Commerce in its surrogate value selection, namely in Commerce’s selection of MSFTI as its main data set in valuing the factor inputs. B. Monthly Statistics of Foreign Trade in India (1) In general The valuation of factors of production in a nonmarket economy is governed by 19 U.S.C. § 1677b(e). As noted previously, this section of the statute instructs Commerce to value factors of production “based on the best available information regarding the values of such factors in a market economy country or countries considered to be appropriate by the [Secretary].” 19 U.S.C. § 1677b(c)(l). In addition, Commerce has promulgated regulations specifying that the information utilized is “normally” to be “publicly available” and that, except to value labor, the Department will normally value all factors using data from a single surrogate country. 19 C.F.R. § 351.408(c). 6 While Commerce has not promulgated additional regulations to govern its selection of data for the valuation of factors of production, it has developed policy preferences relating to its data choices. Specifically, Commerce prefers data that is (1) a non-export average value; (2) most contemporaneous with the period of investigation (“POI”); (3) product-specific; and (4) tax-exclusive. Memorandum to File, From Michael Holton, Subject: Preliminary Determination Factors Valuation Memorandum: Wooden Bedroom Furniture from the People’s Republic of China, Dep’t of Commerce (June 17, 2004), P.R. Doc. 1329 at 2 (“Factors Valuation Mem.”). See, e.g., Polyethylene Retail Carrier Bag Comm. v. United States, 29 CIT -, 2005 WL 3555812, **14-15, **18-25 (Dec. 13, 2005). (2) MSFTI as a primary data set In this investigation, the application of the factors outlined above led Commerce to rely on the Monthly Statistics of Foreign Trade in India data for the valuation of raw materials. Factors Valuation Mem. at 4. MSFTI, published by the Government of India and available through the World Trade Atlas, provides the quantity and value of all Indian imports, reported by Harmonized Tariff Schedule of India (“HTS[I]”) headings and subheadings. Commerce argues that MSFTI data represents the best available information for the valuation of raw material inputs because MFSTI data were publicly available, contemporaneous and representative of all Indian imports, “representative of the inputs in question,” and tax-exclusive. Def.’s Resp. Br. Pls.’ Surrogate Value Related Mot. J. Agency Record 36 (“Def.’s Br.”); see also Issues & Decision Mem., P.R. Doc.1933 Cmts. 10, 17, 25 & 27. Com *1277 merce also notes that it has utilized MSFTI in previous investigations. Id. at 137 (Cmt.10) & 206 (Cmt.25) (Commerce has a long-standing preference for MSFTI unless it is unreliable or distorted). Respondents challenge Commerce’s use of MSFTI, arguing that MSFTI data is overbroad; that better import data is available from Infodrive India and IBIS 7 ; that MSFTI does not always capture the inputs that are used by the Indian wooden furniture industry; that MSFTI data is inaccurate and unreliable and that there are other sources of data that provide specific and accurate domestic data. Pis.’ Br. 32-46. As a threshold matter, Commerce has determined that when it selects import statistics as a means of valuing factors of production for a non-market economy, it would rather choose a broader range of statistics stating that “[a] broad, average price from import statistics is reliable is [sic] because the average includes a range of prices.” Def.’s Br. 52; see also Issues & Decision Mem., P.R. Doc.1933 at 137 & 206. Commerce further elaborated that, “[a]n average price representing a subset of imports is not more accurate than an average price of all imports into India.Def.’s Br. 52; see also Issues & Decision Mem. P.R. Doc.1933, at 206 (Cmt.25) & 214-217 (Cmt.26). Respondents propose Infodrive India and IBIS as alternative sources of import data, rather than MSFTI, arguing that Infodrive India and IBIS provide a detailed breakdown of Harmonized Tariff Schedule (“HTS”) subheadings. This line-by-line information, Respondents claim, would allow Commerce to make a more precise match between the input used by the Chinese manufacturers and the surrogate Indian import. Issues & Decision Mem., P.R. Doc.1933 at 131 (Cmt.10); Pls.’ Br. 36. All parties agree that Infodrive India and IBIS do not provide data on all imports into India. 8 The court finds that, in general, Commerce’s preference for a broader data set is reasonable and supported by substantial evidence. In essence, Commerce is arguing that where it has a choice between underinclusive data (which does not capture all the inputs used by wooden furniture manufacturers) or overinclusive data (which includes some data which is not used as an input, but captures all the inputs because of its broad range), it will choose overinclusive data. As Commerce is faced with a choice between two imperfect options, it is within Commerce’s discretion to determine which choice represents the best available information. Commerce’s decision to rely on the MSFTI data, as a general database, to value raw material factors of production, instead of Infodrive India or IBIS data, is *1278 supported by substantial evidence in that the record supports Commerce’s conclusion that the MSFTI data is more inclusive than the Respondents’ alternatives. (3) MSFTI as applied to individual factors. However, while Commerce’s choice of MSFTI data, as a general database, rather than Respondents’ alternatives, is appropriate, Commerce’s individual determinations, on a factor input by factor input basis, must also be supported by substantial evidence. If Commerce’s specific data choices do not actually include or capture the factor or input it is estimating, or a reasonably comparable item, such a choice is not supported by the record; for example, if Commerce were estimating the cost of hard-cover law casebooks by relying on the average cost of paperback legal thrillers, while another data set provides the cost of hard-cover textbooks, its choice could not be sustained. Cf. Hebei Metals & Minerals Imp. & Exp. Corp. v. United States, 29 CIT-,-, 366 F.Supp.2d 1264, 1272-73 (2005) (“Hebei II”) (Commerce cannot assume that the “others” category includes the coal at issue, without establishing what category of coal was used by the respondents, and without establishing that the category did “cover the type of coal used in Hebei’s production process.”); Guangdong Chems. Imp. & Exp. Corp. v. United States, 30 CIT-, -, 414 F.Supp.2d 1300, 1311 (2006) (“Even if the court were to conclude that Commerce produced substantial evidence demonstrating that azelaic and sebacic acid are priced similarly, that would not justify Commerce’s decision to abandon a more produet-specific data source.”). Additionally, if the data is heavily weighted by a disproportionate inclusion of higher priced or lower priced data, such that Commerce is systematically overvaluing or undervaluing the factors of production, the broad range of the statistics would not, in and of itself, render the data reliable. See Goldlink Indus., 30 CIT at -, 431 F.Supp.2d at 1334 (“Since the presumption is that NME data is distorted, Commerce must find a reasonable surrogate value. Logically then, Commerce cannot use a surrogate value if it is also distorted, otherwise defeating the purpose of using a surrogate value rather than the actual export value.”). In addition, although the court has, in specific instances, approved the use of import data, and the use of MSFTI data in particular, the court has always emphasized that in order for import data to be used, there must be reason to believe that the industry in question would use imported inputs. 9 See Hebei I, 28 CIT at-, 2004 WL 1615597, *8 (“Commerce here did not explain why an Indian manufacturer would pay for imported coal”); Yantai *1279 Oriental Juice Co. v. United States, 26 CIT 605, 617 (2002) (“Here, Commerce has produced no evidence tending to lead to the conclusion that India’s domestic AJC producers would use imported as against domestic coal.”); Wuhan Bee Healthy Co. v. United States, 29 CIT-, -, 374 F.Supp.2d 1299, 1310 (2005) (Commerce gives “no reason[ ] as to why imported coal provides the best surrogate value.”). If it is unlikely that the domestic industry would use imported inputs, and there is domestic data available, then Commerce’s choice of import data to value factor inputs may not be reasonable. One example of when the domestic industry would choose to source its factor inputs domestically would be when the price of the imported good is markedly higher than the price of the domestic good. Hebei II, 29 CIT at -, 366 F.Supp.2d at 1274 (“[T]he preference for domestic data is most appropriate where the circumstances indicate that a producer in a hypothetical market would be unlikely to use an imported factor in its production process. The most obvious circumstance occurs where the import price is significantly greater than the domestic price.”). Another example would be when the import statistics themselves demonstrate that the specific input is not imported into the country. While Commerce may establish criteria in order to guide its data selection process, this does not relieve Commerce of the obligation to evaluate the relative accuracy of domestic and import data in valuing factors of production. See Yantai 26 CIT at 617 (“Commerce nowhere explains how the use of seemingly more expensive imported coal data is the best available information establishing the actual costs incurred by Indian AJC producers.”). Commerce itself, when noting potential price discrepancies in import data, has employed benchmarks in order to evaluate the reliability of MSFTI data. China First Pencil, 30 CIT -, 427 F.Supp.2d 1236 (on voluntary remand, Commerce examined MSFTI data for the price of pencil cores by comparing it with price quotes from the United States, since Commerce was unable to obtain price quotes from India). (a) Mirrors The court turns to Respondents’ challenge of the use of specific MSFTI data for the valuation of mirrors, paints, and cardboard. Respondents claim that the MSFTI data is demonstrably inaccurate for valuing these specific factors of production. Pis.’ Br. 27, & 40-46. Respondents attempt to demonstrate that MSFTI is inaccurate in valuing mirrors by alleging that (1) there is evidence that Indian furniture manufacturers do not use imported mirrors in their manufacturing process; (2) in this instance MSFTI is not inclusive of the mirrors utilized; and (3) the MSFTI data utilized was overly distorted by the inclusion of specialty mirrors. (i) Imported Mirrors As Inputs Respondents provide a range of data to demonstrate that the import prices for mirrors selected by Commerce are considerably higher than domestic prices, suggesting either that the import prices are distorted or that Indian producers would not use imported mirrors as an input for wooden bedroom furniture. See, e.g., Hebei II, 29 CIT at-, 366 F.Supp.2d at 1274 (companies are less likely to use imported inputs when the prices of imported inputs are higher than domestic inputs). Respondents argue that the actual price for 5mm mirrors used in wooden furniture production in India is within a range of between $9.35/m2 and $13.13 /m2 instead of the Final Determination value of $60.13 /m2 for mirror imports (a range of 4.6 to 6.4 times greater than the alternate prices). Pis.’ Br. 27. Repondents refer to: (a) record data from an Indian glass industry publication (“Glass Yug”); (b) actual average period of investigation prices paid *1280 for mirrors used by two Indian wooden furniture producers (Tarun Vadehra and Highland House); 10 (c) and actual average prices paid for mirrors by a large Indonesian wooden bedroom furniture producer, Goldfmdo. Id. Additionally, there is evidence on the record that at least one furniture producer in India only sourced mirrors domestically during the period of investigation. See Ernst & Young Private Ltd., Report on the Agreed Upon Procedures on the Purchases Ledger of Highland House Private Ltd. (Apr. 1, 2004), Attach, to Letter from John D. Greenwald, Wilmer, Cutler, Pickering, Hale & Dorr, LLP., on behalf of Lacquer Craft Mfg. Inc. & Markor International Furniture (Tianjin) Manufacture Co., Ltd., to The Honorable Donald L. Evans, Secretary of Commerce, Re: Wooden Bedroom Furniture from the People’s Republic of China: Surrogate Country Submissions (Apr. 20, 2004), P.R. Doc. 770 at 56 (fr.13) (“As per the list provided by Lacquer Craft, Ernst & Young was required to compile prices of standard 5/6mm thickness type plain mirrors purchased by Highland House during the period. During the period, Highland House purchased plain mirrors from domestic suppliers only.”). Commerce did not address the assertion that imported mirrors were not used by Indian furniture manufacturers. Instead of addressing the Respondents’ concerns with the MSFTI data, Commerce chose ■ to attack the quality of the data proffered by Respondents,, claiming that the unreliability of the data negated its ability to serve even as a means of evaluating the MSFTI data. The court finds that Commerce’s determination that the MSFTI data was accurate with respect to mirrors is not supported by substantial evidence, because Commerce did not evaluate the inaccuracies of the MSFTI data set. See Shanghai Foreign Trade Enters. Co. v. United States, 28 CIT-,-, 318 F.Supp.2d 1339, 1352 (2004)(Commerce’s determination was not supported by substantial evidence when “Commerce summarily discarded the alternatives as flawed but did not evaluate the reliability of its own choice.”). Commerce argues that Yantai, Hebei II, and Shanghai Foreign Trade are inapplicable here as “Commerce did not summarily discard the alternatives as flawed without evaluating the reliability of its own choice.” Def.’s Br. 47. However, as explained below, here Commerce’s reasons for rejecting other data were either applied on an inconsistent basis, or did not actually provide a rational reason for rejecting the data. (ii) Glass Yug Commerce provided three reasons for dismissing the Glass Yug data: (1) the data do not cover the entire POI; 11 (2) the *1281 Glass Yug data are very specific and Commerce does not know whether or not Respondents use the same type of mirrors; and (3) the Glass Yug data are not specific enough because it does not contain information on the sizes of the mirrors and whether or not the mirrors are beveled. Issues & Decision Mem., P.R. Doc.1933 at 204-206 (Cmt.25). First, regarding the POI, Commerce may not always elevate contemporaneity, or its desire to have data covering an entire POI, over the need for accuracy. See Hebei II, 29 CIT at-, 366 F.Supp.2d at 1275 (finding that contemporaneity is but one factor to be considered in looking at data and noting that “the Court has previously found contemporaneity to be insufficient to explain why an import price is the best available information for establishing the actual costs incurred by a producer.” (citing Yantai, 2002 WL 1347018, *9)). If the domestic data proves to be otherwise accurate, then it may well be preferable to use accurate data, compiled over a shorter period of time, then inaccurate or unreliable data that covers the entire POI. Commerce’s second and third contentions rely on Commerce’s desire not to have underinclusive data. Essentially, Commerce is arguing that since it cannot match Glass Yug data with Respondent data, either because the Glass Yug data are not specific enough or because the data are too specific, use of the Glass Yug data would lead to an underinclusive data set. 12 Respondents have asserted that they have placed evidence on the record that shows the width of their mirrors. Respondents note that Shing Mark has reported its mirrors to be 3mm, 5mm, and 6mm in depth, Steptoe & Johnson, LLP, Case Brief on Behalf of Shing Mark Co., Ltd., (Oct. 6, 2004) Attach, to Letter from Jack R. Hayes, Steptoe & Johnson, LLP, on behalf of Shing Mark Co., Ltd. to the Honorable Donald L. Evans, Secretary of Commerce, Re: Antidumping Duty Investigation on Wooden Bedroom Furniture from the People’s Republic of China: Submission of Case Brief on Behalf of Shing Mark Enterprise Co., Ltd., (Oct. 7, 2004), P.R. Doc. 1854, fr. 58 (“Shing Mark Case Brief’), while there is evidence on the record indicating that Dorbest’s mirrors are 5mm in thickness, Dorbest Ltd., Second Supplemental Sections C & D Response (June 15, 2004), Attach, to Letter from Jeffrey S. Grimson, Kaye Sholer, LLP, on behalf of Dorbest Ltd., to the Honorable Donald L. Evans, Secretary of Commerce, Re: Response to DOC’s June 8, 2001, Second Supplemental Sections C-D Questionnaire in Wooden Bedroom Furniture from China, (Inv. No. A-570-890), (June 15, 2004), Prop. Doc. 599, fr. 314 & 318 (“Dorbest June 15 Submission”); See Pis.’ Br. 42. Respondents maintain that this is sufficient information for Commerce to match Glass Yug data with Respondents’ inputs, because Glass Yug data provided prices for mirrors with thickness of 2.5, *1282 3.5, 4.0, 5.0 and 6.0 millimeters. 13 Pis.’ Br. 42; Shing Mark Apr. 16 Submission, P.R. Doc. 761, fr. 446. Additionally, it appears inconsistent for Commerce to require specificity for one data set, while allowing for a broader data set that has no indication either as to whether it includes mirrors with “beveling” or etching, as is the case with MSFTI. See Hebei II, 366 F.Supp.2d at 1273 (“A broad and unsupported coal value falls short of a substantial evidentiary basis just as a narrow and unsupported coal value does.”); Shanghai, 318 F.Supp.2d at 1352 (Commerce must apply its standards as consistently to its selected data set as to other data sets). At the very least, none of Commerce’s arguments with respect to Glass Yug address why this data should not be viewed as probative towards a view that Commerce’s chosen valuation is too high and/or inaccurate. (iii) Tarun Vadehra, Highland House and Goldfindo Commerce rejected the Tarun Vadehra and Highland House information because it found that the information was not “publicly available” insofar as it was not information that could “be duplicated by the Department, the Petitioners, or anyone else that lack[ ] access to the confidential records from which they were derived.” Issues & Decision Mem., P.R. Doc.1933 at 162-63. (Cmt.17). Respondents argue that this is not how Commerce has generally defined publicly available. Pis.’ Br. 34. The court does not have to reach this issue with regard to the Tarun Vadehra and Highland House data, because Commerce also explained that it found that this data was not representative of all Indian prices. Though Commerce considers several issues in evaluating the merits of various data, see supra, p. 30, Commerce’s determination is supported by substantial evidence in that the data from two Indian producers is not representative of the country as a whole. See Retail Carrier Bags, 29 CIT at-, 2005 WL 3555812, *17 (finding that the selection of import statistics in valuing black inks on the basis of the import data being “country-wide” data was supported by substantial evidence, as evidence submitted to demonstrate that Indian import statistics counted as an even smaller percentage of sales of the relevant inks was not provided during *1283 the administrative review.). Commerce also implicitly rejected the data provided by Respondents from the Indonesian company, Goldfindo stating that “the Department has not considered the factor values derived form [sic] Goldfindo because Gold-findo is an Indonesian company and the Department has determined to use India as the surrogate country in this investigation.” Issues & Decision Mem., P.R. Doc. 1933 at 163 (Cmt.17). The court notes that though Commerce is not obligated to value its factors of production from just one surrogate country, see, e.g., 19 U.S.C. § 1677b(e)(1); Nation Ford Chem. Co. v. United States, 166 F.3d 1373, 1378 n. 5 (Fed.Cir.1999) (the antidumping duty statute “does not preclude consideration of pricing or costs beyond the surrogate country if necessary”), Commerce has also noted that the same argument against using Tarun Vadehra and Highland House data — namely that the data do not cover country-wide prices and are too narrow to be considered for surrogate valuation— applies as well to Goldfindo data. Issues & Decision Mem., P.R. Doc.1933 at 163 (Cmt.17) (“the Department’s practice is to not use price quotes to value factors when other usable, reliable information is available.”). (iv) The MSFTI data is either noninclusive or distortive of mirror inputs Respondents also argue that the Info-drive India and IBIS data indicate that the mirrors reported as being imported under subheading 7009.91.00 of the Harmonized Tariff Schedule of India (“HTS[I]”) included Taiwanese exports of rearview mirrors for automobiles to an Indian company by the name of “Enginetech” or mirrors from Germany including a “chiara” bathroom mirror and a telescopic mirror. Issues & Decision Mem., P.R. Doc.1933 at 194 (Cmt.25) & 183 (Cmt.2.4); see also Pis.’ Br. 26. According to Respondents, these are higher-priced specialty mirrors, the inclusion of which has a distortive effect on the valuation of the mirror inputs used in furniture production. Respondents also claim that, in addition to examining the line-by-line data available from Infodrive India to ascertain what precise goods were listed under HTS[I] subheading 7009.91.00 (“mirrors, other, unframed”), they examined Taiwan’s corresponding export data for “mirrors, other, unframed” and that the Taiwanese export data show a low volume of exports to India under this HTS subheading. Issues & Decision Mem., P.R. Doc.1933 at 191 (Cmt.24). Respondents further claim that the Taiwan export data show a correspondingly larger volume of exports under the HTS subheading for rearview mirrors for automobiles. Id. Respondents theorize that the Indian government’s updating of the Tariff Classification to the eight-digit subheading level on February 1, 2003 (immediately prior to the POI) led to problems in the classification of entries. Id. at 182 (Cmt.24). Respondent Shing Mark argues that the “reclassification of the Tariff Schedule resulted in the creation, renaming, moving, or deletion of various headings and subheadings which, in turn, resulted in the misclassification of entries in this new tariff system.” Id. Commerce does not directly address this allegation, nor the apparent discrepancy between Taiwanese export data and Indian import data. Commerce rather points to the existence of an HTS[I] subheading for rearview mirrors which, Commerce argues, would negate the possibility of rear-view mirrors being classified under subheading 7009.91.00. Id. at 205 (Cmt.25). This does not address the issue of potential misclassification — if rearview mirrors are misclassified under “mirrors, other, unframed” the existence of a subheading for rearview mirrors proves nothing. *1284 (v) Infodrive India Rather than directly addressing Respondents arguments, Commerce rejected any attempt to use Infodrive India data to show that the MSFTI data was inaccurate on three bases: (1) Commerce could not use the more specific data provided in Infodrive India because Respondents’ description of their inputs was not specific enough to allow Commerce to make an exact match between Infodrive India and the Respondents’ inputs; (2) the data in Infodrive consisted of “non-quantifiable unit measurement^]”; and (3) Commerce found Infodrive India’s data to be unreliable. Issues & Decision Mem., P.R. Doc. 1933 at 139 (Cmt.10) & 203 (Cmt.25). The court will address each point in turn. 14 With respect to the first point, Commerce found that the input descriptions provided by Respondents did “not provide sufficient descriptions or distinguishable characteristics that would allow the Department to search the voluminous Info-drive India data and IBIS data to obtain accurate surrogate-value information.” Issues & Decision Mem., P.R. Doc.1933 at 138 (Cmt.10). Regardless of whether such a search is possible, this point does not render the Infodrive India or IBIS data ineffective as a benchmark or other means of testing the MSFTI data. If Infodrive India does, as contended by the Respondents, provide a line-by-line item breakdown of imported inputs into India, then a claim that there is no way to match inputs with the Infodrive India data is not responsive to whether or not Infodrive India casts light on potential inaccuracies in the MSFTI data set. Turning to the second aspect of Commerce’s analysis, Commerce concluded that Infodrive India information is not usable because the unit measurements vary and are non-quantifiable. Respondents note, however, that “[i]f anything, the fact that the units of measure differ greatly [in Infodrive India] supports Dorbest’s argument that the import statistics incorporate a hodgepodge of product (much of which was misclassified) and therefore is not suitable for use as a surrogate value.” Reply Br. Pis. Dorbest Ltd., Rui Feng Woodwork (Dongguan) Co., & Rui Feng Lumber Supp. Rule 56.2 Mot. J. Agency R. Relating to Issues Surrogate Country, Surrogate Values & Financial Ratios (“Dorbest Reply Br.”) 6, fn.9. The court finds merit in Respondents’ argument. Given that Respondents offered the Info-drive India data to demonstrate alleged inaccuracies and misclassifications in the MSFTI data for India, Commerce’s reason for rejecting the data set only buttresses Respondents’ argument, and does not provide a reason for not using Infodrive India to ascertain if there are problems with MSFTI’s data set with respect to mirrors. Finally, Commerce determined that it was Infodrive India data, and not MSFTI data, that was unreliable. In its Issues and Decision Memorandum that accompanied Commerce’s final determination, Commerce stated: The only information on the record that India’s HTS reclassification resulted in any misclassifications under the Indian Tariff Schedule is from the Infordrive [sic] India data. In fact, we found that the MSFTI information from the World Trade Atlas does not contain the same *1285 misclassification as those contained in Infodrive India. Therefore, we find that, if India’s reclassification of the Tariff Schedule resulted in any misclas-sifications of import items, it is Infodrive India’s data that is unreliable because these data are the only data that report such misclassification. The Department observes further that the World Trade Atlas reports the official MSFTI data which may account for Infodrive India’s misclassifieations. Issues & Decision Mem., P.R. Doc.1933 at 138-39 (Cmt.10). It appears to the court that, in essence, Commerce is making two points: (1) the MSFTI data is correct because it is corroborated in the World Trade Atlas even though the World Trade Atlas simply compiles the MSFTI data reported to it by the Indian government; (2) because MSFTI and the World Trade Atlas report the official data, they are presumed to be accurate and if only one source contradicts this data, it is the contradictory source that must be unreliable and inaccurate. Respondents argued before Commerce that Commerce had utilized Infodrive India data in a previous investigation: Certain Color Television Receivers From the People’s Republic of China, 69 Fed.Reg. 20,594 (Dep’t Commerce April 16, 2004) (notice of final determination of sales at less than fair value and negative final determination of critical circumstances). 15 Dorbest Reply Br. 5. Commerce, in using the Infodrive data in Color Televisions determined that: the data on which Infodriveindia [sic] is based is not private at all, but rather is Indian customs data. Because we initially shared TCL’s concerns about the source of this data, early on in this investigation, we contacted Infodrive India Pvt. Ltd. (Infodrive), the company responsible for maintaining the Infodri-veindia [sic] website, and inquired about its data collection methods. According to Infodrive officials, Infodrive: 1) obtains the information in question from official Indian customs data; 2) receives daily customs data transmitted each month from the Indian customs department; and 3) presents the Indian customs data exactly as it is received, without additions or deletions. Memorandum from Laurie Parkhill to Jeffrey A. May, Issues and Decision Memorandum for the Antidumping Duty Investigation of Certain Color Television Receivers from the People’s Republic of China, at 43 (Cmt.9)(April 16, 2004) available at http://ia.ita.doc.gov/frn/summary/ prc/04-8694-l.pdf. 16 Therefore, according to Commerce’s own explanation, Infodrive India presents Indian government import data that it receives on a monthly basis from the Indian customs department. Moreover, Infodrive India data appears to be the same data provided by MSFTI (through the World Trade Atlas) in a disaggregated form, providing descriptions of the items that are imported and classi *1286 fied under a particular HTS[I] subheading. The view that this is a subset of the same data set, but more detailed, is supported by the assertion by Respondents that “the total Rupee value of imports from Taiwan from Infodrive is 2,665,-062.75, while the total imports in the MSFTI from Taiwan is 0.002665 Billion, or 2.665,000 (World Trade Atlas data rounds figures).” Dorbest Reply Br. 6. Therefore, if Infodrive India data provides a breakdown of the import data reported in MSFTI, it is unreasonable for Commerce to conclude that Infodrive data is unreliable or contains misclassifications, while simultaneously claiming that MSFTI is both reliable and contains no inaccuracies. In addition, because the record suggests that the disaggregated line-item data indicates that items manufactured by “En-ginetech” are potentially rearview 17 there is a sufficient reason to further investigate whether or not rearview mirrors (and other speciality mirrors) are included in the basket category selected by Commerce to value mirrors. Regardless of whether or not Commerce finds it appropriate to use the Infodrive India data to value mirrors, the Infodrive India data can prove to be illuminating as to the nature of the product actually being valued within a specific (and in this case basket) 18 HTS[I] subheading. (vi) Commerce’s evaluation Having concluded that Infodrive India data was unreliable, Commerce examined the MSFTI data in order to ascertain whether or not there was any distortive effect as a result of imports from Taiwan and Germany. Commerce employed the MSFTI data for mirrors from Indonesia as a benchmark 19 , and calculated the POI *1287 average prices for Indonesia and found that the POI average price from both Taiwan and Germany were lower than that of Indonesia.' Issues & Decision Mem., P.R. Doc.1933 at 204 (Cmt.25). Commerce concluded that this examination of the data demonstrated that “to the extent that any imports were misclassified from Taiwan there is no record evidence that these were distortive as the value comparison described above clearly indicates.” Id. at 205 (Cmt.25). While this attempt at benchmarking may indicate that the Taiwanese and German data were at least as representative of the goods classified in 7009.91.00 as the Indonesian imports, it does not address the claim that the Taiwanese and German mirrors include specialty mirrors, and that specialty mirrors lead to a higher valuation of the mirrors that are used in making wooden furniture sets. If the distortion alleged by the Respondents is limited to a problem with the data for imports from Taiwan and Germany, then this evaluation by Commerce would demonstrate that the problems with the Taiwanese and German data were not distortive. However, if the problem with the mirror data is endemic to the subheading chosen by Commerce to value the mirrors, then the distortions could also exist within the Indonesian data, and the evaluation does not demonstrate that the Taiwanese and German data do not have distortions. Additionally, testing import data against import data does not answer the broader issue raised by Respondents, i.e., that the import data suggests higher prices for imported mirrors than domestic mirrors, which would suggest that furniture manufacturers would not purchase imported mirrors as inputs. From this examination of the record and analysis it appears that Commerce never examined or explained evidence on the record that would seem to indicate (1) that the Indian furniture industry does not use imported mirrors as an input; or (2) if the Indian furniture industry does use imported mirrors as an input, that the data Commerce is employing actually does not capture the price of plain mirrors utilized in furniture production in India (i.e., Commerce only reviewed whether the data is indeed overinclusive of plain mirrors, rather than whether the data was not inclusive at all or distorted). See, e.g., Issues & Decision Mem., P.R. Doc. at 198 (Cmt.25) (“Respondents [contend] that the core problem is that the Indian import statistics are shaped by imports of products not used in furniture production.”). Commerce must reexamine the MSFTI data, in comparison to other data on the record, and/or determine that the data does not include specialty mirrors. In sum, on this issue, Respondents have placed on the record data from four different sources indicating prices that are significantly lower than the import-data-based values selected by Commerce, yet Commerce has not evaluated the MSFTI mirror data vis-a-vis the benchmarks offered by the Respondents, or vis-a-vis any other non-MSFTI source. 20 Commerce *1288 has rendered it impossible to demonstrate any inaccuracies or distortions .in the data it has selected, by requiring data sets being offered for comparisons to be faultless. This is a standard that even Commerce’s data set cannot meet. Respondents have also placed on the record a colorable claim that the data set includes data that is either not representative of the factor input or has a highly distortive effect on the factor valuation. Because Commerce has not analyzed these factors, or Respondents’ claims, the court cannot conclude that Commerce’s conclusions with regard to mirrors are supported by the record. (b) Paints Respondents allege that using MSFTI to provide a surrogate value for .paints is not in accordance with law or supported by substantial evidence, because it leads to an inaccurate valuation of paint inputs. Respondents allege that they have placed evidence on the record that indicates that the surrogate value chosen by Commerce is higher than the price paid by wooden bedroom furniture manufacturers, and that the HTS[I] heading chosen to value paint is overly broad, insofar- as it also captures paint allegedly used for automobiles and shipbuilding. Pis.’ Br. 26-27. Respondents submitted information from (1) a major Indian paints supplier, Asia Paints; (2) Infodrive and IBIS; and (3) Highland House and Tarun Vadhera, that they claim provides better and more accurate information for valuing the paint factor input, in addition to demonstrating that the value selected by Commerce is inaccurate. Id. 26-27, 32-34; Commerce rejected all surrogate values proffered by Respondents, and chose to value paint under a four digit heading for paint, HTS 3208. Issues & Decision Mem., P.R. Doc.1933 at 214 (Cmt.26). - Respondents specifically allege that the surrogate value chosen by Commerce for paint is overvalued, pointing to information they have placed on the record that their price range for paint extends from $2.18 to $4.03 a kilogram, while the surrogate value selected by Commerce was $4.48/kg. 21 They suggest that the very fact that the values they have culled from various data sources are lower than Commerce’s chosen values is in and of itself sufficient reason for Commerce’s valuation to be doubted. However, it appears to the court that this is not the test suggested by Hebei I, 28 CIT-, 2004 WL 1615597 and Hebei II, 29 CIT-, 366 F.Supp.2d 1264. Hebei I & Hebei II stand for the proposition that import prices should not be used when an industry uses domestic inputs, and that one means of ascertaining whether or not domestic inputs are used is by comparing domestic to imported prices. Hebei II 366 F.Supp.2d at 1274. However, in the case at bar, in the table comparing various paint values provided by Respondents, Respondents provide information from Info-drive India from two companies, “Advance Paints and Linea Coats PVT, that were exhibitors at Tndiawood 2004’ and, unlike the other importers, import the types of *1289 paint ... that are used to finish wooden bedroom furniture.” Pis.’ Br. 26. Respondents have also, therefore, indicated that the value that Commerce selected includes the types of paint used in wooden furniture production. Id. The court’s analysis for mirrors found that the selection of MSFTI as a data source for the valuation of mirrors was not supported by substantial evidence in the face of evidence that (1) companies do not source their mirrors internationally; (2) the MSFTI data selected was either not inclusive of the type of mirror used in wooden furniture production or was distorted by the inclusion of specialty mirrors when the factor input was plain mirrors. For paint, there is evidence on the record that the right type of paint is imported. Additionally, though Respondents have placed evidence on the record that the major paint importers as listed in Info-drive India are companies that “purchase or supply paints used by the automobile, computer, chemical and shipbuilding industries” and that “none of these importers have anything to do with furniture production,” Pis.’ Br. 26, they have not alleged or demonstrated that these paints are distor-tive of the final surrogate value. Therefore, Respondents have not alleged a sufficiently colorable claim that the surrogate value chosen by Commerce is either not inclusive of the factor input, or that it is overly distortive of the value of the factor input. Thus, Respondents’ argument becomes an allegation that Commerce’s choice for surrogate value is overinclusive and that instead Commerce should use, inter alia, Asian Paints (India) Dealers Price List. Issues & Decision Mem., P.R. Doc.1933 at 217-223 (Cmt.27). Respondents assert that this list is publicly available, tax-exclusive and contemporaneous with the POI. Additionally, Respondents contend that this list “describes the maximum retail prices to be charged by all Indian dealers, is representative of a large sample of domestic prices because Asian Paints is a major producer of paint products that are sold and distributed throughout India, [and] is not aberrational.... ” Id. at 218. Respondents contend that for at least Shing Mark, the Asian Paints price list includes products that are comparable to the reported factors of production, such as Asian Paints “Melamyne Sealer” and “Me-lamyne Glossy.” Id. Commerce rejected the use of the Asian Paints price list, stating that it does not “best represent the respondents’ paint inputs.” Id. at 222. Given that Shing Mark listed over 100 different paint inputs and Lacquer Craft listed over 230 paint inputs, Commerce found that a single product list from a single producer would not represent the “broad” multitude of factors employed by the various respondents. Id. Commerce also determined that “a single price list from a domestic Indian producer is not a representative sample of the domestic prices charged for the respondents’ finishing factors.” Id. at 223. Finally, after examining the differences in retail prices reported on the Asian Paints website, Commerce found that the Asian Paints price list was not indicative of the prices paid by end-users, but rather was a list of prices paid by dealers, and was therefore not indicative of the prices paid by furniture manufacturers. Id. Effectively, Commerce found that the prices from the Asian Paints price list were both un-derinclusive, and at the wrong level of trade to use for surrogate valuation. As stated in our analysis above, supra at 32-33, when Commerce is faced with a choice between two sub-optimal data sources, i.e., underinclusive vs. overinclu-sive data, Commerce’s choice between the two is reasonable and supported by substantial evidence. Chia Far Indus. Facto *1290 ry Co. v. United States, 28 CIT-,-, 343 F.Supp.2d 1344, 1352 (2004) Because Commerce’s choice of MSFTI data to value paint inputs 22 in light of its alternatives was a reasonable determination of the best available data, Commerce’s selection of MSFTI data is appropriate. 23 , 24 See Goldlink, 30 CIT at -, 431 F.Supp.2d at 1327. (“The Court’s role in the case at bar is not to evaluate whether the information Commerce used was the best available, but rather whether a reasonable mind could conclude that Commerce chose the best available information.”) (citation omitted). (c) Cardboard Respondents also challenge Commerce’s valuation of packing cardboard, which it assigned a value of $1.10/kg based on HTS[I] subheading 4808.9000 (“[pjaper and paperboard, corrugated (with or without glued flat surface sheets), creped, crinkled, embossed or perforated, in rolls or sheets, other than paper of the kind described in heading 4803 — other.”). Pis.’ Br. 27. Respondents assert that this valuation is too high, basing their assertion on data gathered from Tarun Vadehra and Highland House in India and Goldfindo in Indonesia. According to Respondents, the cardboard prices paid by these wooden furniture manufactures ranged from $0.31/kg to $0.65/kg. Id. Respondents object to the utilization of MSFTI data in this instance because these values are not specific to the type of cardboard used by the Respondents. Respondents suggest that Highland House and Tarun Vadehra provide a better, more specific valuation for packing materials, as the packing materials employed by these furniture producers are known in detail, as opposed to the broad, general MSFTI data. Id. at 29. Once again, Commerce was faced with a choice between two imprecise, not-perfect surrogate values. The MSFTI data runs the risk of being overbroad, in capturing more factor inputs than are actually utilized by the wooden furniture industry in India, while the data from the two Indian furniture producers runs the risk of being too narrow, such that the factor input in question is not actually captured by the proffered values. Though Respondents have asserted that the cardboard data is distortive, they have not produced evidence to indicate in what way the data is distortive. Issues & Decision Mem., P.R. Doc.1933 at 225 (Cmt.29). Respondents have not alleged that the industry does *1291 not use the inputs captured by MSFTI, nor that the valuation is distorted by the inclusion of more expensive specialized products. Respondents’ assertion in this instance is limited to pointing out how certain producers in India pay less for their cardboard inputs. Pis.’ Br. 27 & 36. Though one could conclude from this evidence that this lower price is reflective of the real experiences of all Indian producers, one could also conclude that these are specific prices for specific producers, reflective of their unique product needs, and thus not reflective of the experience of all Indian producers. Commerce examined and evaluated the data sources before it, and chose between two imperfect data sources. Commerce’s conclusion, that MSFTI represented the best available information in this instance, was reasonable and supported by substantial evidence. C. Wage rate One of the primary factors of production for any product is the cost of labor. Commerce treats the wage rate differently from all other factors of production; for labor, Commerce employs “regression-based wage rates reflective of the observed relationship between wages and national income in market economy countries,” 19 C.F.R. § 351.408(c)(3). Using this regression analysis, Commerce determines the relationship between countries’ per capita Gross National Product (“GNI”) and their wage rates; Commerce approximates the wage rate of the PRC by using the PRC’s GNI as the variable in the equation that was the result of the regression. See Wooden Bedroom Furniture from the People’s Republic of China: Final Results of Redetermination Pursuant to Court Remand Orders (Dep’t Commerce Aug. 1, 2005) (“Remand Determination”). 25 Stated mathematically: WagePE0 = Y + X * GNIPRC where Y is a constant 26 (as determined by the regression model), X is the coefficient (as determined by the regression model), GNIprc is the per capita GNP of the PRC, and WagePR0 is the derived wage rate for the PRC. Id. at 9. For wage rate data used to calculate the regression, because of “the practices of the respective data sources,” “[t]here is normally a two-year interval between the current year and the most recent reporting year of the data required for [Commerce’s] methodology.” Remand Determination at 4. Therefore, Commerce uses the “most recent reporting year” provided by each country and inflates those values, i.e., multiplies the values by the rate of inflation. Commerce calculates the wage rate regression once a year and uses that regression to calculate the wage rate for all investigations and administrative reviews in NMEs conducted during that year. For this investigation, Commerce calculated its regression after choosing a wage rate data set for fifty-four market economy countries, including the United States, as reported to the World Bank for 2001. See Remand Determination at 4-5 & 14. *1292 Respondents and AFMC challenge numerous aspects of Commerce’s calculation of the wage rate here. Broadly speaking, Respondents and AFMC allege three types of errors: (1) that Commerce’s method for calculating the wage rate is facially unlawful; (2) that even if the regulation is not facially unlawful, the manner in which Commerce implements its regulation is unlawful; and (3) that Commerce erred in its choice of data. 27 (1) Facial Challenge As noted above, the antidumping statute requires that Commerce’s “valuation of the factors of production shall be based on the best available information regarding the values of such factors in a market economy country or countries considered to be appropriate by the administering authority.” 19 U.S.C. § 1677b(c)(l) (emphases added). In defining the market economy countries to which Commerce may turn, Congress further requires that Commerce [S]hall utilize, to the extent possible, the prices or costs of factors of production in one or more market economy countries that are— (A) at a level of economic development comparable to that of the nonmarket economy country, and (B) significant producers of comparable merchandise. 19 U.S.C. § 1677b(c)(4) (emphasis added). Commerce’s regulation provides that “[f]or labor, [Commerce] will use regression-based wage rates reflective of the observed relationship between wages and national income in market economy countries.” 19 C.F.R. § 351.408(c)(3). Respondents argue that although the section 1677b(c)(4) permits Commerce to use data from only “comparable” market economies, Commerce’s regulation does not require such comparability in the selection of economies for the calculation of the wage rate. Respondents further state that Commerce intended to include the word “comparable” in its regulation, and unintentionally omitted that word from its final rule. Therefore, Respondents argue, Commerce’s regulation is void on its face. The court disagrees. Although Commerce’s regulation does not specifically provide that Commerce must choose comparable market economies, it does hot suggest the opposite either. Rather, the regulation is silent as to how Commerce will select market economies for its data set. As such, even if Respondents were correct that the anti-dumping statute permits use of data only from comparable market economies, Commerce could conceivably be faithful to both its regulation and Respondents’ interpretation of the antidumping statute by using data from only comparable market economies. Accordingly, Respondents fail to state a case for the facial invalidity of Commerce’s regulation. Cf. Reno v. Flores, 507 U.S. 292, 301, 113 S.Ct. 1439, 123 L.Ed.2d 1 (1993); INS v. Nat’l Center for Immigrants’ Rights, Inc., 502 U.S. 183, 188 & 194, 112 S.Ct. 551, 116 L.Ed.2d 546 (1991). (2) As Applied Invalidity Respondents next argue that Commerce’s selection of market economies for its regression in this instance violates the *1293 antidumping statute. Specifically, Respondents argue that most of the countries in the data set have a level of economic development far exceeding that of the PRC. For example, among the nations included within the data set is the United States (a nation with one of the highest GNPs). See Import Administration, Expected Wages of Selected Non-Market Economy Countries, Expected Wage Calculation: 2008 GNI Data, Regression Analysis: 2003 GNI Data, Revised November 2005, http://ia.ita. doc.go v/wages/03wages/110805-2003-Tables/03wages-110805.html (last visited August 10, 2006). Therefore, Respondents contend, Commerce’s regression methodology employs inappropriate countries for its calculation of the PRC’s wage rate. While it is true that Commerce’s regression analysis uses data from non-comparable countries, the result (in theory) derives what should be a comparable wage rate for the PRC. As Commerce has explained “[t]he regression that results provides a formula that, when applied to the non market economy country’s GNI, enables Commerce to determine in an accurate, fair and predictable manner, the labor wage rate of a market economy country at a comparable level of development.” Def.’s Resp. Parties’ Comments Remand Results Re: Commerce’s Labor Wage Rate 24 (“Def.’s Resp. Parties Cmts.”). It may be the case that there is a relationship between a country’s GNI and its wage rate and that accuracy would be greatly enhanced by using a broader data set of nations than just those at a comparable level of development to the PRC. Under such circumstances, using a broader data set may constitute the “best available information” and recourse to a broader range of market economy countries could be “appropriate” in advancing one of the antidumping statute’s purposes, i.e., to calculate the dumping margin as accurately as possible. This view of the “best available information” and “appropriate” market economy country or countries requirements is reflected in numerous court decisions. See, e.g., Shakeproof Assembly Components Div. of Ill. Tool Works, Inc. v. United States, 268 F.3d 1376, 1381 (Fed. Cir.2001) (“we have specifically held that Commerce may depart from surrogate values when there are other methods of determining the ‘best available information’ regarding the values of the factors of production.”) (citations omitted); Nation Ford Chem. Co. v. United States, 166 F.3d 1373, 1378 n. 5 (Fed.Cir.1999) (the antidumping duty statute “does not preclude consideration of pricing or costs beyond the surrogate country if necessary”); Lasko Metal Prods. Inc. v. United States, 43 F.3d 1442, 1446 (Fed.Cir.1994) (relying on the international market price for the factors of production); Globe Metallurgical, Inc. v. United States, 28 CIT -, -, 350 F.Supp.2d 1148, 1158 (2004) (“Under some circumstances, however, where the NME country has become a market economy, the post-NME values will best serve as representative of the values of the NME country as if it were a market economy.”). Indeed, the court in Nation Ford explicitly approved the notion of creating a “hypothetical” market economy to approximate the production experience of the PRC. Nation Ford, 166 F.3d at 1378. Consequently, Commerce’s calculation, at least in theory, produces a hypothetical wage rate for the PRC, which is therefore by definition a wage rate for a producer country at a comparable level of development, as required by 19 U.S.C. § 1677b(c)4. As such, so long as a reasonable mind can find that Commerce’s data choice in using the regression methodology constitutes the “best available information” in a particular case, nothing precludes Commerce from using sources outside a surrogate country to value the factors of production. Cf. Nation Ford, 166 F.3d at 1378 (the “best available information” test *1294 must be evaluated under the circumstances of the case). This, however, raises the main thrust of Respondents’ arguments here. Specifically, Respondents claim that (1) the manner in which Commerce created the regression model was arbitrary and therefore unsupported by substantial evidence; (2) the results of the regression model may be distorted; and (3) Commerce has failed to explain why it uses the PRC’s GNI but not its wage rate. The court considers each claim in turn. (a) Creation of the Regression Model Commerce selected the wage rate data for its regression from the Yearbook of Labour Statistics, published by the International Labour Organization (ILO). Remand Determination at 5. Commerce follows a three-step process in choosing which data it will use. First, Commerce establishes a minimum standard for data. Under this standard, Commerce will only use data if that data is (i) less than five years old; (ii) reports wages for both male and female employees; and (iii) covers “different types of industries].” Remand Determination at 6. Next, because the ILO database includes multiple expressions of wage rates, Commerce must choose which data meeting its minimum threshold it will use for a particular country. Commerce prioritizes data using the following criteria (in this order of precedence): (i) “ ‘[w]orker [c]overage,’ ie., coverage of different types of workers, such as wage earners or salaried employees”; (ii) “ ‘[t]ype of [d]ata,’ ie., the unit of time for which the wage is reported, such as per hour or per month”; and (iii) “ ‘[s]ource ID,’ ie., a code for the source of the data.” Id. Within each factor, Commerce has preferences. For example, Commerce “generally prioritizes ‘wage earners,’ ‘employees’ and ‘total employment,’ in that order for the parameter ‘Worker Coverage.’ ” Id. at 7. “Finally, it is the Department’s normal practice to eliminate aberrational values (ie., values that vary in either direction in the extreme from year to year) from the wage rate dataset” Id. Using these criteria, Commerce used the wage rate data of fifty-four countries. 28 Here, Commerce has acknowledged (a) the desirability of a broader data set in its own justification for the creation and utilization of a regression model for wage rates, see Def.’s Br. 21 (“Due to the variability of wage rates in countries with similar per capita GNI, a more accurate result would be obtained by utilizing data from multiple countries.”) (citations omitted); id. at 5 & 24 (“A relatively broad data set helps to prevent bias and ensure that the regression is statistically sound.”); Rules and Regulations 19 CFR Parts 351, 353 and 355, Antidumping Duties; Countervailing Duties, 62 Fed. Reg. 27,296 at 27,367 (May 19, 1997) (“Preamble 19 CFR § 351 ”) (“averaging of multiple data points ... should lead to more accurate results ....”); (b) that additional countries for which it had available data (may have) met its selection criteria, Final Determination, 69 Fed.Reg. 67,313, 67,317; Issues & Decision Mem., P.R. Doc.1933 at 180 (Cmt.23) 29 ; and (c) *1295 use of these additional countries would yield a better result than a regression model without such countries, id. Under such circumstances, Commerce’s exclusion of the countries that met its own selection criteria was arbitrary and therefore unsupported by substantial evidence. Accordingly, Commerce’s regression model cannot constitute the “best available information.” Indeed, even if the statute did not require the use of “best available information” an arbitrary application of selection criteria is inherently unreasonable and, therefore, renders the conclusion’unsupported by substantial evidence. Cf. Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Fam Mut. Auto. Ins. Co., 463 U.S. 29, 48, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). Nor can Commerce hide behind its own regulation — Commerce’s regulation, when interpreted, must yield a reasonable interpretation of the statute. Arbitrariness is inherently unreasonable. Cf. Allied Pac. Food (Dalian) Co. v. United States, 30 CIT -,-, 435 F.Supp.2d 1295, 1320 (2006) (finding non-adherence to selection criteria to be reversible error). Unsurprisingly then, Commerce does not contest this premise. Instead, Commerce stated that it would not increase the data set here because: (i) such an undertaking would require a significant change in the data set that should be subject to comment from the general public; and (ii) this would require more time than was available in the investigation in order to “determine an accurate construction of a new dataset and to conduct a new regression analysis”. Issues & Decision Mem., P.R. Doc.1933 at 180 (Cmt.23); Remand Determination at 22; see also Def.’s Resp. Parties Cmts. 28. The court will consider each rationale. (i) Notice and Comment Rulemaking Commerce’s first argument, ie., that the data set in question must be developed through notice-and-comment rulemaking, appears to be inconsistent with Commerce’s past practice. Commerce has in the past updated and expanded the number of countries within the data set without resorting to notice and comment rule-making. In fact, during the investigation here, Commerce used a basket of fifty-six countries, but during the voluntary remand, used a basket of only fifty-four. Remand Detemination at 13. No notice- and-comment rulemaking was used to effect the change. Commerce has also, over time, expanded its data set of countries from forty-five countries to fifty-six countries without vetting its choices through notice-and-comment rulemaking. During the notice-and-comment period for this regulation, Commerce foresaw utilizing forty-five countries for the regression analysis. See Proposed Rules 19 CFR Parts 351, 353, and 355 Antidumping Duties; Countervailing Duties, 61 Fed.Reg. 7308, 7345 (Dep’t Commerce Feb. 27, 1996) (“Proposed Rule ”) (noting that the methodology would employ approximately forty-five countries); Import Administration, Expected Wages of Selected Nonr-Market Economy Countries, http://ia.ita.doc.gov/ wages (last visited October 5, 2006) (“For previous updates, we utilized the wage and gross domestic product (GDP) data of at least 45 market economies collected from publicly available sources such as the International Labour Organization and the World Bank/International Monetary Fund.”). By the year 2002, Commerce had increased this data set to fifty-six countries. Expected Wages of Selected Non-Market Economy Countries, Corrected 2000 data Revised September 2002, http://ia.ita.doc.gov/wages/corrected00 *1296 wages/correctedOOwages.htm. Commerce has even changed the underlying data it uses in its methodology without notiee- and-comment rulemaking. See Expected Wages of Selected Non-Market Economy Countries, http://ia.ita.doc.gov/wages (“However, beginning with last year’s update (May 2000), we chose to use per capita gross national product (GNP) instead of GDP.”). Consequently, under the regulation, Commerce has expanded and contracted its data set at will, as well as altered the underlying data. Cf. Qwest Corp. v. FCC, 258 F.3d 1191, 1206 (10th Cir.2001) (“the FCC is not required to begin a new notice-and-comment period every time it fixes a technical bug in its computer program.”). Moreover, it is fundamental administrative law that when a regulation is unlawful, or a finding of fact unsupported by substantial evidence, Commerce must change its practice or conclusion. See, e.g., Allegheny Ludlum Corp. v. United States, 367 F.3d 1339 (Fed.Cir.2004) (noting that Commerce employed a different test than that provided for in the regulation found unlawful by the Court of International Trade). Contrary to Commerce’s reasoning, it may not continue to unlawfully apply a regulation or finding of fact to a party. Accordingly, this defense must be rejected. (ii) Deadlines Commerce’s second argument is that, given its statutory deadlines for completing investigations, Commerce could not consider available information in updating its regression model. Congress was certainly sensitive to this concern by limiting Commerce’s choice of data to that “available” during the investigation. But in recognizing this concern, Congress nonetheless required that if information was available, i.e., placed on the record, Commerce was compelled to consider it. Therefore, Commerce’s defense runs directly against its statutory duty. Consequently, Commerce’s second defense must also be rejected. The difficulty here is more pronounced than it might be in another context. Commerce’s problem here is entirely self-inflicted — it chose such an “extremely complex” methodology. Remand Determination at 2. When adopting such an “extremely complex” methodology, Commerce should have been aware that correcting it would also be complex. Therefore, Commerce cannot reasonably defend its actions here by invoking problems of its own creation. (b) Distortion of Regression Model Commerce’s regression model also appears to produce distorted results. Consider, as an example, the prediction of India’s labor wage rate. In its Remand Determination, Commerce calculated the constant to be 0.392, the coefficient to be 0.00048, and India’s 2002 GNI (expressed in U.S. dollars) to be $470. See Exhibit III to Remand Determination at pp. 2, 3, 5 & 6. Commerce’s model therefore predicts India’s wage rate to be' $0.6176 per hour. 30 However, the labor wage Commerce used for India was $0.21 per hour— approximately one-third of what the model predicts. Id. at 3. Therefore, assuming that the Indian labor wage rate is accurate and representative, it appears that the regression contains a bias or distortion. Given that India was the country Commerce found most comparable to the PRC with respect to all other factors of production, it seems likely then that a similar distortion or inconsistency could result in computing the wage rate of the PRC. At the least, there appears to be an inconsistency or distortion in Commerce’s figures, a distortion Commerce does not explain. *1297 This possible distortion also appears when one considers the effect of the constant, ie., 0.392. If a country had no GNI, ie, GNI = 0, then the model would predict a wage rate of $ 0.392. 31 Thus, Commerce’s regression model appears to overstate wage rates of low-income countries. This would appear problematical here where Commerce is attempting to value labor in such a low-income country, ie., the PRC. While Commerce’s model may be the best information available on the record, here Commerce has failed to give a viable explanation for its choice in light of this possible distortion of its predicted wage rate of countries such as the PRC. Rather, despite these possible anomalies, Commerce’s only apparent justification is that use of the regression model leads to an accurate, fair, and predictable computation of the wage rate. See Remand Determination at 22. In the face of what appears to be flaws in Commerce’s figures, the first observation, ie., accuracy, is no more than a conclusory statement. Nor does the court find that the claims of fairness and predictability are supportable. In adopting its regulation Commerce commented that “[i]t also is fairer [to use the regression method], because the valuation of labor will not vary depending on which country [Commerce] selects as the economically comparable surrogate economy.” Preamble 19 CFR § 351, 62 Fed.Reg. at 27,367. At the same time, in this case, Commerce used a wage rate of $0.85 which appears to be higher than each of the possible surrogate countries. Commerce should examine and balance these competing claims to “fairness.” As for predictability, first, because the computation of normal value for merchandise from an NME is based on many different factors calculated during an investigation, the added predictability of a wage rate (if there is any) has limited utility— assuredly not enough to permit an inaccurate computation. Second, Commerce recalculates the wage rate annually. It is therefore unclear to what degree the use of the regression here guarantees predictability. Certainly, the data set Commerce used here was not available when the Respondents were importing the merchandise under investigation — therefore, they gained no benefit from the predictability (if any) of its use. Third, if the regression model is inaccurate, Commerce’s use thereof could simply be adding a tariff to Respondent’s merchandise without there existing a violation of the antidumping statute. In other words, predictable or not, Commerce’s antidumping authority would be of uncertain legitimacy. Such a method would also be contrary to congressional intent. See S.Rep. No. 100-71 at 106 (1987) (quoted above). While the court will give more deference to longstanding agency practices, this deference is neither automatic nor unlimited. See, e.g., AK Steel Corp. v. United States, 226 F.3d 1361 (Fed.Cir.2000). *1298 Therefore, on this record, and lacking sufficient justification from Commerce for its choices made, the court cannot conclude that Commerce’s selected calculation of its regression model is reasonable. 