TARANTO, Circuit Judge.
The Commerce Department determined that a Vietnamese manufacturer of wind towers was selling its products in the United States at about 51.5% below normal value, a figure that Commerce calculated using methods made applicable by statute when imported goods come from a non-market economy, as the wind towers at issue here do. The company challenges three aspects of Commerce’s calculation upheld by the Court of International Trade: Commerce’s selection of data to determine the weight of the manufacturer’s products; Commerce’s presumption-based premise that the company’s supplier received subsidies from the Korean government; and Commerce’s calculation of certain overhead expenses for inclusion in the base of costs that go into normal value. We reverse as to Commerce’s weight calculation; affirm as to Commerce’s treatment of Korean subsidies; and vacate and remand as to Commerce’s overhead-expense calculation.
BACKGROUND
A
In December 2011, the Wind Tower Trade Coalition petitioned the Department of Commerce to impose antidumping duties under 19 U.S.C. § 1673
et seq.
on wind towers imported into the United States from Vietnam.
See
Utility Scale Wind Towers From the People’s Republic of China and the Socialist Republic of Vietnam, 77 Fed. Reg. 3,440, 3,440 (Dep’t of Commerce Jan. 24, 2012). The Coalition alleged that such imported towers were being sold in the United States at less than fair value.
Id.
The Commerce Department conducted an investigation into whether, in particular, CS Wind Vietnam was engaged in such dumping.
As relevant here, CS Wind produces wind towers in Vietnam and ships them in sections to the United States, where they are assembled and erected. J.A. 73; Utility Scale Wind Towers from China
&
Vietnam, Inv. No. 701-TA-486, 2013 WL 1155424, at *5 (USITC Feb. 2013).
As part of its investigation, Commerce calculated the “normal value,”
ie.,
the price at which the product is sold or offered for sale in the exporting country. 19 U.S.C. § 1677b(a)(l)(B). If the normal value exceeds the price at which the product is sold in the United States, and other required findings- are made, Commerce is to make a finding of dumping and impose a duty based on the difference — the dumping margin.
Id.
§§ 1673, 1677(35)(A);
see Dorbest Ltd. v. United States,
604 F.3d 1363, 1367 (Fed. Cir. 2010);
Ningbo Dafa Chem. Fiber Co., Ltd. v. United States,
580 F.3d 1247, 1250 (Fed. Cir. 2009). In percentage terms based on export prices for the towers shipped to the U.S., Commerce calculated a 51.5% “weighted average dumping margin” for CS Wind. 19 U.S.C. § 1677(35)(B); Utility Scale Wind Towers From the Socialist Republic of Vietnam: Final Determination of Sales at Less Than Fair Value, 77 Fed. Reg. 75,984, 75,988 (Dep’t of Commerce Dec. 26, 2012)
(2012 Final
Determination).
It is undisputed here that Vietnam has a “nonmarket economy,” in which prices “do not reflect the fair value of the merchandise.” 19 U.S.C. § 1677(18)(A). Because CS Wind operated in a nonmarket economy, Commerce calculated the normal value pursuant to 19 U.S.C. § 1677b(c)’s special rules for such economies.
See Dorbest,
604 F.3d at 1367 (describing such rules). Under those rules, Commerce was to calculate a normal value for the wind towers based on the “value of
the factors of production utilized in producing the merchandise” plus “an amount for general expenses and profit plus the cost of containers, coverings, and other expenses”; and “the valuation of the factors of production” was to 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)(l). Commerce was to use, “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.”
Id.
§ 1677b(c)(4). The object, under that approach, is “to construct a hypothetical normal value for the merchandise that is uninfluenced by the nonmarket economy.”
Jiaxing Brother Fastener Co. v. United States,
822 F.3d 1289, 1292 (Fed. Cir. 2016);
see Downhole Pipe & Equip., L.P. v. United States,
776 F.3d 1369, 1375 (Fed. Cir. 2015). Here, Commerce used “surrogate values” from India, a market economy, to calculate values for various elements of the normal value of CS Wind’s towers. J.A. 1298-99.
Three of Commerce’s determinations are pertinent to the present appeal. First, after translating certain Indian prices into U.S. dollars per kilogram, Commerce had to multiply that per-kilogram price by the weight (in kilograms) of the CS Wind components. In arriving at the weight of CS Wind’s products, Commerce decided not to use the weights CS Wind reported for its various factors of production. Instead, it used the weights indicated on certain packing slips prepared by one of CS Wind’s customers for the necessary transocean shipping of the sections of the towers. J.A. 126-27. CS Wind challenges that decision.
