Before its amendment in 1994, the antidumping statute directed that when a party to an antidumping proceeding failed to provide properly requested information, the Commerce Department was required to “use the best information otherwise available” in determining the appropriate antidumping duty rate. 19 U.S.C. § 1677e(e) (1988). Pursuant to that provision, the Commerce Department developed a protocol that allowed it to use the highest dumping margin established in any prior administrative review or in the less-than-fair-value proceeding as the “best information available” (or “BIA”) rate for an administrative review in which the foreign exporter did not comply with Commerce’s inquiry. The question presented in this case is whether, when the prior antidumping duty rate that Commerce relied upon to establish the BIA rate has been invalidated in the course of judicial review, Commerce may nonetheless continue to rely on that antidumping duty rate as the BIA rate in any pending administrative review proceeding. We hold that it is improper for Commerce to continue to use, as the BIA rate, an antidumping duty rate that has been vacated as erroneous.
I
Appellant D & L Supply Company, an American company, imported iron castings from the People’s Republic of China during the period relevant to this case. Plaintiff Guangdong Metals & Minerals Import & Export Corporation, a Chinese exporter, supplied iron castings to D & L during that period.
On March 19, 1986, the Department of Commerce found that iron eastings from China were being sold in the United States at less than fair value. After the International Trade Commission determined that those imports were causing injury to U.S. producers of similar castings, Commerce issued an anti-dumping order and set an antidumping duty rate of 11.66 percent for the Chinese exporters.
In the course of its annual administrative reviews of the antidumping order, Commerce adjusted the antidumping rates for iron castings several times. For 1987-88 Commerce calculated the rate to be 24.21 percent, and for 1988-89 Commerce calculated the rate to be 45.92 percent. In connection with the administrative review for 1989-90, Guangdong was the only Chinese manufacturer or exporter that responded to Commerce’s anti-dumping questionnaire. Commerce initially established a preliminary rate of 63.28 percent for Guangdong, but when it issued its final results for the 1989-90 administrative review, Commerce raised that rate to 92.74 percent. At the same time, Commerce declared that the 92.74 percent rate for Guangdong would also apply to all other Chinese manufacturers and exporters of iron castings.
While Commerce was still conducting the administrative review for 1989-90, it initiated the administrative review for 1990-91. In *1222connection with the 1990-91 review, however, no Chinese exporter, including Guangdong, responded to Commerce’s questionnaire. Commerce therefore resorted to its BIA procedures to determine the antidumping duty rates for that period. Basing the BIA rate on the highest antidumping duty rate from any prior administrative review, Commerce looked to what was then the rate for 1989-90 — 92.74 percent — and it set the 1990-91 BIA rate for all Chinese manufacturers and exporters at that level.
Commerce’s various decisions provoked litigation. First, D & L, Guangdong, and a group of other importers (U.V. International et al.) brought suit in the Court of International Trade contesting the final rate of 92.74 percent for 1989-90. Then, when Commerce selected the contested 92.74 percent rate as the BIA rate for 1990-91, D & L, Guangdong, and U.V. International et al. appealed that determination as well.
The Court of International Trade subsequently vacated the 92.74 percent rate for 1989-90 with instructions that Commerce recalculate the antidumping duty rate for that period. At the same time, the court remanded the 1990-91 BIA determination, directing Commerce to “reevaluate the situation [for 1990-91] and deem whether the rate from [the 1989-90 review] is still appropriate for use as BIA.” D & L Supply Co. v. United States, 841 F.Supp. 1312, 1314-15 (CIT 1993).
On remand, Commerce determined that it had overstated the antidumping duty rate for 1989-90, and it reduced the rate for that period from 92.74 percent to 31.05 percent (later revised to 31.51 percent). Commerce refused, however, to adjust the BIA rate for 1990-91, which had been based on the now-discredited 92.74 percent rate for 1989-90. D & L and the other plaintiffs again sought review of the 1990-91 BIA rate in the Court of International Trade, but the court upheld the 92.74 percent rate for that period. D & L and the other U.S. importers have appealed from that ruling.