32 Cf. Qwest, 258 F.3d at 1206 (plaintiff “has not presented any evidence that the model overall produces such inaccurate results that it cannot form the basis of rational decision-making.”) (emphasis removed). Additionally, Commerce uses the GNI of the PRC in its wage rate calculation while, at the same time, declining to use wage rate data from the PRC. See Remand Determination at 9. Respondents object to the use of the GNI from the PRC in calculation of the wage rate, pointing out that the entire reason behind using surrogate value data is that Commerce has determined that the PRC is an NME, and, as such, Commerce does not trust the wage and price data that emerges from the PRC to be truly reflective of market-forces. The GNI, representing the per capita income of a country, is in part based on the wages in a given country, as wages affect income. Commerce, relying on its regulation, failed to address the substantive point raised in this argument. To be sure, despite the presumption that data from an NME is unreliable, Commerce may be justified in finding some data from the PRC reliable while finding other data from the PRC unreliable, e.g., Commerce may conclude that the PRC’s GNI data is reliable whereas the PRC’s wage rate data is not. However, at the very least, Commerce must justify why it made such a finding. See, e.g., Allied Pac. Food, 30 CIT at -, 435 F.Supp.2d at 1320. Therefore, this matter is also remanded for an explanation as to why Commerce finds the PRC’s GNI data to be sufficiently reliable to utilize in a regression analysis. * X X * * * The court expresses no opinion on whether Commerce’s regression methodology is salvageable. 33 The court notes, however, that this will be Commerce’s third try to rectify problems with its methodology in this case. 34 *1299 Upon consideration on remand, Commerce must explain why it is using the data set it employs in face of the objections noted above and address any apparent statistical anomalies. If Commerce finds it cannot offer such explanation, Commerce shall use the “best available information” as required by the antidump-ing statute. (3) Proper Data Set AFMC also disagrees with Commerce’s approach for a different reason. During the proceedings on remand, Commerce admitted that “[a]fter extracting the selection of datapoints from the larger dataset, Commerce} did not retain the full underlying ILO dataset” it had used during the investigation. Remand Determination at 17. Therefore, because the original data set was lost, Commerce used a data set compiled a month after the publication of the Final Determination in the Remand Determination. Id. at 10-12, 17. Commerce did note that the data set it used was “drawn from the same ILO database that existed during [Commerce’s] investigation.” Id. at 17. However, Commerce also noted that “while [Commerce] would prefer to use a 2002 wage rate data set that was extracted in October 2004, such a data set is not available.” Id. Commerce further commented that “parties provided [no] information to suggest that there are any material differences between the data-sets.” Id. AFMC argues that it was improper for Commerce to resort to a data set compiled after the investigation. In supporting its arguments, AFMC points to what it claims are “material” differences (presumably outcome determinative differences) between the data sets. AFMC further questions when the data was extracted, claiming that the data upon which Commerce relied was not “available” during the investigation; it avers that because Commerce did not retain the original data set, Commerce cannot claim that the information was actually available during the pendency of the investigation. Much of AFMC’s legal premise is correct. First, when Commerce uses data, it must include that data as part of the record. 19 U.S.C. § 1516a (b)(2); see also 19 U.S.C. § 1677f (a)(3); 19 C.F.R. §§ 351.104(a)(1). Commerce must, in turn, file this record with the clerk of this Court within forty days of the service of the complaint filed under 19 U.S.C. § 1516a. See 28 U.S.C. § 2635(b); USCIT R. 73.2. Moreover, just as “substantial evidence review (on the record) would not be a meaningful exercise if the ‘evidence’ that comprised the record was obtained through an arbitrary procedure,” Decca Hospitality Furnishings, LLC v. United States, 29 CIT-, ---, 391 F.Supp.2d 1298, 1305 (2005), substantial evidence review would not be meaningful if Commerce fails to turn over the relevant evidence to the court to review. Even absent this requirement, however, Commerce additionally erred in failing to retain information relevant to ongoing litigation; such an error may render the court less able to perform its function and result in a miscarriage of justice. AFMC is also right in claiming that Commerce may only resort to the “best available information.” See 19 U.S.C. § 1677b(c)(l). Although the statute does not define the time period in which “availability” is measured, given that administrative law defines “available” in terms of the underlying investigation, “available” must mean “available during the investigation.” See, e.g., Vt. Yankee Nuclear Power Corp. v. Natural Resources Def. Council, Inc., 435 U.S. 519, 555, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978) (“[T]he role of a court in reviewing the sufficiency of an agency’s consideration of environmental factors is a limited one, limited both by the time at *1300 which the decision was made and by the statute mandating review.”); see also Co-Steel Raritan, Inc. v. ITC, 357 F.3d 1294, 1316-1317 (Fed.Cir.2004); cf. 19 U.S.C. § 1677f(a)(3); S. REP. NO. 96-249 at 247-48 (1979), reprinted in 1979 U.S.C.C.A.N. 381, 633 (“Judicial review of determinations subject to the provisions of subsection (a)(1) would proceed upon the basis of information before the relevant decision-maker at the time the decision was rendered including any information that has been compiled as part of the formal record.”). However, these legal principles notwithstanding, 19 U.S.C. § 1677b(c)(1) only applies to the information itself, ie., the source data, apart from any manipulations or analysis of that data, ie., the actual data set. After all, Commerce routinely compiles data sets to value factors of production; if “information” included these manipulations then, after a determination issues, Commerce could never go back and fix errors in its analysis without creating new “information” unavailable during the investigation. Cf. infra at 100 n. 36. Consequently, to allege that Commerce has relied on data unavailable during the investigation, a party must allege more than that the data set as used was unavailable; it must allege that the underlying data which was then used in the data set was unavailable during the investigation. With this foundation in place, AFMC fails to sufficiently allege reversible error. First, it appears that AFMC’s actual claim is that the data was not extracted rather than the data set was unavailable. See, e.g., AFMC’s Resp. Br. Final Result of Redetermination Re: Wage Rate Calculations 4. Simply because data was not “extracted” does not mean that it was unavailable. Therefore, Commerce’s isolated statement in its Remand Determination (and others in Commerce’s brief) cannot support AFMC’s argument. AFMC’s argument also contradicts Commerce’s comment that the information was unavailable during the investigation. Remand Determination at 17. Although Commerce’s inability to present the data set for examination may make it more difficult for parties to contest Commerce’s assertion, all the data upon which Commerce relied is publicly available. 35 As such, AFMC could still have presented evidence before the closure of the investigation. Given these considerations, the court rejects AFMC’s argument. D. Financial Ratios Because firms have “general expenses and profits” not traceable to a specific product, in order to capture these expenses and profits, Commerce must factor (1) factory overhead (“overhead”), (2) selling, general and administrative expenses (“SG & A”), and (3) profit into the calculation of normal value. 19 U.S.C. § 1677b (c)(1); see also Final Determination, 69 Fed.Reg. at 35, 327; Hebei II, 29 CIT at -, 366 F.Supp.2d at 1277 n. 7; Shanghai Foreign Trade Enters. Co. v. United States, 28 CIT-,-, 318 F.Supp.2d 1339, 1341 (2004); Peer Bearing Co. v. United States, 25 CIT 1199, 1214-15, 182 F.Supp.2d 1285, 1303-04 (2001); cf. 19 U.S.C. §§ 1677b (c)(4), 1677b (b)(3)(B), 1677b (e). As with its calculation of the other factors of production, Commerce uses surrogate values to determine an importer’s financial ratios. In this instance, Commerce uses financial statements from *1301 one or more surrogate company/companies to calculate comparable ratios. 36 In choosing financial statements, Commerce “normally will use nonproprietary information gathered from producers of identical or comparable merchandise in the surrogate country.” 19 C.F.R. § 351.408(c)(4). In choosing from financial statements which meet this criterion, Commerce generally considers the quality, specificity, and contemporaneity of the available financial statements. See Issues and Decision Memorandum at Cmt. 5 accompanying Fresh Garlic From the People’s Republic of China, 67 Fed.Reg. 72,-139 (Dep’t Commerce Dec. 4, 2002) (final results of antidumping duty new shipper review) available at http://ia.ita.doc.gov/ frn/summary/prc/02-30771-l.pdf (“Fresh Garlic”). Commerce may also consider the “representativeness of the production experience of the surrogate producers in relation to the respondent’s own experience!;,]” Issues and Decision Memorandum at Cmt. 9, accompanying Persulfates from the People’s Republic of China, 68 Fed. Reg. 6712 (Dep’t Commerce Feb. 10, 2003) (final results of antidumping duty administrative review) available at http://ia.ita.doc. gov/frn/summary/prc/03-3285-l.pdf. We note again that Commerce must apply its selection criterion in a consistent and uniform manner, otherwise its selection could become arbitrary and capricious. In order to derive “a more accurate portrayal of the economic spectrum,” Commerce may, of course, select more than one surrogate company from which to draw data. Issues and Decision Memorandum at Cmt. 6 accompanying “Fresh Garlic ”, 67 Fed.Reg. 72,139. When this is the case, Commerce usually calculates the sim- *1302 pie average of the selected companies’ financial ratios. Rhodia, Inc. v. United States, 25 CIT 1278, 1285, 185 F.Supp.2d 1343, 1350 (2001); see also Timken v. United States, 26 CIT 434, 466-67, 201 F.Supp.2d 1316, 1346 (2002). Nevertheless, although using multiple financial statements is permissible, as will be discussed below, Commerce is not justified in sacrificing quality for quantity; put differently, Commerce must justify why the data set it chooses is appropriate. Otherwise, Commerce’s decision would not be faithful to its own stated criteria. In the case at bar, the parties submitted financial statements of eighteen companies. Commerce winnowed this list to nine through the course of its investigation. See, e.g., Financial Ratio Memo, P.R. Doc.1931 at Attachs. 1 & 11. Commerce rejected four financial statements because they were not contemporaneous with the period of review. Id. at Attach. 11. Commerce rejected another company’s financial statement because the company was not a producer of the subject merchandise. Id. The parties do not dispute the rejection of these financial statements. However, the parties make numerous other challenges to Commerce’s choices of financial statements and calculation of the surrogate financial ratios: (1) Respondents contend that Commerce improperly rejected the 2003/2004 financial statement of Indian Furniture Products, Inc. (“IFP”); (2) AFMC challenges the inclusion of a financial statement from Jayaraja Furniture (“Jayaraja”); (3) AFMC challenges the inclusion of a financial statement from Evergreen International Ltd. (“Evergreen”); (4) Respondents challenge the inclusion of financial statements from Swaran Furnitures Ltd. (“Swaran”), Niza-muddin Furniture Private Ltd. (“Niza-muddin”), Fusion Design Private Ltd. (“Fusion Design”), and D’nD’s Fine Furniture Pvt., Ltd. (“DnD”); and (5) Respondents challenge Commerce’s rejection of financial statements from three Indonesian companies, Goldofindo, CIPTA, and SIMA (collectively “the Indonesian companies”). 37 Because of the relationship between these five issues, the court will address each in turn. (1) IFP During the investigation, Respondents submitted a 2003/2004 financial statement from IFP, an Indian producer of subject merchandise. Although acknowledging that the IFP financial statement was more contemporaneous with the period of review than the other Indian financial statements used (including one from IFP from fiscal year 2002/2003), Commerce “excluded [IFP’s 2003/2004 financial statement] because it showed no profit for its 2003/2004 fiscal year and [Commerce] had a wealth of financial statements from the previous fiscal year on which to rely.” Issues & Decision Mem., P.R. Doc.1933 at 68 (Cmt.3). Without a stated profit, Commerce is unable to calculate the profit ratio, which is an important factor in their comparison. *1303 Respondents challenge this finding. Despite the fact that IFP’s financial statement showed a loss, Respondents contend that this loss is illusory. Pis.’ Br. 31, 50-51. When closely read, Respondents attest, the financial statement reveals that IFP pays a significantly high interest rate on a loan to its parent. Id. This high rate of interest, Respondents allege, represents profit that IFP diverts to its parent to avoid tax liabilities. Id. Therefore, Respondents assert, IFP is profitable and Commerce’s rejection of the statement was erroneous. Id. Respondents claim that because IFP’s 2003/2004 financial statement demonstrates that IFP is profitable, and given that this statement is the more contemporaneous with the period of review, Commerce should have used (perhaps exclusively) the 2003/2004 IFP financial statement. In the Final Determination, Commerce rejected this argument averring “that the interest owed to IFP’s parent company was just a method for the parent company to extract profits without incurring tax liability, is speculation and does not change the fact that the inter-company interest expense is an expense nonetheless.” Issues & Decision Mem., P.R. Doc.1933 at 68-69 (Cmt.3). The court finds reasonable Commerce’s conclusion that Respondents’ theory is speculative. Respondents provide no evidence to support their profit diversion theory and offer no reason why their theory is the only possible explanation for the allegedly high interest rate. Moreover, even if Respondents’ theory is correct, it was reasonable for Commerce not to prefer (and therefore exclude) a financial statement where one of the accounting maneuvers was, even under Respondents’ account, misleading. Furthermore, Respondents’ allegation would require the court to determine the “real” interest rate as well as IFP’s “real” profit. Because any derivation of the “real” interest rate IFP is paying to its parent, if any, would, on this record, be entirely speculative, deriving IFP’s profit ratio and SG & A might be highly problematic if Commerce were to include IFP’s 2003/2004 financial statement in its calculation. Cf. Financial Ratio Memo, P.R. Doc.1931 at Attachs. 3-6, 10 (including interest payments within SG & A). Because of these complications, it was reasonable for Commerce to conclude that IFP’s earlier financial statements provided more accurate information. (2) Jayaraja AFMC challenges Commerce’s use of Jayaraja’s financial statement. In particular, AFMC alleges that despite “Commerce’s well-established practice to reject” financial statements that contain “no notes” and “no auditor’s statement,” Commerce used Jayaraja’s financial statement which had such deficiencies. AFMC’s Br. 23. In addition, AFMC avers that “Jaya-raja’s financial statements are inappropriate because they report zero depreciation in the profit and loss statement,” id. (emphasis in original), which “renders the Jay-araja statement a significant outlier and a cause of distortion when aggregated with the other financials.” AFMC’s Reply Def. & Def.-Intervenors’ Resps. Opposition AFMC’s Mot. J. Agency R. Re: Selection Surrogate Values & Calculation Financial Ratios 8 (“AFMC Reply Br.”). In its brief, Commerce asserts that Jay-araja’s financial statement was approved by an auditor and was sufficiently detailed such that no auditor’s statement or notes were necessary; in addition, insofar as Commerce is expected to vet financial statements with the same consistency, Commerce further notes that one of the financial statements AFMC proffered, and which Commerce adopted, also lacked such imprimaturs. Def.’s Br. 85. In response *1304 to AFMC’s alternative argument, Commerce argues that general accounting principles require consideration of depreciation. Accounting rules do not require that depreciation be separately or specifically listed — it can be factored into other values. Id. Because the statement was approved by an auditor, Commerce reasons, depreciation must be stated as part of other values. Id. Despite the fact that Commerce outlined AFMC’s arguments in the Issues & Decision Mem. accompanying its Final Determination, Issues & Decision Mem., P.R. Doc.1933 at 48 (Cmt.3), the Final Determination does not directly or indirectly refute these arguments, id. at 67-73. As such, Commerce’s explanations and arguments offered in its brief here are entirely post-hoc rationalizations. Although Commerce’s argument may be compelling, the agency must adopt this position on the agency record if the court is to affirm it here. See, e.g., SKF USA Inc. v. United States, 254 F.3d 1022, 1028 (Fed.Cir.2001). Accordingly, this question is remanded to Commerce. (3) Evergreen AFMC also argued before Commerce that Evergreen’s financial statement should not be included in Commerce’s data set. Evergreen, AFMC contends, is “a significant producer of leather garments as well [as] a manufacturer of furniture.” AFMC Br. 24. Because leather goods are not comparable to the subject merchandise, AFMC avers, including Evergreen’s financial statement was improper. AFMC claims Evergreen’s financial ratios reflect values not attributable to its furniture manufacturing operations and are therefore inappropriate surrogates. In the Final Determination, Commerce conceded that Evergreen produces leather products, but found that it could disaggre-gate leather production from furniture production expenses and/or neutralize any spillover. Specifically, Commerce found that “Evergreen outsources almost the entire production of its leather goods, and as such there is no installed capacity. Accordingly, in our calculation of Evergreen’s factory overhead ratio, we have excluded identifiable manufacturing expenses related to the production of leather goods from the MLE denominator.” Financial Ratio Memo, P.R. Doc.1931 at 2 (citations omitted). Commerce conceded that Evergreen’s leather production was more problematic in the calculation of the other ratios. In particular, Commerce noted, “Evergreen’s SG & A and profit relate to both leather and furniture goods.” Id. To remedy this problem, Commerce included leather related expenses in the denominator of the SG & A and profit ratios. Id. Averring that Commerce’s remedy was insufficient, AFMC argues that “Commerce improperly applied the SG & A ratio for the entire company to the furniture division, even though it recognized that the furniture division was a manufacturing operation, and the leather goods division was an outsourcing operation.” AFMC Br. 25. This, it claims, distorts Evergreen’s SG & A and profit ratios. Id. The court agrees for slightly different reasons. Missing from Commerce’s determination is an explanation as to why the inclusion of Evergreen’s financial statement, in spite of the complication identified by Commerce, adds to the accuracy of its calculation of the surrogate ratios. Particularly problematic is the fact that other financial statements, without such problems, exist. Under such circumstances, Commerce must justify its decision to include statements which it admits are of questionable reliability and thereby unlikely to constitute the best available information. *1305 Nor does the court find convincing Commerce’s argument espoused in its briefs. In its briefs, Commerce argues that Evergreen is a large producer of wooden bedroom furniture. Commerce further contends that AMFC “has not shown that the calculation of SG & A and profit results in a distortion to the surrogate ratios, nor does it cite to any record evidence indicating that another allocation would have been more accurate.” Def.’s Br. 88-89. Rather than justifying their position or providing additional rationales for the inclusion of Evergreen in their calculation, Commerce attempts to shift the burden to AFMC to prove why Commerce’s mathematical manipulation does not cure the facial deficiency of Evergreen’s data, even though there are other reliable financial statements from which Commerce may choose. This is not an appropriate basis upon which to approve Commerce’s selection. (4) Swaran, Nizamuddin, Fusion Design, and DnD Respondents contend that the inclusion of Swaran, Nizamuddin, Fusion Design, and DnD’s financial statements was improper. Pis.’ Br. at 48-50. Respondents claim that the inclusion of these financial statements was improper because the firms from which the financial statements came have different production experiences than their own. Id. More specifically, Respondents allege that: i. “There is no record evidence that Swaran produces wood bedroom furniture as it lists no production equipment among its assets. It therefore must either subcontract its production or produce furniture by hand.” Pis.’ Br. 29 (citations omitted). ii. “Nizamuddin characterizes itself as a ‘handicraft’ producer specializing in ‘carving pearl and wood inlay.’ In addition evidence was submitted that its showroom and workshop combined was no more than 600 square feet and employed about 4 or ' 5 people.” Id. (citations omitted). iii. “Fusion Design is as much a design shop as a manufacturer, creating custom furniture for end-users such as the British High Commission, British Airways, U.S. 'AID, and STARTV.” Id. at 30. iv. “DnD is also a design shop and it is not clear at all that it is even a furniture design shop given its discussion in the technology absorption section of its financial statement regarding efforts ‘being made to increase the shelf life of the product and increase its nutritious value.’ ” Id. at 30 (citations omitted). In response to the argument that the size of these firms led to unrepresentative financial ratios, Commerce acknowledged: The fact that the Indian surrogate has a smaller production capacity than the Chinese respondents “does not lead to the automatic conclusion that its overhead rate is different”, but simply that it may incur less overhead (in the numerator) and consume fewer raw materials (in the denominator). Issues & Decision Mem., P.R. Doc.1933 at 69-70 (Cmt.3) (quoting Persulfates from the People’s Republic of China, 68 Fed. Reg. 6712). In refuting the idea that any such distortion would occur here, Commerce found that: Jayaraja and Akriti [companies included within the data set that Respondents did not directly contest] both experienced production and sales volumes in the range of those experienced by Nizamud-din, Fusion [Design], Swaran, and DnD. Further, the SG & A expenses of Jaya-raja and Akriti demonstrate that small production and sales volumes do not automatically precipitate high SG & A expenses. *1306 Issues & Decision Mem., P.R. Doc.1933 at 70 (Cmt.3). Commerce appears to concede that these firms are facially less representative of Respondents’ production experience. Nevertheless, Commerce claims that this facial distinction is of no moment. In weighing the arguments on this issue, the court notes that Respondents are certainly correct in claiming that a firm’s size may affect certain of its financial ratios— after all, that is why economies of scale are beneficial in certain settings. Indeed, as is recorded in the legislative history of the antidumping statute, “Commerce should seek to use, if possible, data based on production of the same general class or kind of merchandise using similar levels of technology and at similar levels of volume as the producers subject to investigation.” Conf. Rep. at 591(emphasis added). This concern appears relevant here. As reflected in the Amended Final Determination, Nizamuddin, Fusion Design, Swar-an, and DnD had an average SG & A ratio of 32.22% compared with the 14.37% average SG & A rate of the other allegedly similar five companies included in the data set. 38 Commerce does not address these averages. Rather, Commerce responds by pointing to the SG & A ratios of Jayaraja and Akriti, which it claims demonstrate “that small production and sales volumes do not automatically precipitate high SG & A expenses.” Issues & Decision Mem., P.R. Doc.1933 at 70 (Cmt.3). While this may be true, it does not demonstrate the opposite conclusion: that the size of Niza-muddin, Fusion Design, Swaran, and DnD are irrelevant. 39 Commerce’s answer is simply nonresponsive to Respondents’ challenge. Also lacking from Commerce’s analysis is any discussion of why it chose to include the four financial statements at all. In particular, (a) given the facial distinctions, (b) that general principles of economics and intuition would suggest that such a distinction has relevance, (c) evidence to suggest there might be a relationship between production experience and overhead, and (d) financial statements of firms which (appear) to have similar manufacturing experiences to the Respondents, Commerce has an obligation to explain why it included these financial statements. Nor is the court convinced by AFMC’s argument that, in essence, the comparability between these firms and Respondents is close enough. “Comparability” is an *1307 elastic concept that Commerce has latitude to define on a case-by-case basis. Nevertheless, this latitude may vary in relation to the information available. It may be the case that where Commerce is left with the choice of two undesirable (in the absolute sense) options, the comparability of the surrogate firms to the importers need not be great. For example, if Commerce is investigating allegations of dumping cast iron pipe fittings and is left with choosing between the financial statement of a surrogate company producing cast iron brake rotors or “nonspecific information compiled by the Reserve Bank of India,” relying on a company producing cast iron brake rotors may suffice. See Shanghai Foreign Trade Enters. Co., 28 CIT at-, 318 F.Supp.2d at 1342, 1348-49; cf. Yantai Oriental Juice Co. v. United States, 26 CIT 605, 619-20, 2002 WL 1347018 (2002). However, where, as here, there are (perhaps) other surrogate companies which better approximate the manufacturing experience of the importers’ businesses, the comparability test may require limiting the data set to those surrogate companies which reasonably approximate the importers’ manufacturing experience — otherwise, Commerce’s choices may no longer be faithful to its statutory mandate. Put differently, without a convincing explanation, when Commerce has other reliable information available, there is no basis for tainting good data with bad by adopting an expansive definition of “comparability.” With that said, the court expresses no opinion on which company or companies reasonably approximate the Respondents’ production experiences. Commerce is free on remand to find that these financial statements are as reflective of the Respondents’ manufacturing experiences as the other financial statements upon which it relies. In justifying its conclusion either way, Commerce cannot take the inferential leap it took here without addressing contrary evidence. The court further notes again that Commerce must uniformly apply whatever criterion it ultimately adopts. (5) Indonesian Firms Respondents placed on the record financial statements of three Indonesian companies. Commerce rejected these statements because “the record contained a wealth of Indian financial statements” and therefore it “had no reason to look” outside the surrogate country' for financial statements. Issues & Decision Mem., P.R. Doc.1933 at 68 (Cmt.3). This conclusion is consistent with Commerce’s explanation of its regulation, 19 C.F.R. § 351.408(c)(2), which provides (subject to certain exceptions) that Commerce will “value inputs using publicly available information regarding prices in a single surrogate country.” Rules and Regulations 19 CFR Parts 351, 353 and 355, Antidumping Duties; Countervailing Duties, 62 Fed.Reg. 27,296 at 27,366 (May 19, 1997) (emphasis added). Nevertheless, Respondents contend that “[njumerous U.S. importers and retailers provided affidavits demonstrating that Indian furniture producers are not comparable to Chinese or Indonesian furniture producers.” Pis.’ Br. 52. In contrast to the data from Indian companies, Respondents contend that one Indonesian producer, Goldfindo, produces identical merchandise and has a production process “on a scale and with technology that, is very much like Plaintiffs.” Id. at 53.' Consequently, Respondents contend that the data from Indonesia, and specifically Goldfindo, represents the best available information. The court disagrees. Here, based on Commerce’s regulations, Commerce may reasonably rely, even after remand, on financial statements from Indian surrogate companies. Moreover, as Commerce has reasonably interpreted its *1308 statutory mandate, Commerce has full discretion not to use the data from Indonesian companies. Accordingly, Commerce’s determination is sustained. II. VALUING SPECIFIC FACTORS OF PRODUCTION Having addressed Commerce’s selection of certain data sets, the court now turns to how Commerce chose to value specific factors of production. As noted in the introduction, Commerce used Indian import statistics as stated in a database referred to as the MSFTI as a basis for estimating most factor values. Factors Valuation Mem., P.R. Doc. 1329, at 4. MSFTI, in turn, lists pricing information for goods by their tariff classifications. Therefore, to value a factor of production, Commerce must first classify a factor of production under the Indian tariff schedule (the Harmonized Tariff Schedule of India (“HTS[I]”)), and then determine the price listed in MSFTI for goods classified under the tariff provision. 40 It is well-established under United States law that an administering authority can make two types of errors when classifying a product. First an administering authority may misread the tariff schedule, i.e., erroneously interpret it and incorrectly include a product. Second, an administering authority may misunderstand the characteristics, use, or properties of the input material and, therefore, erroneously classify that input. Cf. Bausch & Lomb Inc. v. United States, 148 F.3d 1363 (Fed. Cir.1998). This analytic approach has been consistently applied to classifications throughout U.S. history, and helps to frame the exact question so the court can quickly identify the alleged agency error. Nor is this approach different from the court’s previous reviews of Commerce’s choice of surrogate values. Commerce must articulate in what way the surrogate value chosen relates to the factor input. Hebei II, 29 CIT at-,366 F.Supp.2d at 1273. “It will not do for a court to be compelled to guess at the theory underlying the agency’s action; nor can a court be expected to chisel that which must be precise from what the agency has left vague and indecisive.” Siderca, S.A.I.C., v. United States, 28 CIT -, -, 350 F.Supp.2d 1223, 1236 n. 15 (2004) (quoting SEC v. Chenery Corp., 332 U.S. 194, 196-97, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947)). With that said, the court appreciates that the standard of review here, i.e., the substantial evidence test, is significantly different from that for Customs’ classifications cases in which the court reviews Customs’ finding of fact and conclusions of law largely de novo. Furthermore, Commerce’s goal here is different from Customs’. Whereas in Customs cases determining the proper specific classification is paramount, Jarvis Clark Co. v. United States, 733 F.2d 873, 878 (Fed.Cir.1984), in Commerce cases, Commerce is using the HTS[I] merely to approximate the cost of a factor of production, Nation Ford Chem. Co. v. United States, 166 F.3d 1373, 1377 (Fed.Cir.1999). Therefore, unlike Customs cases, the court will only upset Commerce’s decision if no reasonable mind could find Commerce’s choice of HTS[I] heading or subheading is proper. Pursuant to these principles, the court examines the parties’ arguments regarding the valuation of certain factors of production concerning: (1) hooks, hinges and connectors; (2) resin; (3) styrofoam; (4) cardboard; (5) certain metal components. *1309 (A) Hooks and Connectors In its submissions to Commerce, Dorbest described its hooks and connectors as made out of iron and its hinges as made out of metal. Letter from Jeffrey S. Grimson, Grunfield, Desiderio, Liebowitz, Silverman & Klestadt LLP,- on behalf of Dorbest Limited, to The Honorable Donald L. Evans, Secretary of Commerce, Re: Response to the Department’s Request of HTS Data in Wooden Bedroom Furniture from the People’s Republic of China (Investigation A-570-890), P.R. Doc. 1152 Attach. 1 at 2-3 (fr.8-9) (May 26, 2004) (“Dorbest HTS Submission”)', Issues & Decision Mem., P.R. Doc.1933 at 169 (Cmt.19). For these inputs, Dorbest proposed subheading 8302.1009, HTS[I] (“hardware, fixtures, castors, etc., and parts, base metal: hinges & parts thereof’). Pis.’ Br. 59. In its Preliminary Determination, Commerce used Dorbest’s proposed category. Issues and Decision Mem., P.R. Doc.1933 at 171 (Cmt.19). In its Final Determination Commerce determined “not to use the HTS category 8302.10.09 because that category is no longer a valid HTS category.” Id. Instead, Commerce chose to use subheading 8302.4200, HTS[I] (“mountings, fittings and similar articles” that are “suitable for furniture”) to value hooks and connectors. Id. Dorbest argues that Commerce’s determination is not supported by substantial evidence and that Commerce should have utilized the HTS[I] subheading suggested by Dorbest and inflated the values so that they would be contemporaneous with the POI. Additionally, with respect to hooks and connectors, Dorbest claims that Commerce has neither provided a rationale nor articulated how the subheading proposed is related to the input. For its part, Commerce has suggested that the description of its chosen HTS[I] subheading “closely resembles” the description of the input provided by Dorbest. Memorandum from Aishe Allen, Case Analyst, to File, through Robert Bolling, Program Manager, Re: Analysis Memorandum for the Final Determination of the Antidumping Duty Investigation of Wooden Bedroom Furniture from the People’s Republic of China: Rui Feng Woodwork Co., Ltd. (“Rui Feng Dongguan”), Rui Feng Lumber Development Co., Ltd. (“Rui Feng Shenzen”), and their parent company Dorbest Limited (collectively “Dorbest”), Dep’t of Commerce, Attach, to Letter from Laurie Parkhill, Office Director, International Trade Administration, U.S. Department of Commerce, to Jeffrey S. Grimson, Grun-feld, Desiderio, Lebowitz, Silverman & Klestadt LLP, at 7, P.R; Doc.1936 (fr.9) (Nov, 17, 2004). (“Dorbest Calculation Mem.”) Commerce may have determined here that mountings and fittings or those articles similar to mountings and fittings include hooks and connectors. This, however, is not apparent to the court. Though Commerce has stated that there is a match, Commerce has not articulated in what way the inputs “closely resemble” the HTS[I] subheading description. See Hebei II, 29 CIT at-, 366 F.Supp.2d at 1273. Therefore, with respect to hooks and connectors, Commerce must explain how the chosen subheading is rationally connected to the factor input. If Commerce finds the chosen subheading does not include hooks and connectors, it may revert to the method utilized in the Preliminary Determination of inflating an expired HTS[I] subheading, Pis.’ Br. 69, or find another suitable subheading or data set. (B) Hinges For hinges, Commerce determined that subheading 8302.1000, HTS[I], (“hinges made out of different types of metal”) was *1310 the appropriate classification because Dor-best did not specify, in its description-of this input in its May 26, 2004, submission, the type of metal it uses in its hinges. Issues and Decision Mem,., P.R. Doc.1933 at 171 (Cmt.19); Dorbest HTS Submission, P.R. Doc. 1152 Attach. 1 at 2 (fr.8). Dorbest avers that Commerce’s chosen subheading, though it specifically refers to metal hinges used in furniture, is inappropriate because it covers different types of metal. Pis.’ Br. 69. Dorbest claims that since the hinges Dorbest used in its wooden bedroom furniture production are not made of brass, id., a subheading that includes hinges “of different types of metal” cannot be specific to the input, presumably because it could potentially include brass. Dorbest, however, does not claim that other various types of metal are not used in its hinges, only that its hinges are not made of brass. Neither did Dorbest propose a more suitable, contemporaneous HTS[I] subheading. As Commerce is faced with two suboptimal choices, it is reasonable for Commerce to choose a contemporaneous subheading that, based on the description, appears to resemble the input, over an inflated valuation of an expired category to determine the best available information, i.e., it is reasonable for Commerce to have found that the better information was in finding a valuation of hinges made of various kinds of metal, rather than inflating the values of an expired HTS[I] subheading. (C) Resin Dorbest suggested that Commerce value Dorbest’s resin applique input using expired subheading 3926.4009, HTS[I], (“articles of plastics (inc. polymers & resins)”). Pis.’ Br. 56. Dorbest, after initially labeling this input as “ornament”, renamed it “resin applique” and described it as a “PVC and polymer used for decorating.” Dorbest HTS Submission, P.R. Doc. 1152 Attach. 1 at 5 (fr.11). Because subheading 3926.4009, HTS[I], was phased out, Commerce decided not to value Dorbest’s resin input using Dorbest’s suggested category. Issues & Decision Mem., P.R. Doc.1933 at 170-71 (Cmt.19). Instead, Commerce applied subheading 3926.3090, HTS[I], which covers “other articles of plastics and articles of other materials ... fittings for furniture, coachwork or the like.” Id. Pis.’ Br. 57-58. Subheading 3926.4009, HTS[I], appeal's to apply to ornamental articles while subheading 3926.3090, HTS[I], appears to apply to furniture fittings. In response to a supplemental question from the court as to whether or not Commerce’s selection was based on the nature of Dorbest’s resin applique or an analysis of what is the appropriate subheading, Commerce stated that “[t]he nature of this dispute concerns how Commerce chose to value Dorbest’s resin input, not the nature of the resin input.” Def.’s Resp. Court’s March 10, 2006 Questions 22 (“Def.’s Resp. Court’s Questions”). Commerce further clarified, in its response to the court’s question, that “Commerce selected this category because it most closely resembles Dorbest’s proposed but expired category, and comports with Commerce’s preference that the factor value information be contemporaneous.” Id. at 23 (citations omitted). While the court agrees that contemporaneity is an important factor to consider when evaluating surrogate value information, the use of contemporaneity as the sole justification for its decision does not comport with Commerce’s statements that contemporaneity is but one of several criteria when selecting surrogate value information. See Factors Valuation Mem., P.R. Doc. 1329 at 2; see also Hebei II, 29 CIT at -, 366 F.Supp.2d at 1275. Additionally, it does *1311 not appear that Commerce advanced this argument on the record before the court. Commerce determined that subheading 3926.3090, HTS[I] most closely resembled the subheading advanced by Dorbest (and by implication the factor description advanced by Dorbest). Commerce has not explained in what way the selected subheading resembles the suggested subheading, or, more importantly, the factor input. The court cannot be expected to fathom what the link is between the surrogate value chosen and the factor input. See Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 9 L.Ed.2d 207(1962) (the agency did not “articulate any rational connection between the facts found and the choice made.”); Hebei II, 29 CIT at-, 366 F.Supp.2d at 1273; Siderca, 28 CIT at -, 350 F.Supp.2d at 1236 n. 15. Commerce must provide an explanation or analysis as to why it chose this particular subheading. Based on the record before the court, it is not clear whether Dorbest is representing its “resin applique” input to be ornamentation. If, in the Remand Determination process, Commerce finds that the factor input is ornamental in nature, Commerce needs to then determine what the appropriate subheading would be to reflect that it is ornamental. (D) Styrofoam Dorbest described its styrofoam input as “styrofoam.” It also proffered subheading 3901.1010, HTS[I], covering “Polymers of ethylene, in primary forms: Polyethylene having a specific gravity of less than 0.94.” After submitting this product description and its classification thereof, Dorbest acknowledged that its proposed classification was submitted in error and proposed the subheading Commerce used to classify other respondents’ styrofoam, i.e, subheading 3903.1100, HTS[I], covering “Polymers of styrene, in primary forms: Polystyrene: Expansible polystrene.” Pis.’ Br. 55-56; Def.’s Br. 67. In rejecting Dorbest’s proposed classification, Commerce explained that Dorbest “did not provide Commerce with any reasoning or explanation as to why this value would be more appropriate for Dorbest’s input.” Def.’s Br. 67. Therefore, Commerce employed a basket provision subheading 3921.1100, HTS[I], covering “other plates, sheets, film, foil and strips of plastics: Cellular: Polymers of styrene,” Issues & Decision Mem., P.R. 1933 at 171 (Cmt.19); Pis.’ Br. 55. The court subsequently asked the parties about the nature of Dorbest’s styro-foam, because the subheading suggested by Dorbest and the subheading chosen by Commerce appeared to refer to different forms of styrofoam. 