Second, for certain components,
i.e.,
flanges, welding wire, and wire flux inputs, CS Wind asked Commerce to use the actual prices CS Wind paid for them when buying them from a manufacturer in Korea — a market economy. Commerce denied the request. Based on a determination in earlier proceedings that certain goods exported from Korea are eligible for subsidies, Commerce presumed that CS Wind’s purchases benefited from such subsidies, and it then found that CS Wind had provided insufficient evidence to rebut the presumption. J.A. 65-68. Commerce therefore used Indian surrogate values for the prices of those components, rather than the prices CS Wind actually paid. CS Wind challenges that decision.
Third, Commerce used the financial statements of an Indian company, Ganges International, which sells identical and comparable wind towers, J.A. 50, to calculate the required contribution to normal value from, in particular, “general expenses,” 19 U.S.C. § 1677b(c)(l) — here, overhead, selling, general, and administrative expenses (which, following Commerce’s usage, we call “overhead” for short). J.A. 220-21. One of the Ganges-reported expense line items that Commerce included in overhead was “Jobwork Charges (including Erection and Civil Expenses).” J.A. 737, 204. The following meanings of those terms are not disputed before us. Firm A incurs “jobwork” expenses when it pays Firm B to provide manufacturing
services
for A (presumably, therefore, working with raw materials A has supplied to B), with the resulting manufactured goods then transferred to A for A to sell. In the wind-tower setting, “erection and civil” expenses are payments for preparing the foundation on which to set a tower (“civil”) and for setting up the tower on the foundation (“erection”) — which we infer are payments to outsiders where, as here, they are listed as “inelud[ed]” in “jobwork.”
See
CS Wind Br. 49 (quoting J.A. 1712); U.S. Br. 7 n.2; J.A. 221, 740.
(We may refer to the two activities together as “tower setup.”)
CS Wind asked Commerce to reduce the 212,380,751 rupee figure for those jobwork expenses by certain income amounts for what CS Wind alleges are corresponding items, namely, “Erection income” (90,856,-566 rupees) and “Civil income” (51,931,347 rupees) — totaling 142,787,913 rupees— which Ganges reported as income separate from the income from its “Sales.” J.A. 733. CS Wind’s request would have resulted in including only 69,592,838 rupees of Job-work Charges in overhead (212,380,751 minus 142,787,913), but Commerce denied the request. Instead, it reduced the “Jobwork Charges (including Erection and Civil Expenses)” only by the tiny amount (2,085,-029 rupees) of Ganges-reported income for “Sales — Jobwork,” J.A. 736.
Commerce thus included 210,295,722 rupees of “Job-work Charges (including Erection and Civil Expenses)” as overhead in calculating normal value.
See
J.A. 221-22.
In response to CS Wind’s challenge to that decision in the present litigation, Commerce eventually adopted a different approach to deciding what amount of the “Jobwork Charges (including Erection and Civil Expenses)” to include in overhead. In some but not all of its descriptions, Commerce characterized its new approach as trying, like CS Wind’s approach, to exclude from overhead costs the portion of the “Jobwork Charges (including Erection and Civil Expenses)” line item that were tied to erection and civil
income.
J.A. 161, 175. But whereas CS Wind did so by simply subtracting the “Erection income” and “Civil income” amounts, Commerce sought to achieve a similar goal by a more complex ratio calculation, using certain aspects of the income and expense sides of the financial statements. Commerce’s final approach reduced the 212,380,751 rupees of “Jobwork Charges (including Erection and Civil Expenses)” by 8.62%. J.A. 209. The result was to include more than 194,000,-000 rupees from the Jobwork Charges line item as overhead, far more than the roughly 70,000,000 rupees CS Wind urged. CS Wind challenges Commerce’s final approach to this aspect of the normal-value calculation.
B
In August of 2012, Commerce published a preliminary determination that CS Wind had engaged in dumping. Utility Scale Wind Towers From the Socialist Republic of Vietnam: Preliminary Determination, 77 Fed. Reg. 46,058, 46,058 (Dep’t- of Commerce Aug. 2, 2012). In December of that year it made certain modifications and made its dumping determination final.
2012 Final Determination, supra.
After the International Trade Commission determined under 19 U.S.C. § 1673d(b) that a U.S. industry was materially injured or threatened with material injury by imports of wind towers from Vietnam, Commerce published an amended final determination, correcting ministerial errors in its 2012 final determination. Utility Scale Wind Towers From the Socialist Republic of Vietnam: Amended Final Determination of Sales at Less Than Fair Value and Anti-
dumping Duty Order, 78 Fed. Reg. 11,150 (Dep’t of Commerce Feb. 15, 2013).