II
On appeal, D & L and U.V. International et al. argue that Commerce acted unreasonably by refusing to revise the 1990-91 BIA rate notwithstanding Commerce’s recognition on remand that the 1989-90 rate on which the BIA rate was based was erroneous. Commerce and a group of domestic producers, Alhambra Foundry, Inc., et al., disagree. They argue that it was proper for Commerce to use 92.74 percent as the BIA rate, because that was the applicable antidumping duty rate for 1989-90 at the time Commerce calculated the 1990-91 BIA rate, and they contend that no rule of law requires Commerce to recalculate the BIA rate when the prior rate on which it was based is later held to be erroneous.
The BIA provision in effect at the time Commerce made its margin determinations in this case provided:
In making their determinations under this subtitle, the [Commerce Department] ... shall, whenever a party or any other person refuses or is unable to produce information requested in a timely manner and in the form requested, or otherwise significantly impedes an investigation, use the best information otherwise available.
19 U.S.C. § 1677e(c) (1988). The statute was amended in 1994 and now provides that under such circumstances Commerce shall “use the facts otherwise available in reaching the applicable determination under this subtitle.” 19 U.S.C. § 1677e(a) (1994). Pursuant to section 1677e, Commerce has adopted a two-tiered method for calculating antidumping rates based on BIA Commerce uses the first tier for respondents who fail to cooperate. To calculate the appropriate dumping rate for first-tier firms, Commerce selects the highest margin calculated for any party in the less-than-fair-value investigation or in any administrative review. Commerce uses the second tier for respondents who substantially cooperate but fail to provide the requested information in a timely manner or in the required form. To calculate the appropriate dumping rate for. second-tier firms, Commerce selects the higher of (1) the firm’s less-than-fair-value margin for the subject merchandise, or (2) the highest calculated rate in the current review for the class or kind of merchandise from the same country of origin. See Koyo Seiko Co. *1223v. United States, 92 F.3d 1162, 1167 (Fed.Cir. 1996). Guangdong was accorded first-tier BIA treatment in this case.
Stripped to its essentials, the argument made by Commerce and Alhambra is that there is nothing wrong with using a subsequently invalidated dumping margin as the basis for calculating the BIA rate, because the BIA rate is supposed to be sufficiently high to induce respondents to cooperate with Commerce’s dumping investigations. See Atlantic Sugar, Ltd. v. United States, 744 F.2d 1556, 1560 (Fed.Cir.1984). Because there is no way of knowing what Guangdong’s actual dumping margin was in 1990-91 (as a result of Guangdong’s failure to provide the information Commerce requested), Commerce argues that 92.74 percent cannot be regarded as an inaccurate BIA rate. In addition, Commerce argues that to require recalculation of a BIA rate whenever the prior dumping margin on which the BIA rate was based is overturned would render the process of calculating a BIA rate unduly cumbersome and would be contrary to the important interest in achieving finality for administrative review determinations.
These arguments, although based on valid general considerations, do not justify reliance on an invalidated prior antidumping duty rate as the “best information available.” If efficiency, finality, and the inducement of cooperation were the only goals served by the BIA procedure, the optimal BIA rate would be a rate at least high enough to ensure that an uncooperative respondent could not profitably export its goods. As this court has noted previously, however, “expedition is not the only value needed in the implementation of the [antidumping] statute; accuracy is also important.” Borlem S.A.-Empreedimentos Industriais v. United States, 913 F.2d 933, 939 (Fed.Cir.1990). Congress’s desire for expeditious determinations in trade matters “should not be interpreted as authorizing proceedings that are based on inaccurate data.” Id. at 937. Commerce’s refusal to revise its BIA rate on remand has resulted in what must be characterized as an inaccuracy in the ultimate rate derived, because the BIA rate is based on an antidumping duty rate that has been held to be grossly excessive. .