41 Once again, Commerce’s response to the court’s supplemental question stated that the nature of the dispute was one of classification and not the nature of Dorbest’s input. Def.’s Resp. Court’s Questions 18. Commerce further clarified that “Dorbest advised that its styrofoam input is of ‘plates, sheets, film, foil and strip plastics,’ and Commerce accepted this description in its final determination.” Id.; See Dorbest Limited An-tidumping Duty Investigation: Wooden *1312 Bedroom Furniture from the People’s Republic of China A-570-890 — Period of Review 0Jp/01/2003-09/30/2003 (May 21, 2004), Attach, to Letter from Jeffrey Grimson, Grunfeld, Desiderio, Lebowitz, Silverman & Klestadt LLP, on behalf of Dorbest Limited, to the Secretary of Commerce Re: Supplemental Questionnaire Sections C & D Responses in Wooden Bedrooom Furniture from China (Inv. No. A-570-890), P.R. Doc. 1086 at fr. 49 (May 24, 2004) (“Dorbest Sections C & D Resp.”)(Dorbest describes its input as “styrofoam sheet.”); see also Issues & Decision Mem., P.R. Doc.1938 , at 170 (Cmt.19). Commerce based its surrogate value selection on the information provided by Dorbest as to its input. Def.’s Resp. Court’s Questions 20. An examination of the various descriptions of factor inputs provided by the respondents reveals that, understandably, respondents did not provide an abundance of detail in their descriptions. As such, Commerce credited Respondents’ descriptions of the nature of their products as much as possible, and sought to match the descriptions to HTS[I] subheadings. See, e.g., Dorbest HTS Submission, P.R. Doe. 1152; Def.’s Resp. Court’s Questions 19 & 22. Though it may be the case that the subheading does not accurately capture the nature of Dorbest’s styrofoam, Dorbest did not present arguments or evidence that demonstrated the nature of its styro-foam to be other than what it described. Therefore, Commerce selected an HTS[I] subheading whose description included the word “sheets”, which was a reasonable selection given that Dorbest described its input as “styrofoam sheet.” Accordingly, the court finds that Commerce’s selection of 3921.1000, HTS[I] is supported by substantial evidence. (E) Cardboard Dorbest described its packing cardboard as “paper cardboard for protecting furniture at the time of shipping.” Issues and Decision Mem., P.R. Doc.1933 at 167 (Cmt.19). In its Final Determination, Commerce assigned subheading 4808.1000, HTS[I], for Dorbest’s packing cardboard. Id. at 170. Commerce assigned this subheading because it covered “corrugated paper/paperboard whether or not perforated,”' noting that Dorbest’s verification report did not indicate that the cardboard was made of corrugated paper, despite Commerce’s observation at verification that Dorbest’s cardboard is made of corrugated paper. Id. Throughout the investigation, Dorbest suggested that Commerce value Dorbest’s cardboard input using subheading, 4808. 9000, HTS[I], covering “other paper and paperboard corrugated.” Resp. Def.-Intervenors Dorbest Ltd., Rui Feng Woodwork (Dongguan) Co., Ltd. & Rui Feng Lumber to PI. Am. Furniture Mfrs. Comm. Legal Trade’s Br. Supp. Mot. J. Agency R. Re: Selection Surrogate Values & Calculation Financial Ratios 25-30 (“Dorbest Resp. AFMC”). After the Final Determination, Commerce reviewed the record, which indicated that Commerce does not use perforated cardboard. Memorandum from Barbara E. Tillman, Acting Deputy Assistant Secretary for Import Administration, to Joseph A. Spetrini Acting Assistant Secretary for Import Administration Re: Issues and Decision Memorandum for the Amended Final Determination in the Less-Than-Fair-Value Investigation of Wooden Bedroom Furniture from the People’s Republic of China, P.R. Doc.1996 at 10-12 (Cmt.5) Dep’t of Commerce (Dec. 27, 2004) (“Amended Final Issues & Decision Mem.”). As a result of examining this record evidence, Commerce amended its choice for Dorbest’s cardboard input to subheading 4808. 9000, HTS [I], (“other”) stating that “there is nothing on the record that contradicts Dorbest’s statement that *1313 it did not use perforated cardboard as an input.” Id. at 12. AFMC has questioned Commerce’s decision to use a basket provision, subheading 4808.9000, HTS[I], to classify Dorbest’s cardboard, instead of a specific provision under which Dorbest’s cardboard is classifiable, subheading 4808. 1000, HTS[I]. AFMC asserts that “Commerce’s recitation of evidence that Dorbest did not use perforated cardboard does not constitute substantial evidence supporting the change in classification.” AFMC’s Br. 23 (“AFMC Br.”) (emphasis in original). Logic would dictate that the use of the word “other” in the basket subheading indicates that this subheading should only be used if all other subheadings within that heading are exhausted and have been deemed inappropriate. This view is supported by classification law — under well-established classification principles, it is only appropriate to use a basket provision when no other subheading applies. See, e.g., Witex, U.S.A., Inc. v. United States, 28 CIT -, -, 353 F.Supp.2d 1310, 1319 & n. 16 (2004). 42 It appears on the record before the court, that it is uncontro-verted that Dorbest’s cardboard is “paper [or] paperboard.” There also appears to be no dispute that the cardboard is “corrugated.” Under the common understanding of the phrase “[w]hether or not,” the finding of the perforated nature of the cardboard is irrelevant to the proper classification of the cardboard. Cf. XX Oxford English Dictionary 220 (2nd ed.1989) (wherein the third definition of “whether” is “[w]hichev- ef of the two ... [n]o matter which of the two”); id. at 221 (wherein the sixth definition of “whether” is “whether or no ... less freq. not ... [i] n any case, at all events.”) Webster’s Third new International Dictionary 2,602 (1993) (“whether or no also whether or not adv: in any case”). Reviewed in the light of this analysis of the terms of the subheadings, Commerce’s choice of the catch-all subheading 4808.9000, HTS[I], absent an adequate explanation, is not supported by substantial evidence, where the alternate subheading more clearly fits the description of the factor input. This issue is therefore remanded for Commerce to either explain, with reference to the description of the input, why 4808.9000 is the appropriate classification, or to change the classification accordingly. (F) Iron Components Dorbest described its iron components inputs as “iron-welded shapes and canopies” as well as “iron panel for headboard, and iron panel used to keep the drawer fixed.” Exh. 16 of Dorbest Limited Antidumping Duty Investigation: Wooden Bedroom Furniture from the People’s Republic of China A-570-890— Period of Review 01/01/2003-9/80/2003 Posh-Preliminary Determination Supplemental Questionnaire Response, (July 13, 2004) Attach, to Letter from Jeffrey S. Grimson, Grunfeld, Desiderio, Lebowitz, Silverman & Klestadt LLP on behalf of Dorbest Limited to the Secretary of Commerce Re: Response to DOC’s July 1, *1314 2004 Supplemental Questionnaire in Wooden Bedroom Furniture from China (Inv. No. A-570-890), P.R. Doc. 1514 at fr. 224; Dorbest HTS Submission, P.R. Doc. 1152 at fr.9. For these components Dorbest suggested subheading 7216.9000, HTS[I], covering “angles, shapes & sections of iron & nonalloy steel: Angles, shapes section, iron/nonalloy steel nesoi.” Id. Commerce ultimately selected the subheading proposed by Petitioners, 8302.4200, HTS[I], covering “mountings, fittings and similar articles” that are “suitable for furniture.” 43 Issues & Decision Mem., P.R. Doc.1993 at 171 (Cmt.19). Commerce now requests a remand to exclude non-scope metal canopies from the Dorbest calculations. The court grants this request. The court understands that the remainder of the iron components at issue is the portion of iron components described by Dorbest as “iron panel for headboard, and iron panel used to keep the drawer fixed.” As there is no argument as to the nature of Dorbest’s input, the record before the court, in addition to the Explanatory Notes, demonstrates that Commerce’s choice for iron components was reasonable. An examination of the Explanatory Notes that accompanies the subheading chosen by Commerce indicates that the subheading chosen by Commerce appears to include items such as the ones described by Dorbest in its submission. Accordingly, this aspect of Commerce’s determination is supported by substantial evidence and is otherwise in accordance with law. III. DISCRETE COMPANY-SPECIFIC ISSUES The parties raise additional discrete issues. Dorbest argues that (A) Commerce erred in failing to ©eliminate the spare parts discount adjustment to U.S. price for Dorbest; (ii) remove non-scope metal canopies and other metal parts from Dorbest’s antidumping duty calculation; (iii) treat certain expenses on incoming raw materials as direct expenses for Dorbest. Commerce has requested a voluntary remand on these issues. Def.’s Resp. Pis.’ Party-Specific Mot. J. Agency R. 3. (“Def.’s Party-Specific Resp.”). AFMC argues that (B) Commerce improperly used its export price methodology, as opposed to its constructed export price methodology, for certain sales after erroneously concluding that the sales were made between unaffiliated parties; Additionally, Dorbest asserts that (C) Commerce erred in applying its zeroing methodology. The court will address each issue in turn. (A) Voluntary Remand Issues In its brief, Commerce requested a voluntary remand to (i) eliminate the spare parts discount adjustment to U.S. price for Dorbest; (ii) remove non-scope metal canopies and other metal parts from Dorbest’s antidumping duty calculation; (iii) treat certain expenses on incoming raw materials as direct expenses for Dorbest. Def.’s Party-Specific Resp. 3. *1315 The Court of Appeals for the Federal Circuit has approved the grant of a request for a voluntary remand where “the agency’s concern is substantial and legitimate.” SKF USA Inc. v. United States, 254 F.3d 1022, 1029 (Fed.Cir.2001); Shakeproof Assembly Components Div. of Ill. Tool Works, Inc. v. United States, 29 CIT -, -, 412 F.Supp.2d 1330, 1337 (2005). If the agency’s determination is unsupportable, there exists no good reason to expend judicial resources to state the obvious. Cf. Lawrence v. Chater, 516 U.S. 163, 167, 116 S.Ct. 604, 133 L.Ed.2d 545 (1996) (per curiam). Here, Commerce’s explanation for why a voluntary remand is necessary shows that its concern is substantial and legitimate. Although AFMC disagrees that remand is necessary, it fails to state any good cause for why voluntary remand should not be granted. Accordingly, the court grants a voluntary remand on these issues. The only remaining question regarding the voluntary remand is its scope. Dor-best argues that the court should include issues Dorbest did not raise in its summons and complaint, but which were related to the issue of metal spare parts. Specifically, Dorbest asserts that: Having now agreed that metal spare parts do not belong in the antidumping calculation in the context of the specific claim raised by Dorbest to the Court, [Commerce] should take reasonable steps to reverse the improper instruction it gave to Dorbest [to include all metal spare parts in its calculations]. Dorbest Resp. Court’s Questions March 7, 2006 3(“Dorbest Supp. Br.”). While the court can appreciate that Commerce may have erroneously requested that Dorbest include metal spare parts in its calculations, Dorbest did not preserve these issues on appeal. As such, Dorbest has not appropriately exhausted its administrative remedies. 28 U.S.C. § 2637(d). See also USCIT R. 8(a)(1); United States v. Thibodeaux, 211 F.3d 910, 912 (5th Cir.2000) (failure to brief an issue is waiver). Accordingly, Dorbest’s request is denied. In its brief in response to the court’s questions of July 13, 2006, Dorbest argues that a contrary conclusion is required by Shandong Huarong Mach. Co. Ltd. v. United States, 30 CIT-, 435 F.Supp.2d 1261, 1292-93 (2006), claiming that the Shandong court recognized that raising the issue of commissions in administrative proceedings was sufficient to preclude a finding of failure to exhaust on the more general issue of Commerce’s refusal to make a circumstance of sale adjustment. But as the Shandong court explained, the DefendanNIntervenor had made an argument at the agency level that was equivalent to the argument raised before the court. Id. at 1293. The same cannot be said here. (B) Zeroing Dorbest also attacks Commerce’s zeroing methodology in this case. Under Commerce’s current practice in calculating the dumping margin, Commerce first finds the average normal value. Commerce then subtracts each individual export and constructed export price from this average normal value. If the result is negative, i.e., the particular sale was not dumped, Commerce assigns the transaction a “zero” value. “Commerce [then] calculates the dumping margin by dividing the combined unit margins of the dumped sales by the value of dumped and nondumped U.S. sales.” See Böwe Passat Reinigungs-Und Wäschereitechnik Gmbh v. United States, 20 CIT 558, 570, 926 F.Supp. 1138, 1149 (1996). Commerce’s zeroing metholodgy has been attacked numerous times before this court, see, e.g., NSK Ltd. v. United States., 30 CIT - — , 416 F.Supp.2d 1334 (2006); *1316 SNR Roulements v. United States, 28 CIT -, -; 341 F.Supp.2d 1334 (2004); Böwe Passat Reinigungs-Und Wäschereitechnik, 20 CIT at 558, 926 F.Supp. at 1149; Serampore Indus. Pvt. Ltd. v. U.S. Dep’t of Commerce, 11 CIT 866, 675 F.Supp. 1354 (1987), and before the Court of Appeals for the Federal Circuit, Timken Co. v. United States, 354 F.3d 1334 (Fed. Cir.2004); Corus Staal BV v. Dep’t of Commerce, 395 F.3d 1343 (Fed.Cir.2005). No court has ever disapproved Commerce’s methodology. Recently, in Timken, 354 F.3d 1334, the Court of Appeals for the Federal Circuit identified 19 U.S.C. § 1677(35)(A) (defining dumping margin as the amount normal value exceeds export or constructed export price) as the relevant statutory provision. The Court of Appeals held that section 1677(35)(A), even when read in light of 19 U.S.C. § 1677b(a) (requiring a fair comparison between normal value and export/constructed export price), permitted (although did not require) zeroing. In reaching its conclusion, the court also rejected an argument that zeroing was unlawful because the practice violated the Uruguay Round Agreements as interpreted by the World Trade Organization (“WTO”). The court noted that (a) the United States was not a party to the WTO dispute and that (b) the WTO dispute involved an investigation whereas the challenge before the court involved an administrative review. The Timken court implicitly left the door open to the possibility that (a) other statutory provisions, (b) a decision by the WTO against the United States, or (c) a challenge to an investigation might change the outcome. The court in Corus Staal, 395 F.3d 1343 closed that door. In Corns Staal, the court held that a WTO decision could not compel the United States to abandon a regulatory approach and that any decision to abandon a permissible regulatory approach in light of a WTO decision was left to the unfettered discretion of the political branches. The court also summarily held that Timken foreclosed arguments as to the permissibility of zeroing under U.S. law. As a result of these decisions, there appear to be no grounds to challenge the zeroing methodology. See, e.g., NSK Ltd. v. United States., 30 CIT -, 416 F.Supp.2d 1334, 1338 (2006). Nevertheless, Respondents argue that zeroing is impermissible because (1) 19 U.S.C. § 1677f-l(d)(l)(B) would be rendered superfluous if Commerce were permitted to zero (in the manner it does); (2) Commerce’s application of zeroing here is overly distortive; and (3) Commerce has published notice in the Federal Register that it is seeking comments on whether to replace the methodology. The court is unpersuaded by any of these arguments. Respondents’ argument that section 1677f-l(d)(l)(B) informs the zeroing question is misplaced. The Court of Appeals has already interpreted the statute — Respondents’ argument is no more than an attempt to reargue a settled issue by raising a new argument for interpretation. Because Congress can always correct erroneous interpretations of law, Patterson v. McLean Credit Union, 491 U.S. 164, 172-73, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989), and because predictability in law (especially in the case of foreign commerce) is highly valued, Zenith Radio Corp. v. United States, 437 U.S. 443, 457-58, 98 S.Ct. 2441, 57 L.Ed.2d 337 (1978); Scherk v. Alberto-Culver Co., 417 U.S. 506, 517, 94 S.Ct. 2449, 41 L.Ed.2d 270 (1974), “Considerations of stare decisis are particularly forceful in the area of statutory construction, especially when a unanimous interpretation of a statute has been accepted as settled law for several decades[,]” IBP, Inc. v. Alvarez, 546 U.S. 21, 126 S.Ct. 514, 523, 163 L.Ed.2d 288 (2005). If parties *1317 were free to introduce new theories of interpretation, or invoke canons of interpretation left unaddressed by prior courts, an interpretation of a statute would never become settled. Accordingly, given that this issue has been unanimously addressed by multiple opinions, the court finds Respondents’ argument barred by stare deci-sis. Nor are Respondents’ claims here that zeroing is overly distortive availing. The Court of Appeals for the Federal Circuit has categorically approved zeroing. Moreover, despite the observations of this court that it is “wary of [such] a methodology that intentionally minimizes the impact of nondumped transactions by manipulating the data of potentially equalizing sales,” Corus Staal BV v. United States DOC, 27 CIT-,-, 259 F.Supp.2d 1253, 1263 (2003), and that this methodology “introduces a statistical bias in the calculation of dumping margins,” Bowe Passat, 20 CIT at 570, 926 F.Supp. at 1149, the court has nevertheless unanimously (albeit not enthusiastically) sustained the zeroing methodology. Respondent’s last argument similarly fails. Although Commerce may be in the process of changing its methodolgy, no rule has been set forth which necessarily would apply retroactively. Therefore, given that Commerce’s current practice has been found permissible, that practice was applicable during this investigation, and no new rule with retroactive effect has been promulgated, Respondents fail to state a case requiring remand on this issue. IV. FACTS OTHERWISE AVAILABLE/AD VERSE INFERENCES AFMC further raises complaints about Commerce’s decision not to apply adverse inferences in other specific instances. Given the inherent difficulties in obtaining information, gaps in the factual record are largely inevitable. When such gaps arise, section 1677e(a) requires Commerce to use “facts otherwise available” to fill gaps in the administrative record. Because Commerce’s use of “facts otherwise available” is borne out of necessity, the specific reason for the gap is “of no moment,” Nippon Steel Corp. v. United States, 337 F.3d 1373, 1381 (Fed.Cir.2003), and Commerce is required to use facts otherwise available in calculating the dumping margin as accurately as possible. See 19 C.F.R. § 351.308; F.LLI De Cecco Di Filippo Fara S. Martino S.p.A. v. United States, 216 F.3d 1027, 1032 (Fed.Cir.2000) (“De Ceceo ”). Nevertheless, section 1677e also recognizes that parties may bear responsibility for informational gaps. Appropriately, in order to prompt parties to timely and diligently provide such information, id., and because a party’s failure to provide requested information may lead to an inference that the party is attempting to conceal damaging information to its case, see Rhone Poulenc, Inc. v. United States, 899 F.2d 1185, 1190 (Fed.Cir.1990); Shanghai Taoen Int’l Trading Co. v. United States, 29 CIT -, -, 360 F.Supp.2d 1339, 1344 (2005), when Commerce finds that a party has “failed to cooperate by not acting to the best of its ability ... [to meet] a request for information from” Commerce, section 1677e(b) grants Commerce the discretion to use adverse inferences when “selecting” from the array of “facts otherwise available.” See 19 U.S.C. § 1677e(b). That this authority is discretionary is manifested by section 1677e(b)’s use of the permissive term “may,” which stands in contraposition to section 1677e(a)’s use of the mandatory term “shall.” See AK Steel Corp. v. United States, 28 CIT,-, 346 F.Supp.2d 1348, 1355 (2004). See generally Martin v. Franklin Capital Corp., 546 U.S. 132, 126 S.Ct. 704, 708-709, 163 L.Ed.2d 547 (2005) (the word “may” con *1318 notes discretion where the statute has in other instances used the word “shall”); Jama v. Immigration & Customs Enforcement, 543 U.S. 335, 125 S.Ct. 694, 697, 160 L.Ed.2d 708 (2005) (same). In exercising its discretion to use adverse inferences in its selection of facts otherwise available, section 1677e(b) first requires that the requirements of section 1677e(a) be satisfied. See Gerber Food (Yunnan) Co. v. United States, 29 CIT -, -, 387 F.Supp.2d 1270, 1284 (2005); New World Pasta Co. v. United States, 28 CIT --,-, 316 F.Supp.2d 1338, 1350 (2004). Section 1677e(a) requires that there be a gap in the record of verifiable information due to a party’s failure to supply necessary or reliable information in response to an information request from Commerce. See NTN Bearing Corp. of Am. v. United States, 368 F.3d 1369, 1377 (Fed.Cir.2004) (“All that is required is that the necessary information be unavailable on the record.”); Nippon Steel, 337 F.3d at 1381; SAA at 869, as reprinted in 1994 U.S.C.C.A.N. 4040, 4198 (subsection 1677e(a) pertains to situations “where requested information is missing from the record or cannot be used because, for example, it has not been provided, it was provided late, or Commerce could not verify the information.”). Second, and inextricably intertwined with the first requirement, this gap in the record must be factual in nature; where a gap is not factual in nature, section 1677e is not triggered. See, e.g., Gerber Food (Yunnan), 29 CIT at-, 387 F.Supp.2d at 1285 (finding that an assessment rate is not a fact); NSK Ltd. v. United States, 29 CIT-,-, 358 F.Supp.2d 1276, 1290 (2005) (use of a methodology is not a fact); Tung Fong Indus. Co. v. United States, 28 CIT,-, 318 F.Supp.2d 1321, 1335 (2004) (a calculation is not a fact); cf. Transcom, Inc. v. United States, 294 F.3d 1371, 1381 (Fed.Cir.2002) (applying a presumption did not invoke Commerce’s authority to use the best information under the predecessor statute to modern section 1677e). Third, Commerce must take into consideration the requirements of section 1677m (d). Generally speaking, section 1677m (d) requires Commerce to inform parties of their deficient submissions and, when practicable, allow the party remedy or explain this deficiency. 