CS Wind filed an action in the Court of International Trade challenging Commerce’s determination under 19 U.S.C. § 1516a(a)(2) & (b)(1)(B)© and 28 U.S.C. § 1581(c). As relevant here, the Court of International Trade on March 27, 2014, affirmed in part and remanded in part. J.A. 100-40. It affirmed Commerce’s use of the packing-weight figures rather than the component weights CS Wind reported. J.A. 126-27. It also affirmed Commerce’s determination not to use the Korean prices for certain components. J.A. 137. But it remanded on the issue of jobwork expenses, holding that if Commerce, using the Ganges financial statements, were to account for the reported jobwork charges including erection and civil expenses, it must also account for the reported erection income and civil income. J.A. 123-24. In particular, the court ruled that the two must be “treated similarly under Commerce’s practice, either including both as overhead or excluding both from the calculation, unless Commerce explains why different treatment is warranted.” J.A. 124.
On remand, Commerce issued a determination on July 29, 2014. J.A. 145-81. It abandoned its initial inclusion of all job-work charges, including civil and erection expenses, in favor of a new approach that Commerce here describes as attempting to exclude “the proportion of the jobwork expenses relating to erection and civil activities, so that jobwork expenses and associated income were treated consistently.” U.S. Br. 10;
see
J.A. 160-62, 175-78. On review, the Court of International Trade on November 3, 2014, again remanded. J.A. 183-99. It concluded that “Commerce is still treating expense and income line items differently without stating an acceptable reason,” and it required “recalculation or further explanation.” J.A. 196-97. On January 20, 2015, Commerce issued its
Final Redetermination,
fundamentally following its initial redetermination approach but making some modifications. J.A. 203-16. This time, the Court of International Trade affirmed, producing a final judgment. J.A. 218-34.
We have jurisdiction to hear CS Wind’s appeal under 28 U.S.C. § 1295(a)(5).
DISCUSSION
We begin with CS Wind’s challenge to the calculation of the weight of its products. In determining the weight of the CS Wind products (to be multiplied by the surrogate per-kilogram values), Commerce had available two sources of information: CS Wind’s listing of the components of the wind towers and their weights, produced to and verified by Commerce during the investigation, J.A. 124-25, 317-43; and packing slips containing customer-supplied (not manufacturer-supplied) weight estimates for tower sections to provide center-of-gravity information for the transocean shipping, J.A. 835,
see
J.A. 56-57. Commerce chose to use the weights reflected on the packing slips, J.A. 55-57, which were higher than the weights CS Wind reported for the components, J.A. 1341. The choice of higher weights increased the calculated “normal value” and, therefore, the dumping margin and the duty.
Commerce does not dispute that, in this decision, it was seeking to use the best available evidence for an accurate assessment.
See Shakeproof Assembly Components, Div. of Illinois Tool Works, Inc. v. United States,
268 F.3d 1376, 1382 (Fed. Cir. 2001) (“In determining the valuation of the factors of production, the critical question is whether the methodology used by Commerce is based on the best available information and establishes antidump-ing margins as accurately as possible.”);
see Ningbo Dafa,
580 F.3d at 1257 (same). Commerce necessarily decided, therefore, that the packing lists provided more accurate information about weight than did CS Wind’s records. The question before us is whether that determination is supported by substantial evidence. 19 U.S.C. § 1516a(b)(l)(B)(I);
see F.lli De Ceceo Di Filippo Fara, S. Martino S.p.A. v. United States,
216 F.3d 1027, 1031 (Fed. Cir. 2000).
Substantial evidence “means such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.”
Universal Camera Corp. v. N.L.R.B.,
340 U.S. 474, 477, 71 S.Ct. 456, 95 L.Ed. 456 (1951). “The substantiality of evidence must take into account whatever in the record fairly detracts from its weight.”
Gerald Metals, Inc. v. United States,
132 F.3d 716, 720 (Fed. Cir. 1997) (internal quotations omitted, alterations in original). Here, the question is whether the record supplies a basis for Commerce reasonably to find that the packing-weight information was more accurate than the CS Wind component-weight information. We conclude that Commerce has not provided a sufficient basis for using the packing weights rather than the component weights reported by CS Wind.
CS Wind documented the sources, including commercial invoices, for the weights it reported to Commerce, J.A. 822, 824, 827, and Commerce verified those figures to the extent it deemed necessary (making certain adjustments).
On the other hand, Commerce “acknowledge[d] that the packed weights are based on certain estimations,” J.A. 58; U.S. Br. 21 (“It is undisputed that Packed Weight is an estimated weight.”), made by customers, not the manufacturer. “Commerce determined that the total Packed Weight of a section is based on center-of-gravity calculations provided by CS Wind’s customers for purposes of optimally positioning the wind tower section on the shipping vessel to maintain correct balance.” U.S. Br. 18, citing J.A. 57, 835.
In nevertheless choosing the customer-estimated figures over the manufacturer-reported, Commerce-verified ones, Commerce gave what amounts to a single reason — which, we conclude, lacks the eviden-tiary support that would be required in order for it to justify choosing the packing weights. Commerce stated that it was “unreasonable to assume that the weight of the wind tower section recorded in the packing lists is
so grossly overestimated
as to chance the misplacement of the wind tower section ón a shipping vessel and risk an imbalance of the vessel or rolling of the tower section in transit.” J.A. 57 (emphasis added; footnote citing J.A. 1341 omitted);
see
J.A. 58 (“considering the importance of the use of the packed weight for shipping purposes, it is not unreasonable to assume that the packed weights and the [component] weights should be similar”). In this court, Commerce confirms what the italicized phrase indicates: the basis of Commerce’s choice was the
“material extent
of the discrepancy between the two weights” (the packing weights. and the component weights). U.S. Br. 22 (emphasis added). Underscoring the centrality to Commerce’s rationale of the size of the weight discrepancy, Commerce states three times that the packed-list weight for “the internal components” was “nearly double” CS
Wind’s reported weight for those components. U.S. Br. 13,18, 22.
The problem with that basis grows out of the fact that the doubling is only of a very small fraction — the “internal components” — of the overall tower weights. As the Commerce-cited J.A. 1341 indicates, the entire weight discrepancy between the CS Wind figures and the packing-list figures lies in the internal components, and that discrepancy as a percent of the weight of the overall towers is less than 4%. But there is no evidence that either (a) a mere 4% difference in overall weight or (b) the specific difference in weight figures for the small internal-components portion of the towers would make a difference in maintaining balance on the vessels used for transportation here. And we have no basis for thinking that either premise is a matter of common knowledge or otherwise can be presumed true without evidence. In the absence of such evidence, there is no reasonable basis for Commerce’s conclusion that it should assume that the packing-list weight is more accurate because the shipping-balance purpose demanded the assumption.
Because the reason Commerce offers for using the packed weights is without record support, we find Commerce’s choice to be unsupported by substantial evidence. We therefore reverse the Court of International Trade’s affirmance of that choice and direct Commerce to use the manufacturer-reported weights in its calculation.
B .
We turn next to the issue of Korean subsidies. CS Wind purchased three categories of components from a supplier in Korea and exported those components to Vietnam. J.A. 65. Under the statute, if Commerce determines that “broadly available export subsidies existed” with respect to such a foreign purchase, Commerce may “disregard” the presumably subsidized prices, using surrogate values to calculate normal value instead. 19 U.S.C. § 1677b(c)(5). Here, Commerce relied on previous determinations to find that Korea maintains “broadly available, non-industry-specific export subsidies.” J.A. 65, 1510 n.3. Commerce ultimately relied on that basis to reject use of CS Wind’s Korean purchase prices and use surrogate values instead. J.A. 66. In this court, CS Wind has not challenged Commerce’s conclusion that export subsidies are generally available, so we simply accept that conclusion here, without reviewing its basis. CS Wind Br. 33-40. But CS Wind contends that the evidence required Commerce to find that no subsidies affected CS Wind’s particular purchases and therefore to use the actual prices CS Wind paid for those items, not surrogate values for those items, in the calculation of normal value. The Korean prices are lower than the surrogate values, so using them would lower the “normal value” and- hence the dumping margin and resulting duties.
Commerce relied on the generally available Korean subsidies to reject use of the Korean prices, concluding that it had a reasonable basis to believe or suspect that CS Wind’s purchases of flanges, welding wire, and wire flux benefited from such subsidies and that CS Wind did not persuasively show there was in fact no such benefit. J.A. 65-68. Nothing in the statute precludes that approach to choosing whether to use surrogate values or particular market purchases here. Indeed, the statute states that Commerce “may disregard price or cost values without further investigation if the administering authority has determined that broadly available export subsidies existed.” 19 U.S.C. § 1677b(c)(5);
see also
H.R. Rep. No. 100-576, at 590-91 (1988) (Conf. Rep.) (“In valuing [the factors of production], Commerce shall avoid using any prices which it has reason to believe or suspect may be
dumped or subsidized prices. However, the conferees do not intend for Commerce to conduct a formal investigation to ensure that such prices are not dumped or subsidized, but rather intend that Commerce base its decision on information generally available to it at that time.”). And we have been given no basis on which to conclude that this approach, essentially a presumption-based approach, is an unreasonable way of implementing the statute.