44 See, e.g., Ta Chen Stainless Steel Pipe, 298 F.3d at 1341; NEC Corp. v. United States, 151 F.3d 1361, 1374 (Fed.Cir.1998); Gerber Food (Yunnan), 29 CIT at -, 387 F.Supp.2d at 1280-81. But cf. NTN Bearing Corp. v. United States, 368 F.3d 1369, 1376 (Fed.Cir.2004) (explaining that Commerce need not request information in order to later apply adverse inferences). Fourth, Commerce must determine that this deficiency was due to a party’s failure to act to the “best of its ability” in responding to an information request from Commerce. Under this test, Commerce must make (a) an “objective showing” that a reasonable and responsible importer would have been able to provide such information and (b) a “subjective showing” that the respondent failed to cooperate by not maintaining records or failing to put maximum effort in acquiring such information. See Nippon Steel, 337 F.3d at 1382; *1319 Tung Fong Indus. Co. v. United States, 28 CIT-,-, 318 F.Supp.2d 1321, 1335 (2004) (where Commerce fails to explain how a party could have met a deadline, Commerce may not resort to the use of adverse inferences); China Steel Corp. v. United States, 27 CIT-, 264 F.Supp.2d 1339, 1360-61 (2003). Only when each of these elements are found does Commerce then have discretion to elect whether or not to apply adverse inferences, notwithstanding that its discretion as to how it applies its discretion is bounded by the requirements of law. Compare Timken Co. v. United States, 354 F.3d 1334, 1346 (Fed.Cir.2004) (“Commerce has discretion to apply adverse inferences.”) with De Cecco, 216 F.3d at 1033 (holding that Commerce’s selection of facts otherwise available was overly punitive and, therefore, not in accordance with law). Because this court only has authority to review a final determination to ensure that Commerce has acted in accordance with law and that its conclusions are based on substantial evidence, see 19 U.S.C. § 1516a(b)(l)(B)(i); cf. 19 U.S.C. § 1516a(b)(l)(A) & (B)(ii) (judicial review provided to ensure that an action by Commerce is not “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”), unless a party alleges that Commerce has exercised its discretion in an unlawful manner, De Cecco, 216 F.3d at 1033; D & L Supply Co. v. United States, 113 F.3d 1220, 1224 (Fed. Cir.1997), or that the factual predicates of Commerce’s decision were unsupported by substantial evidence, see, e.g., Kao Hsing Chang Iron & Steel Corp. v. United States, 26 CIT 536, 542 (2002), this court may not disturb Commerce’s decision, 45 Ta Chen Stainless Steel Pipe, 298 F.3d at 1340; SNR Roulements, 28 CIT at -, 341 F.Supp.2d at 1353 (“Because Commerce is not bound by prior decisions based on different facts ... ”); AK Steel, 28 CIT -, 346 F.Supp.2d at 1356. In this case, the parties allege that Commerce impermissibly failed to apply adverse inferences where it should have and did apply adverse inferences when it should not haye. Specifically, AFMC argues that Commerce failed to apply adverse inferences (A) against all parties with regard to their HTS[I] submissions. Additionally, (B) Dorbest argues that Commerce improperly applied adverse inferences in rejecting an offset claim for its wood scraps. (A) Factor Inputs Before Commerce, AFMC argued that various respondents failed to cooperate to the best of their abilities by failing to justify their proposed HTS[I] classifications. In particular, AFMC maintained that: [T]he respondents failed to provide complete and accurate information for Commerce to use in calculating the final dumping margins. Specifically, the respondents either failed to respond to Commerce’s request for information as to the classification of raw material inputs under the Indian Harmonized Tariff Schedule ... provided information that was completely useless to make such classifications, or provided classifications that were too vague or broad to permit accurate classification under the Indian HTS at the eight-digit level. *1320 AFMC’s Br. Reply Def.’s & Def.-Interve-nor’s Resp. Op. AFMC’s Mot. J. Agency R. Re: Selection Surrogate Values & Calculation of Financial Ratios 4. In its determination, Commerce did not dispute that there were deficiencies in Respondents’ answers. Notwithstanding this fact, Commerce declined to use adverse inferences in its selection of facts otherwise available. In particular, Commerce found that: This investigation has presented a host of complex issues with respect to HTS categories and factor valuations, given the hundreds of inputs that are necessary to produce the subject merchandise. It is important to recognize that the breadth of the information we have requested in this investigation is substantial. We have balanced that recognition with the importance of ensuring that the information we receive is adequate for the purpose of calculating an accurate antidumping margin. We have examined each of the Petitioners’ criticisms of the respondents’ HTS and valuation recommendations carefully to ensure that the values we apply to the respondents’ factors are supported by the weight of the evidence on the record. Issues & Decision Mem., P.R. Doc.1933, at 160-61 (Cmt.16). While' Commerce was not satisfied with some of the responses and selected its own values, given the burdensome request, it did not find that imperfect responses amounted to a failure to cooperate. AFMC claims that Commerce’s Final Determination erred by failing to sufficiently analyze and respond to its arguments. Although acknowledging that Commerce need not respond to every argument raised by parties, AFMC claims that here Commerce did not meet the minimum threshold of laying out a clear “path.” The court disagrees that remand is required here. In discussing the volume of information requests Commerce made to the parties, and the limited time for response permitted them, Commerce implicity found that Respondents’ errors were excusable. Certainly, the record supports the factual basis of Commerce’s finding. Nor can the court say Commerce’s result was unlawful. As discussed above, section 1677e (b) grants Commerce discretion not to use adverse inferences. Given that Commerce’s factual finding was well supported by the record, there is no basis here for reversing Commerce’s conclusion. Indeed, granting AFMC’s request would convert section 1677e(b)’s use of the term “may” into “shall” — a result contrary to the intent of Congress. 46 *1321 (B) Wood Scraps Dorbest asserts that when Commerce calculated the constructed export price of its merchandise, Commerce failed to deduct the economic value of certain byproducts of its manufacturing process. Specifically, Dorbest claims that it generates wood scraps and scrap cardboard in its production process and that these byproducts “re-enter the production process” except for portions which are “burned in the kitchen for cooking and for hot water for the dormitory.” Dorbest Sections C & D Resp., P.R. Doc. 1086 at fr. 48. In its Determination, Commerce found that Dorbest failed to explain whether it based its “scrap-allocation methodology on sales or production figures for scrap.” Therefore, Commerce found inadequate “Dorbest’s explanation for its calculation of the by-product offset ... because Dorbest did not provide worksheets or any other evidence on the record to demonstrate how it calculated its wood scrap and cardboard scrap offset.” Issues & Decision Mem., P.R. Doc.1933 at 231 (Cmt.33). Prior to denying Dorbest’s offset request, Commerce did not provide Dorbest notice that Dorbest had failed to sufficiently document its claimed deduction. Dorbest claims that Commerce’s failure to notify Dorbest of the deficiencies in its submissions, and Commerce’s subsequent failure to credit Dorbest’s claimed deductions, contravenes Commerce’s duties imposed by 19 U.S.C. § 1677m(d) 47 and, therefore, Commerce’s Determination im-permissibly resorted to using facts otherwise available under 19 U.S.C. § 1677e (a). The court disagrees. Despite Dorbest’s contention, here Commerce did make a follow-up request for Dorbest to explain its numbers. Dorbest Sections C & D Resp., P.R. 1086, fr. 48. This follow-up inquiry discharged Commerce’s obligations under section 1677m(d). Once Commerce did this, no more was required of Commerce under 1677m(d). V. CONCLUSION In summary, the court finds as follows: (i) Commerce’s selection of India as the surrogate country is affirmed; (ii) Commerce’s calculation of the labor rate is remanded; specifically Commerce is to either (a) justify why its data set constitutes the best available information; or (b) incorporate those countries meeting its criteria into the data set; and (c) reconsider its use of its methodology or an alternative method for determining the labor rate for the PRC in this case; (iii) Commerce’s decision not to use the spring 2005 data set in its calculation of the labor rate is affirmed; (iv) Commerce’s selection of surrogate companies for the computation of the financial ratios is remanded; specifically, Commerce may exclude IFP’s 2004 financial statement; Commerce must explain its inclusion of Jayaraja’s financial statement; Commerce must explain its inclusion of financial statements from Swaran, Niza-muddin, Fusion Design, and DnD; (v) Commerce’s use of MSFTI, in general, is affirmed; (vi) Commerce’s use of MSFTI to value mirrors is remanded for a fuller explanation or use of a better data set; (vii) The use of MSFTI to value paints is affirmed; (viii) The use of MSFTI to value cardboard is affirmed; (ix) Commerce’s valuation of hooks and connectors is remanded for a fuller explanation or use of a different subheading; *1322 (x) Commerce’s valuation of resin is remanded; (xi) Commerce’s valuation of hinges is affirmed; (xii) Commerce’s valuation of styrofoam is affirmed; (xiii) Commerce’s valuation of cardboard is remanded; (xiv) Commerce’s valuation of iron components is affirmed; (xv) Commerce’s decision not to apply adverse facts on its valuation of input materials is affirmed; (xvi) Commerce’s denial of Dorbest’s by-product offset claim is affirmed; (xvii) Commerce’s request for a voluntary remand regarding metal spare parts is granted; Commerce shall limit itself to addressing only the stated issues in its request for voluntary remand; (xviii) Commerce’s request for a voluntary remand on the question of non-scope metal canopies and other metal parts is granted; (xix) Commerce’s request for a voluntary remand regarding its treatment of certain expenses on incoming raw materials is granted; (xx) Commerce’s use of zeroing is affirmed. Commerce shall have until March 2, 2007 to issue a remand determination. Parties’ comments shall be due by March 23, 2007. Rebuttal comments shall be due by April 13, 2007. SO ORDERED. Footnotes [1] In investigations of merchandise produced in non-market economies (''NMEs”), if the “prices of the goods produced in an NME are subject to discrepancies which distort their value,” Magnesium Corp. of Am. v. United States, 20 CIT 1092, 1095, 938 F.Supp. 885, 890 (1996), the antidumping statute authorizes Commerce to approximate normal value based on the cost of producing the foreign merchandise (with a margin of profit factored in), see 19 U.S.C. § 1677b(c) (2000). To value these factors of production and profit, Commerce must use the “best available” pricing and cost data for the factors of production taken from “surrogate” economies/companies. Id. Antidumping investigations are bifurcated between Commerce and the International Trade Commission ("ITC”). The ITC makes injury determinations and Commerce makes less-than-fair value ("LTFV”) determinations. If the ITC determines that á domestic industry is materially injured or is threatened with material injury and Commerce determines that the merchandise is being sold at LTFV, Commerce issues an antidumping order directing the United States Customs Service to collect antidumping duties "equal to the amount by which the normal value exceeds the export price (or the constructed export price) for the merchandise.” 19 U.S.C. § 1673 (2000). The "export price” or "constructed export price” is the price at which the subject merchandise is sold in the United States market. 19 U.S.C. § 1677a. "Normal value” is the price of the foreign merchandise in its country of origin, in an appropriate third country, or the foreign product's cost of production. See 19 U.S.C. § 1677b. [2] Other aspects of Commerce's final determination here were challenged in Decca Hospitality Furnishings, LLC v. United States, 29 CIT,-391 F.Supp.2d 1298(2005) and Guangzhou Maria Yee Furnishings, Ltd. v. United States, 29 CIT-,-, 412 F.Supp.2d 1301 (2005). [3] Hereinafter cited as 19 U.S.C. § 1677 et seq. 2000 ed. [4] All references to the Code of Federal Regulations are to the 2003 edition. [5] According to the legislative history of the governing statute, "[t]he term ‘significant producer’ includes any country that is a significant net exporter and, if appropriate, Commerce may use a significant net exporting country in valuing factors.” Conference Report to the 1988 Omnibus Trade & Competitiveness Act, H.R.Rep. No. 100-576, at 590 (1988) ("Conf.Rep.")at 590. This, however, does not, preclude Commerce from finding that a country who is a producer of the subject merchandise but not a net exporter is a significant producer of the subject merchandise based on other criteria. [6] 19 C.F.R. § 351.408(c)(l)-(2) reads: (c) Valuation of factors of production. For purposes of valuing the factors of production ... under section 773(c)(1) of the Act the following rules will apply: (1) Infonnation used to value factors. The Secretary normally will use publicly available information to value factors.... (2) Valuation in a single country. Except for labor, as provided in paragraph (d)(3) of this section, the Secretary normally will value all factors in a single surrogate country. [7] Respondents describe Infodrive and IBIS data as using "customs data to identify inter alia, the specific type of merchandise being imported, quantity of the imports and their customs value, the origin of the merchandise and the importer.” Pis.’ Br. 25. Shing Mark contends that Infodrive India and IBIS report "the date of entry, the Indian HTS, the importer of record, the import description, the quantity, the value, the unit measure, foreign port, foreign country, Indian port, and method of shipment.” Issues & Decision Mem., P.R. Doc.1933 at 131 (Cmt.10). [8] The parties do not seem to agree as to the exact percentage of imports captured by these two different databases. Commerce claims that Infodrive India only accounts for 60 percent of Indian imports, and within that percentage, not all of that is usable, so the amount is actually less. Issues & Decision Mem., P.R. Doc.1933 at 139-40 (Cmt.10). Respondents claim that "Infodrive India reports commercially significant quantities of roughly 73 percent of imports.” Id. Regardless of which percentage is the actual percentage, it is uncontroverted that these two databases do not capture the full universe of imports into India. [9] Respondents also argue that the use of import data is contrary to the clear intent of the statute which requires that factors of production be valued “in” a comparable market economy country. In addition to the fact that this court, and the Federal Circuit have affirmed the use of import statistics to value factors of production (in certain circumstances), see, e.g., Nation Ford Chemical Co. v. U.S., 166 F.3d 1373 (Fed.Cir.1999); China First Pencil Co. v. United States, 30 CIT-, 427 F.Supp.2d 1236 (2006); Wuhan Bee Healthy Co. v. United States, 29 CIT -, 2005 WL 2886229 (Nov. 2, 2005), the fact that Commerce is using import data to value the factors of production does not mean that they are not valuing those factors of production "in” the comparable market economy country. The prices that producers pay for the imported inputs are not created in a vacuum; they are created by market forces, and indicate the value that producers are willing to associate with those imported prices. Therefore, the use of import prices in and of itself is not contrary to the statute. The use of the import statistics, however, still has to be supported by substantial evidence on a case-by-case basis. [10] According to respondents, the "Highland House data were compiled and verified as complete and accurate by the independent accounting firm of Ernst & Young.” Pis.' Br. 24. [11] In making this comment, Commerce also emphasized that "among Indian mirror manufacturers there was 'intense competition that resulted in the downward prices for mirrors.” Issues & Decision Mem., P.R. Doc.1933 at 207 (Cmt.25). The article in which the mirror data appears provides the data that illustrates the effects of said price war. Glass Yug 9 (Apr.-June 2003), Exh. 48 to Steptoe & Johnson, LLP Submission of Surrogate Values for the Factors of Production for Shing Mark Co. Enterprise, Ltd. (Apr. 16, 2004) Attach, to Letter from Thomas L. Trendl, Steptoe & Johnson, LLP on behalf of Shing Mark Co. Enterprise, Ltd. to the Honorable Donald L. Evans, Secretary of Commerce, Re: Wooden Bedroom Furniture from the People’s Rebublic of China: Submission of Swrogate Values for the Factors of Production, (Apr. 16, 2004), P.R. Doc. 761, fr. 446 ("Shing Mark April 16 Submission”). However, as Commerce noted, intense competition is not in and of itself a *1281 reason to reject mirror prices. Issues & Decision Mem., P.R. Doc.1933 at 207 (Cmt.25). Indeed, one of the reasons that Commerce engages in a factor valuation process for non-market economy countries is in order to have factor input prices that are determined by the interaction of competitive forces in a market economy context. [12] Commerce stated specifically, in its Issues & Decision Memorandum: the Department will consider domestic prices where sufficient record evidence demonstrates that the range of grades exist in the reported inputs and a difference of the domestic and import price appear to be caused by the breadth of category. Additionally, the Department determines that Glass Yug is not the best available information due to a lack of detail put on record by the respondents for their factor inputs of mirror and glass and the lack of specific information for the prices reported in Glass Yug. Issues & Decision Mem., P.R. Doc.1933 at 206 (Cmt.25). [13] If Commerce required more specific information than this in order to properly ascertain the most accurate and appropriate surrogate value, such as whether or not mirrors are cut down to a smaller size, or if they are beveled, Issues & Decision Mem. P.R. Doc. 1933 at 207 (Cmt.25), Commerce could request such information, see Hebei II 366 F.Supp.2d at 1273, ("During its investigation or upon remand, Commerce should have established the category of coal used by Hebei or at least established the category or categories of coal normally used to produce the subject merchandise."). The court notes that, as Respondents point out in their brief, "[t]he thickness range of the Indian import data is, of course, completely unknown.” Pls.' Br. 42. Commerce relies upon MSFTI data in order to ensure that it captures within its range of data the inputs actually used, and so is operating on the assumption that the broader data will somehow capture the correct factor input whereas if a narrower category is chosen, there is a chance that it will be valuing the wrong factor of production. This position, however, is only reasonable if (1) Commerce does not have evidence in the record that the narrower data set is of the same inputs as used by Respondents; and (2) the larger data set captures the factor of production in question. Respondents have asserted that the thickness of the mirrors they use is of the same thickness as the Glass Yug data. Pls.’ Br 42. Shing Mark Apr 16 Submission, P.R. Doc. 761, fr. 446; Shing Mark Case Brief, P.R. Doc. 1854 at 36 (fr.58); Lacquer Craft May 26 Submission, P.R. Doc. 1135 at fr. 9-10; Dorbest June 15 Submission, Prop. Doc. 599, fr. 314 & 318. Respondents also assert that there is nothing on the record reflecting any consumption of mirrors outside the range. Pls.’ Br. 42. [14] Commerce also points to the "lack of contemporaneous data for the entire POI” as another reason for rejecting such data. Issues & Decision Mem., P.R. Doc.1933 at 204 (Cmt.25). Again, this argument does not address why this data cannot be used as a reference point to evaluate other data. Also, as noted in the courts previous discussion, contemporaneity, in and of itself should not be viewed as the sole reason to discard data; rather the quality of the data needs to be viewed in its totality. See supra p. 41. [15] This decision was recently affirmed in Sichaun Changdong Electric Co. v. United States, 30 CIT-, 2006 WL 2927571 (Sept. 14, 2006)(upholding Commerce's use of Info-drive data rather than MSFTI data where Infodrive was product specific and contemporaneous). [16] See also Memorandum from Alice Gibbons, Analyst Office of AD/CVD Enforcement to File, Re: Placing Information on the Record Regarding Infodriveindia.com in the Anti-dumping Duty Investigation on Color Television Receivers from the People’s Republic of China (PRC) 2 (Nov. 17, 2004), Attach, to Letter from Eric C. Emerson, Thomas J. Trendl, & Jack R. Hayes, Steptoe & Johnson LLP, to the Honorable Donald L. Evans, Secretary of Commerce, Re: Wooden Bedroom Furniture from the People’s Republic of China: Additional Comments on “Valuation of the Factors of Production”, P.R. Doc. 907 at fr. 125-129 (May 14, 2004). [17] Petitioners claim that the information listed in Infodrive India indicates that the mirrors shipped by Enginetech are "mirror plates” and do not indicate that these were actually rearview mirrors. Issues & Decision Mem., P.R. Doc.1933 at 202 (Cmt.25). Respondents have placed data on the record, however, that, if correct, demonstrate that Enginetech is an autoparts company that sells rearview mirrors for automobiles. See Dor-best Reply Br. 7; Exh. 9 to Letter from John D. Greenwald, Wilmer, Cutler, Pickering, Hale & Dorr, on behalf of, Lacquer Craft Manufacturing Co., Ltd. & Marltor International Furniture (Tianjin) Manufacture Co., to the Honorable Donald L. Evans, Secretary of Commerce, Re: Wooden Bedroom Furniture from China: Surrogate Value Submission, P.R. Doc. 1695, fr. 123 (Aug. 17, 2004) ("Lacquer Craft Aug. 17 Submission”) (referencing www.enginetech-autoparts.com). Respondents also claim that the Taiwanese imports from Enginetech were described as "6 inch mirror with housing”, "MB 100 Mirror plates with lettering,” "TWM mirror plate with lettering.” Dorbest Reply Br. 6; Steptoe & Johnson, LLP, Submission of Factual Information for Valuing Factors of Production of Shing Mark Enterprise Co., Ltd., (Aug. 17, 2004) Attach, to Letter from Eric C. Emerson, Steptoe & Johnson, LLP on behalf of Shing Mark Enterprise Co., Ltd., to the Honorable Donald L. Evans, Secretary of Commerce Re: Antidumping Duty Investigation on Wooden Bedroom Furniture from the People’s Republic of China: Submission of Factual Information for Valuing Factors of Production, Pub. 1694 fr. 161-162 (Aug. 17, 2004)("Shing Mark Aug. 17 Submission”); Lacquer Craft Aug. 17 Submission, Pub. 1695 at fr. 95. The information placed on the record indicates at least a colorable claim that the surrogate value selected by Commerce includes data from rearview or other specialty mirrors. [18] Commerce has previously noted the broad aspect of an HTS[I] subheading. Freshwater Crawfish Tailmeat from the People’s Republic of China, 64 Fed.Reg. 27,961, 27,962 (Dep't Commerce, May 24, 1999) (final results of new shipper review) ("[I]mport data from basket categories can be too broad to be reliable.”). [19] Commerce used the Indonesian data to analyze this issue despite the fact that Commerce considers Indonesia to be a country with subsidies and therefore excluded those products originating from Indonesia from its surrogate value calculation in the Final Determination. Issues & Decision Mem., P.R. Doc. 1933 at 204 (Cmt.25). [20] Commerce does try to explain, post-hoc, the differences between the import prices and the prices provided by Respondents: "... all of the values to which [sic] respondent claims are ‘mutually reinforcing' were championed by respondent companies that had a keen interest in selecting low values. It is hardly shocking that the values they propose as 'more accurate' are at the lower end of the spectrum. Conversely, Indian import values are more representative of prices from all of India because they are not derived from the few values hand picked by respondents. Therefore, it is perfectly logical that they are higher; however, higher values do not equate to inaccurate values, as respondents erroneously assert.” Def.'s Br. 64. This explanation, however, merely relies on the blanket statement that import prices are more representative of prices from all of India. As explained above, this rationale only holds true if the import prices used capture the actual inputs in ques *1288 tion. Accordingly, Commerce's rationale does not constitute a reasonable examination or explanation as to what may be driving the difference in prices. [21] Respondents have divided their paints usage into five categories: thinner, lacquer, sealer, stain and glaze. For thinner, respondents provided values of between $1.18/kg-— $1.36/kg with a final determination valuation of $2.83/kg. The final determination valuation for laquer, sealer, stain and glaze was $4.48/kg each. For laquer, the respondents provided values, of between $2.11/kg — $3.89/ kg; for sealer $2.42/kg — $3.89/ kg; for stain $2.18/kg — $3.89/kg; and for glaze $2.99/kg— $4.03/kg. Pis.’ Br. 27. ' Respondents also provided a per liter price for the paint, but the court could not ascertain from the record before it the conversion factor for liters to kilograms for the various types of paint. [22] The court’s holding here is limited to finding that Commerce may reasonably determine that MSFTI represents the best available information for valuing paint. This holding, however, is not preclusive of Commerce finding a more narrow subheading (to the six or eight digit level) or a series of subheadings, to value paints, instead of the broad 4 digit heading of HTS[I] chosen in the final determination. [23] AFMC avers that Commerce's selection of a separate HTS[I] subheading for thinner was unsupported by substantial evidence and not in accordance with law, alleging that the aggregation of products into "thinner” when the thinner was not an independent factor of production, but rather was added to other finishes, was not appropriate. The court finds this argument to be without merit. Commerce was able to find a more specific match for the product reported to be "thinner,” and therefore it was reasonable for Commerce to use that valuation for that input. Cf. Guangdong Chems.,30 CIT at-, 414 F.Supp.2d at 1311. Therefore, Commerce's selection of a separate valuation for "thinner” is appropriate. [24] It is not clear whether or not Respondents are also advocating for the valuation of paint using either Infodrive India or Tarun Vadehra and Highland House data. To the extent Respondents are, the court finds that its analysis above, with respect to overinclusive or under-inclusive data and the select nature of Info-drive India, Tarun Vadehra and Highland House data, see supra at 32-33,44-46, applies equally here. [25] Following the commencement of this litigation, Commerce requested, and was granted, a voluntary remand to correct some flaws in its wage rate calculation. The determination under review is Commerce's results pursuant to this voluntary remand. This is the only aspect of Commerce’s determination that the court is considering after a remand. [26] This term is also known as the y-intercept. The regression model attempts to fit a line through various data points, plotted along the x and y axes (here the x axis representing GNI and the y axis representing the wage rate). The constant, or y-intercept, is the point at which the line predicted by the regression equation crosses the y-axis. See, e.g., Lawrence Hamilton, Data Analysis for Social Scientists 309 (Wadsworth Publishing Co.1996). [27] As Commerce noted in its Final Determination, and in its brief, Commerce is currently seeking comments on its calculation methodology for its NME Wage Rate Methodology. Expected Non-Market Economy Wages: Request for Comment on Calculation Methodology, 70 Fed.Reg. 37,761 (Dep't. Commerce June 30, 2005). While the court's holding infra is that Commerce is not required, by the statute, to limit its data set in its regression analysis to economically comparable countries, in the manner calculated by Respondents, nothing stated by the court here would preclude Commerce from considering Respondents' arguments in Commerce's reexamination of its NME wage rate methodology. [28] The initial data set that Commerce employed in this investigation had fifty-six countries. In the Remand Determination, Commerce determined that the data from two countries were not reliable. Remand Determination at 13. As a result, the data set employed in December 2004 consisted of only fifty-four countries. [29] Specifically Commerce stated: the Department agrees in part with Dorbest that a recalculation of the regression analysis may require the Department to expand the basket of countries it includes in its regression analysis. A review of the data shows, however, that it may be appropriate *1295 to include substantially more than the nineteen countries which Dorbest identifies. Issues & Decision Mem., P.R. Doc.1933 at 180 (Cmt.23). [30] I.e., WageIndia = 0.392 + .00048 * 470 = $0.6176. [31] Of the countries Commerce found economically comparable to the PRC, only one country has a higher wage rate than this baseline (without adding in the value of GNI*0.00048). Predicted Country_Wage rate_GNI Wage rate India_$31_470 $0.62 Pakistan_$.36 420_$0.59 Sri Lanka_$33_850_$0.80 Indonesia *_$.35_n/a_n/a Philippines_$.81_1030_$0,89 * Indonesia was not included within Commerce's regression model; this number was provided to Commerce by Respondents. All data came from, or was compiled using data obtained in, Exhibit III to Remand Determination at 3 & 5. For reference, the PRC's GNI for 2002 was $960. See Exhibit II to Remand Determination at 1. It is not clear to the court whether the source of this apparent distortion is in the underlying wage rate data set, in Commerce's calculation, or in the regression model itself. [32] In AFMC’s response to the court’s questions of July 13, 2006, AFMC pointed out that some changes to the data set or the regression model would create "distortions and results that are less reliable than Commerce's existing methodology." AFMC Resp. Court's July 13, 2006 Qs. 19. AFMC points to the fact that Commerce's data set of 56 countries leads to an adjusted R square of 0.92 while if only the LI [low-income] and LMI [lower-middle income] countries were used, the adjusted R square is only 0.47. Id. This argument ignores two crucial points. First, accepting all of AFMC's points as true, these are points addressing the reasonableness of the model, through an evaluation of the statistical soundness of the model, that Commerce would have to make to demonstrate reasonableness in the face of other discrepancies and distortions. Secondly, as to the point that restricting the data set to only low income and lower-income countries would reduce the adjusted R-squared, this does not address what the outcome would be were the data set to be expanded to include all countries that meet Commerce's criteria. Commerce is, of course, free, upon remand to consider these issues and respond appropriately. [33] For instance, Respondents have alleged that Commerce's data set appears to be heter-oscedastic (that is, the variance between the predicted data point and the actual data point varies over the data set). Respondents also allege that the appropriate means of correcting for such heteroscedasticity would be to employ a Generalized Least Squares Model instead of an Ordinary Least Squares Model in Commerce's regression analysis. The court, of course, does not know whether or not this would improve the accuracy of the model, but this is an example of a step Commerce could consider on remand. [34] Commerce made adjustments during ministerial error proceedings and during the voluntary remand. [35] Commerce obtains its wage rate data from the International Labor Organization, Yearbook of Labour Statistics; country-specific consumer price index and exchange rate data from the International Monetary Fund, International Financial Statistics; and country specific GNI data from the World Bank, World Development Indicators. See Remand Determination at 5. [36] These values are calculated as follows. Factory overhead includes such costs as the cost of machinery, spare parts, and rent. Commerce adds together all such costs, as expressed on a surrogate company's financial statement, to get the total overhead expenditure ("Overheads”); Commerce then divides the result by the surrogate firm’s material, labor, and energy costs ("MLES”). See, e.g., Memorandum from Jon Freed, Case Analyst, to File Re: Final Determination Financial Ratio Memorandum: Wooden Bedroom Furniture from the People’s Republic of China, P.R. Doc. 1931 at Attachs. 2-10. ("Financial Ratio Memo”). Finally, Commerce multiplies the result by the derived manufacturing cost of the product in question of the investigated firm ("MLEp”). The result is the overhead that may be allocated to the normal value of the merchandise in question ("Overheadp”). Stated mathematically: Overhead^ _* MLEp = Overheadp MLEs Next, Commerce adds the surrogate firm's MLE and Overhead (together "the cost of manufacturing”) and determines an amount for general expenses ("SG & As”) including, for example, expenses such as bank charges, travel expenses, and office supplies. See Magnesium Corp. of Am. v. United States, 166 F.3d 1364, 1371-72 (Fed.Cir.1999); Financial Ratio Memo, P.R. Doc.1931 at Attachs. 2-10. Commerce then calculates the ratios of the surrogate firms' SG & A to its cost of manufacturing and multiplies this ratio by the sum of MLEp and Overheadp; the result is the SG & A that may be allocated to the merchandise in question ("SG & Ap”). Stated mathematically: SG & As _*(MLEp + Overheadp)=S G & Ap MLES + Overheads Last, Commerce adds an amount for profit. Commerce initially calculates the surrogate company’s profit ratio which is the ratio of the surrogate company's before-tax profit ("profits”) over hie sum of MLES, Overheads, and SG & As. Financial Ratio Memo, P.R. Doc.1931 at Attachs. 2-10; Shanghai Foreign Trade Enters. Co., 28 CIT at -, 318 F.Supp.2d at 1341. Commerce then multiplies this result by the investigated company’s derived MLEp, Overheadp, and SG & Ap. The result is the profit that may be allocated to the merchandise in question ("profitp”). Stated mathematically: profits _*(ML MLES + Overheads + SG & As ip + Overheadp + SG & Ap)=profitp [37] AFMC also argues that Commerce “failed to treat the salaries ... as SG & A expenses rather than MLE expenses for DnD and Evergreen,” AFMC’s Br. Supp. Mot. J. Agency R. Re: Selection Surrogate Values & Calculation of Financial Ratios 26 ("AFMC Br.”), and that Commerce failed to categorize certain expenses of Jayaraja, Evergreen, Nizzamuddin, Swaran, and DnD, id. at 28-29. Because of the court's disposition of issues (2), (3) and (4) above, the court reserves judgment on these subsidiary issues until Commerce issues a Remand Determination. [38] The Amended Final Determination found the following SG & A ratios: SG & A SG & A Company Ratio_Company_Ratio IFP_24.38% Nizamuddin 31.51% Evergreen 6.90% Fusion Design 34,39% Akriti_13.53% Swaran_47.30% Jayaraja 16.61% DnD_15.66% Raghbir 10.44%_ Attach. 1 to Memorandum from Jon Freed, Case Analyst, to File, through Robert Bolling, Program Manager, Re: Amended Final Determination Financial Ratio Memorandum: Wooden Bedroom Furniture from the People's Republic of China, Dep't of Commerce (Dec. 27, 2004), P.R. Doc.2004 ("Amended Financial Ratio Memorandum."). Admittedly, because these numbers were announced for the first time in its Amended Final Determination, Commerce did not rely on them. Nevertheless, because Commerce amended its ratio calculations, either Commerce made an implicit finding that the changes did not upset its previous findings in its Final Determination or, alternatively, the matter must be remanded for Commerce to reconsider this evidence. See, e.g., SEC v. Chenery Corp., 318 U.S. 80, 94, 63 S.Ct. 454, 87 L.Ed. 626 (1943); cf. Borlem S.A. — Empreedimentos Industriais v. United States, 913 F.2d 933, 937 (Fed.Cir.1990). [39] In order for this evidence to have bite, Commerce would essentially have to establish that because Jayaraja's and Akriti's SG & A ratios do not appear (superficially speaking) to be related to production volume, there may not be, in general, a relationship between size and financial ratios for other furniture manufacturers. [40] Because India is a member of the Harmonized Tariff System, classifications under the HTS[I] are based on an international nomenclature common to many countries (including the United States). [41] The court assumes that the HTS[I] is identical to the HTS[US] up to the six digit level. As such, the Explanatory Notes to the Harmonized Commodity Description and Coding System ("Explanatory Notes”) for the HTS[US] (2006) provide illumination for the HTS[I], The Explanatory Notes for Chapter 39 state that “Primary forms,” which was a phrase employed by Dorbest in its suggested subheading, is defined as ''liquids and pastes ... [b]locks of irregular shape, lumps, powders. ...” Explanatory Note 6 to Chapter 39. Note 10 to Chapter 39 states that the phrase "plates, sheets, film, foil and strip applies only to plates, sheets, film, foil and strip ... and to blocks of regular geometric shape....” Explanatory Note 10 to Chapter 39. [42] In Polyethylene Retail Carrier Bag Comm. v. United States, the court noted that Commerce in the past has found the basket categories to be potentially unreliable if a more representative alternate surrogate is available. Polyethylene Retail Carrier Bag Comm. v. United States, 29 CIT -, -, 2005 WL 3555812, *18 (2005); See, e.g., Tetrahydrofur-furyl Alcohol from the People’s Republic China, 69 Fed.Reg. 3887, 3892 (Dep't Commerce Jan. 27, 2004) (notice of preliminary determination of sales at less than fair value); Freshwater Crawfish Tailmeat from the People’s Republic of China, 64 Fed.Reg. 27,961, 27,962 (Dep’t Commerce May 24, 1999) (final results of new shipper review) (“import data from basket categories can be too broad to be reliable.”). [43] Note E, to Section 8302 of the Explanatory Notes states: Mountings, fittings and similar articles suitable for furniture. This group includes: (1)Protective studs (with one or more points) for legs of furniture, etc.; metal decorative fittings; shelf adjusters for book-cases, etc.; fittings for cupboards, bedsteads, etc.; keyhole plates. (2) Corner braces, reinforcing plates, angles, etc. (3) Catches (including ball spring catches), bolts, fasteners, latches, etc. {other than key-operated bolts of heading 83.01). (4) Hasps and staples for chests, etc. (5) Handles and knobs, including those for locks or latches. Commodity Description and Coding System, 3 Explanatory Notes 1372 (3d ed.2002) (emphasis in original). [44] Section 1677m(d) provides, in relevant part: If the administering authority or the Commission determines that a response to a request for information under this subtitle does not comply with the request, the administering authority or the Commission (as the case may be) shall promptly inform the person submitting the response of the nature of the deficiency and shall, to the extent practicable, provide that person with an opportunity to remedy or explain the deficiency in light of the time limits established for the completion of investigations or reviews under this subtitle. [45] The reason for this discretion is self-evident. See, e.g., Henry J. Friendly, Indiscretion About Discretion, 31 Emory L.J. 747, 761 (1982) (in justifying deference to trial courts noting that "[t]he district judge must be master of how to get cases to trial, and has had opportunities for frequent observation of the offending counsel which would not emerge from a cold record.”). [46] The court notes that AFMC’s arguments are unclear. To the extent AFMC is arguing that Respondents did not provide an accurate description of their inputs, it is correct that only respondents are in possession of this information and have a duty to disclose such facts when so requested by Commerce. If a party fails to provide such facts, Commerce may apply adverse inferences. However, once Respondents supply these facts, Commerce must then apply the law to these facts. This process is analytical and, therefore, not subject to an application of adverse inferences. After all, Respondents are in no better position than any other party to classify these inputs. Inasmuch as AFMC is arguing that respondents’ difficulty in applying a third-country’s law triggers adverse inferences, they misunderstand the nature and role of adverse inferences in the process. The court notes that although classifying goods appears to be simple, it is quite complicated. Routinely, the Court of Appeals for the Federal Circuit remands Customs cases back to this court to develop the factual record (this of course after a protest proceeding before Customs and review by this court). Accordingly, if Commerce is going to use the HTS [I], it must (and did) accept the burden to ask the follow-up questions and conduct the analysis that such a choice will necessitate. [47] See supra note 44, p. 131.