Thus, with CS Wind not challenging Commerce’s finding that subsidies were available for export transactions in Korea, Commerce reasonably required CS Wind to demonstrate that it received no subsidies for the particular purchases in question.
See Hangzhou Spring Washer Co. v. United States,
387 F.Supp.2d 1236, 1248 (Ct. Int’l Trade 2005);
Luoyang Bearing Corp. v. United States,
347 F.Supp.2d 1326, 1342 (Ct. Int’l Trade 2004). We conclude that substantial evidence supports Commerce’s conclusion that CS Wind did not make its case.
CS Wind has relied on the following bases for its contention that its purchases did not benefit from the Korean export subsidies: a statement from a Finance Manager stating that CS Wind itself had not received subsidies on the purchase in question, J.A. 673; emails from two individuals at two vendors — one a “Deputy General Manager/Team Leader, Sales
&
Marketing team,” the other a Product Manager — both emails indicating that the vendors did not apply for or receive export subsidies, J.A. 675, 677; and the contention that the initial purchase of the components at issue here was made by CS Wind Corporation, a Korean company that owns CS Wind Vietnam (to which they were then shipped), and the Korean parent was not eligible for any export subsidies. CS Wind Br. 38-41.
Commerce could reasonably reject CS Wind’s case as not comprehensive or definitive enough, at least in light of other evidence, to show that subsidies did not affect the purchases, at issue. The generally available subsidies were for exports. Commerce cites several Certificates of Origin that, while listing CS Wind Vietnam as the “Consignee,” list CS Wind’s supplier as the “Exporter,” which therefore could well have taken advantage of export subsidies. A box labeled “Declaration by the Exporter” is signed by the supplier, not by CS Wind Vietnam. And several invoices list CS Wind Corp. in Korea, not CS Wind Vietnam, as the “Shipper/Exporter.” The parent corporation could have taken advantage of the export subsidies. An invoice and a certificate of inspection, though showing the product as being purchased by CS Wind’s Korean parent, referred to the “MidAmerican (Vietnam)” project, perhaps indicating that the manufacturer was aware that the product was destined for Vietnam. And “Certificate^] of Material” from the supplier list the customer not as CS Wind Corp. in Korea, but as CS Wind, Ltd.,
i.e.,
the Vietnamese company.
This issue required a judgment about evidence. Commerce reasonably made that judgment, finding that CS Wind did not demonstrate that the purchases at issue were unaffected by the generally available export subsidies in Korea. Commerce could therefore choose to use surrogate values for those components of the wind towers, rather than the prices of the Korean purchases.
C
Finally, we consider CS Wind’s challenge to Commerce’s determination of how much of the “Jobwork Charges (including Erection and Civil Expenses)” line item on the Ganges financial statements to include as overhead expenses. CS Wind challenges (a) Commerce’s rejection of its proposal simply to subtract from the amount of that
expense line item the income line items for “Erection income” and “Civil income” and (b) Commerce’s ultimate adoption instead of a complicated alternative approach. We conclude that a further remand is needed, because Commerce has failed to meet its obligation to set forth a comprehensible and satisfactory justification for its approach as a reasonable implementation of statutory directives supported by substantial evidence.
Under the review provision invoked by the parties, we are obliged to set aside Commerce’s determination if it is “unsupported by substantial evidence on the record[] or otherwise not in accordance with law.” 19 U.S.C. § 1516a(b)(l)(B)(I). To fulfill that obligation, we insist that Commerce “examine the record and articulate a satisfactory explanation for its action.”
Yangzhou Bestpak Gifts & Crafts Co., Ltd. v. United States,
716 F.3d 1370, 1378 (Fed. Cir. 2013). Although we uphold “a decision of less than ideal clarity if the agency’s path may reasonably be discerned,”
Bowman Transp., Inc. v. Arkansas-Best Freight Sys., Inc.,
419 U.S. 281, 286, 95 S.Ct. 438, 42 L.Ed.2d 447 (1974), the required explanation must reasonably tie the determination under review to the governing statutory standard and to the record evidence by indicating what statutory interpretations the agency is adopting and what facts the agency is finding. Such an explanation enables us to fulfill our review function and also to avoid making choices reserved to the agency,
i.e.,
to avoid violating the principle of
SEC v. Chenery,
318 U.S. 80, 88, 63 S.Ct. 454, 87 L.Ed. 626 (1943), under which a reviewing court “may not affirm on a basis containing any element of discretion — including discretion to find facts and interpret statutory ambiguities — that is not the basis the agency used, since that would remove the discretionary judgment from the agency to the court.”
ICC v. Brotherhood of Locomotive Eng’rs,
482 U.S. 270, 283, 107 S.Ct. 2360, 96 L.Ed.2d 222 (1987).
In another case in which we remanded to Commerce for a more satisfactory explanation, we explained that it is
necessary for Commerce to explain the factual settings for the calculations at issue, and explain exactly how those calculations are made. The antidumping statute is highly complex and often confusing, and we accordingly rely on Commerce in its antidumping determinations to make sense of that statute. The more complex the statute, the greater the obligation on the agency to explain its position with clarity. If the Court of International Trade and this court are to play their statutorily required roles in reviewing Commerce’s determinations, it is important that we have clear guidance from Commerce as to what is actually happening. [¶] Once Commerce explains its actual methodology for the calculation of constructed value profit, it should explain why its methodology comports with the statute.
SKF USA, Inc. v. United States,
263 F.3d 1369, 1382-83 (Fed. Cir. 2001);
see Pension Benefit Guar. Corp. v. LTV Corp.,
496 U.S. 633, 654, 110 S.Ct. 2668, 110 L.Ed.2d 579 (1990) (reading precedent as “mandating that an agency take whatever steps it needs to provide an explanation that will enable the court to evaluate the agency’s rationale at the time of decision”);
Burlington Truck Lines, Inc. v. United States,
371 U.S. 156, 167-68, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962);
see also NMB Singapore Ltd. v. United States,
557 F.3d 1316, 1320 (Fed. Cir. 2009);
Timken U.S. Corp. v. United States,
421 F.3d 1350, 1354-57 (Fed. Cir. 2005) (discussing 19 U.S.C. § 1677f(I)).
Two aspects of this requirement are worthy of particular note here. First, an agency’s “experience and expertise” (U.S. Br. 45) presumably enable the agency to provide the required explanation, but they do not substitute for the explanation, any more than an expert-witness’s credentials substitute for the substantive requirements applicable to the expert’s testimony under Fed. R. Evid. 702,
see Carnegie Mellon Univ. v. Marvell Technology Group, Ltd.,
807 F.3d 1283, 1302-03 (Fed. Cir. 2015). The requirement of explanation presumes the expertise and experience of the agency and still demands an adequate explanation in the particular matter.
See Burlington Truck Lines,
371 U.S. at 167-68, 83 S.Ct. 239. Second, an agency’s statement of what it “normally” does or has done before
(e.g.,
J.A. 160, 1422) is not, by itself, an explanation of “why its methodology comports with the statute.”
SKF USA,
263 F.3d at 1383. Whether it does so in a particular agency decision or in a cited earlier decision, the agency must ground such a normal or past practice in the statutory standard.
In this case, Commerce has not provided the needed explanation setting forth the interpretations and evidence-based factual findings .that establish the required connection from statute to determination. We remand for Commerce to provide that A-to-Z explanation. Here we identify some of the uncertainties that we are left with upon reading what Commerce has said so far. We do not intend this recitation to be comprehensive or to suggest the absence of simple ways of resolving them. The task on remand is for Commerce to lay out a reasoned grounding, in the statute and evidence, for whatever choice it ends up making about what portion of “Jobwork Charges (including Erection and Civil Expenses)” to include in overhead in calculating the normal value of the wind towers at issue.
We begin with the legal source of the authority Commerce is exercising in imposing duties on the imports here, based on calculations of normal value, 19 U.S.C. § 1677b(c). Commerce does not appear to dispute that the statute should be interpreted in accordance with a simple core idea: expenses should be included in calculating normal value for the merchandise at issue only to the extent one would expect a fair sales price for that merchandise to be set to recoup such expenses, so that expenses separately recouped by income other than receipts from selling that merchandise should not be built into the “normal .value” of the merchandise.
If Commerce has a different statutory interpretation, it should articulate and justify it on remand. The analysis of what expenses should be included in overhead should then be carefully justified in terms of the adopted statutory interpretation.
As to the particular expenses at issue here: CS Wind’s consistently promoted option is simply to subtract the erection and civil
income
from the expense line item that includes erection and civil
expenses.
As we currently understand the matter, one possible scenario supporting that position would be the following: Ganges does
essentially no tower setup through its own employees but hires subcontractors for all such work (erection and civil activities), pays the subcontractors (incurring erection and civil expenses), and then charges its tower customers for such setup (receiving erection and civil income). If that were an accurate description of how Ganges' conducts its business, the case for CS Wind’s subtraction approach, with a possible small adjustment, would seemingly be strong. Ganges would be getting paid for all of its tower-setup expenses separately from what it receives in selling the towers at issue, so one would not expect that a fair sales price of the towers would be set to recoup those expenses. And the tower-setup
income
amount would correspond to the
expense
amount being recouped separately from the tower-sale price (with a possible adjustment to account for, say, a contractor’s markup by Ganges).
Aspects of the Ganges financial statements are relevant to the likelihood of that scenario or some variant. As to separate billing, Ganges states that “Erection of Steel Structures is recognised on completion of individual erection activity & Civil contracts are recognised on Percentage of completion method.” J.A. 740. As to subcontracting, the wording of “Jobwork Charges (including Erection and Civil Expenses)” carries a strong implication on its face: if jobwork charges are payments to outsiders, as appears undisputed, the “included]” erection and civil expenses would seem to be payments to outsiders as well, as appellee Wind Tower Trade Coalition argued to Commerce. J.A. 1539 (“ ‘Job-work [Charges] (including Erection and Civil Expense)’ reasonably only includes payments to third party contractors for their labor.”). That does not necessarily mean that
all
of the tower-setup
income
is for subcontracted-out activity, but the rupee amounts are consistent with that scenario: the roughly 140,000,000 rupees of erection and civil income could represent a pass-through (plus markup) of a substantial share of the roughly 212,000,000 rupees of jobwork expenses Ganges incurred (say, roughly 127,000,000 plus a 10% markup). In addition, as Commerce has noted, J.A. 207; U.S. Br. 46, Ganges says in its financial statement that it “is primarily engaged in Manufacturing
&
Trading activities and geographically operating in Domes[ti]c as well as export market,” J.A. 741 — which is at least consistent with the idea that Ganges subcontracts out all tower-setup work and does not perform any itself.
The methodology Commerce ultimately settled on seems to reject any scenario in which more than a small fraction of the Ganges tower-setup income represents outsourced tower-setup work (Commerce reduced the 212,380,751 rupee expense item by 18,307,221 rupees, about 13% of the tower-setup income of 142,787,913 rupees). If Commerce thinks practically all subcontractor-cost-pass-through scenarios for tower setup are unlikely to match reality for Ganges, we are uncertain as to precisely why. If Commerce thinks that it simply cannot tell how much tower-setup work Ganges does in-house rather than through subcontractors, it would seem relevant to weigh uncertainties about the role Ganges plays in tower setup against any uncertainties seemingly built into the more complicated methodology Commerce adopted.
Under that methodology, Commerce undertook to include as overhead all but some tower-setup-related percentage of the 212,380,751 rupees listed for “Jobwork Charges (including Erection and Civil Expenses).” To determine what percentage to exclude from the 212,380,751 rupee expense figure, Commerce turned its attention to the income side of the Ganges financial statements. It determined “all of the income items reflected in Ganges’ fi
nancial statements that can reasonably be associated with jobwork” and “the percentage that erection/civil income represents” of that amount,
ie.,
it calculated a ratio of erection/civil income to “the total income that can be reasonably associated with job-work.” J.A. 161. As to what income items belong in the denominator of that income ratio, Commerce finally settled on three items in addition to erection and civil income themselves (90,856,566 rupees and 51,931,347 rupees): sales of jobwork (2,085,029 rupees), sales of finished goods (1,440,021,110 rupees), and (what we understand to be sales of) scrap (62,484,632 rupees). J.A. 206-08, 733, 736. The dominant figure in that income ratio plainly is the sale of finished goods.
Commerce then adjusted the income ratio in a way that made essentially no difference in the resulting ratio. It stated that, except for the (tiny) sales-of-jobwork item, the income from all the just-mentioned items not only could be “associated with jobwork” but also “relates to raw materials and direct labor.” J.A. 209. On that basis, Commerce reduced each included income item (except sales of jobwork) in a way that sought to exclude “the amount of revenues associated with raw materials and labor.” J.A. 215. To do that, Commerce turned back to the expense side of the Ganges financial statements and calculated the ratio of “the sum of raw materials and direct labor expenses” to “the sum of Ganges’ raw materials, direct labor, energy, and manufacturing overhead expenses.” J.A. 215. That ratio was 82.03%. Commerce then used the residue of that number — namely, 17.97% — and multiplied each item in the income ratio except for sales of jobwork, both numerator and denominator, by 17.97%.
See
J.A. 215. Because the sales-of-jobwork figure is so small, and every other figure in the income ratio was multiplied by the same number, this barely changed the resulting income ratio — which became 8.62%. Commerce then multiplied 8.62% by the 212,380,751 rupees listed for “Jobwork Charges (including Erection and Civil Expenses)” and subtracted that amount (18,307,221 rupees) from 212,380,751 rupees. J.A. 215. That calculation produces a final figure of just over 194,000,000 rupees from that line item to include as overhead expenses in the calculation of normal value.
Commerce’s explanation for its approach is not satisfactory. Commerce has not clearly explained the logic, in terms keyed to the statute, of turning to the income side of the financial statements and using a (particular) ratio of certain income items as a way of apportioning the expense item at issue, “Jobwork Charges (including Erection and Civil Expenses).” Moreover, when saying what it was doing, Commerce said that it was trying to discern the amount of the Jobwork Charges tied to, variously, erection and civil “income,” “activities,” or “expenses.”
E.g.,
J.A. 161, 175, 176, 205, 207. The differing formulations can refer to Ganges-performed tower setup, Ganges-purchased tower setup, or both, but Commerce was not clear in making those distinctions. Commerce was likewise unclear when it referred to identifying income items that “can be reasonably associated with jobwork,” J.A. 161, 205, since it did not say whether “jobwork” referred to Ganges-performed services sold to others or Ganges-purchased services bought from others. Because Commerce is not clear in each “erection and
civil” and “jobwork” reference who is doing, selling, and buying what, it is difficult to follow Commerce’s logic.
We are uncertain, too, about the justification for using a single loose standard asking what “can be reasonably associated with” jobwork (for Ganges or perhaps by Ganges) in making all-or-nothing decisions about whether to build certain categories of income into the income ratio. The two main “sales” items (finished goods and scrap) and the tower-setup income items might well differ greatly in how much they realistically involve (someone’s) jobwork. Our uncertainty extends to the justification for the expense-based ratio that Commerce calculated to multiply against all but one item in the income ratio before arriving at the income ratio (used finally to reduce the Jobwork Charges line item). More explanation is needed not only of why that particular expense-based ratio serves to capture some proper portion of the items in the income ratio, but also of why it is proper to use the same expense-based ratio for all such items.
We have been illustrative, not exhaustive, in identifying our concerns. On remand, Commerce’s task is not to provide isolated responses to our concerns. It is to provide a coherent, full explanation of a final overhead determination, laying out and justifying each step so that not only are our concerns addressed but, more broadly, we may see how the ultimate result is grounded in a justified statutory interpretation and the evidence of record.
We note here a particular legal issue that warrants more attention from Commerce. In this proceeding, faced with the task of trying to interpret the financial statements of non-party Ganges, Commerce said that it “cannot go behind” those statements, U.S. Br. 42; J.A. 206, 175 — that it “will only seek information from within the surrogate financial statements,” J.A. 175. We understand Commerce to mean that it lacks authority even to ask Ganges for information, even if Ganges is free to decline to provide the requested information and no matter how important, simple, or objective the information might be. On remand, Commerce should set out a legal justification for that stated constraint on its question-asking authority or correct our understanding of what Commerce has said it cannot do. If Commerce concludes that it does have authority to make inquiries of Ganges, Commerce should explain why it is reasonable to refrain from making them, generally or in this particular matter. In so ordering, we are seeking explanations, not prejudging their legal soundness.
We reiterate that we are not here prescribing the proper overhead-expense calculation, generally or in this matter. Nor are we ruling that Commerce’s current result is incorrect — that it cannot be properly justified. We are remanding because we conclude that Commerce has not explained its determination sufficiently to allow us to conduct the judicial review to
which CS Wind is entitled to ensure that the agency’s exercise of power adheres to the authorizing law and respects the record evidence.
Conolusion
We reverse the Court of International Trade’s affirmance of Commerce’s use of packing weights rather than component weights in its calculation of surrogate values. We affirm the Court of International Trade’s affirmance of Commerce’s determination not to use Korean purchase prices for flanges, welding wire, and wire flux. We vacate the Court of International Trade’s affirmance of Commerce’s overhead determination with respect to job-work charges, erection expenses, and civil expenses. We direct the Court of International Trade to remand the matter regarding the overhead issue for Commerce to proceed in accordance with this opinion.
No costs.
REVERSED IN PART, AFFIRMED IN PART, AND VACATED AND REMANDED IN PART