The procedure that Commerce has followed in this case is contrary to the clear import of the applicable statutory language. The pre-1994 version of section 1677e, which applies to this case, requires the use of the “best information ... available” to calculate the dumping margin when a party does not cooperate with Commerce. The statutory directive that Commerce use the best information available is intended to serve “the basic purpose of the statute — determining current margins as accurately as possible.” Rhone Poulenc, Inc., v. United States, 899 F.2d 1185, 1191 (Fed.Cir.1990); see also Allied-Signal Aerospace Co. v. United States, 996 F.2d 1185, 1190, 1191 (Fed.Cir.1993). Information that has conclusively been determined to be inaccurate does not qualify as the “best information” under any test, and certainly cannot be said to serve the “basic purpose” of promoting accuracy.
Basing the BIA rate on a calculated rate that is determined to be incorrect is also inconsistent with the policies that the two-tiered BIA scheme are supposed to serve. The purpose of using the highest prior antidumping duty rate as the BIA rate is twofold: it offers some assurance that the exporter will not benefit from refusing to provide information, and it produces an antidumping duty rate that bears some relationship to past practices in the industry in question. The rationale for selecting the highest prior antidumping duty rate, rather than some other prohibitively high rate, is the “common sense inference” that “the highest prior margins are the most indicative of current market conditions.” Allied-Signal Aerospace Co., 996 F.2d at 1192; see also Rhone Poulenc, Inc., 899 F.2d at 1190. While the highest prior margin is obviously not a precise indicator of current dumping practices, it provides at least some guidance as to the probable dumping margin in the period for which the exporter is not providing information, and it is preferable in that respect to an arbitrarily selected figure that has no pretension to accuracy. Commerce’s theory — that a prior rate can properly be used as the BIA rate even if *1224that rate is found to be grossly excessive— focuses entirely on the interest in inducing foreign exporters to cooperate with Commerce’s investigations, and it ignores the interest in selecting a rate that has some relationship to commercial practices in the particular industry. If the antidumping duty rate that Commerce uses as the basis for its BIA rate is seriously flawed — as was the case with the 92.74 percent rate at issue in this case — the rate ceases to have any utility as a rough index of conditions in the industry.
Commerce’s practice of refusing to adjust the BIA rate when the underlying calculated rate is changed does not even consistently serve the purpose of maximizing the pressure on foreign exporters to cooperate with Commerce’s investigations. The government advised us at oral arguinent that if the prior antidumping duty rate on which the BIA rate was based had been adjusted upward rather than downward, Commerce still would not have adjusted the BIA rate, but would have stuck with the prior rate that was in effect when it selected the BIA rate. In that setting, of course, Commerce’s policy would work to the advantage of uncooperative foreign exporters and would not serve the deterrent purpose that the government claims for it. In sum, we are persuaded that for a regulatory scheme that purports to base the BIA rate on actual dumping margins from prior years, reliance on margins that have been demonstrated to be inaccurate is irrational.
This is not to say that Commerce must wait until a particular antidumping duty rate has been judicially blessed or has otherwise become final before that rate can be used as the basis for calculating a BIA rate. A margin that has not yet been overturned is presumed to be accurate and can properly be used in the BIA determination. Thus, there was nothing wrong with using the 92.74 percent rate as the BIA in this case until that rate (as it applied to both Guangdong and, by extension, to all others subject to the anti-dumping order) was invalidated in the 1989-90 administrative review proceedings. Moreover, we do not hold that a BIA rate that has become final must be reopened if the prior rate on which it was based is subsequently overturned. Commerce could rationally consider interests in finality to preclude reopening BIA antidumping margins on such grounds once they have become final and no longer subject to further review. But when the dumping margin on which the BIA rate is based is invalidated before the BIA rate has become final, it is irrational to ignore the invalidity of the underlying rate and to uphold the BIA rate as purportedly based on the “best information available.”
We do not share the concern expressed by Commerce and the domestic producers that intolerable delays will result from prohibiting the use of invalidated rates as BIA in any administrative review that has not become final. As we have noted, nothing in our analysis would require Commerce to postpone selecting a BIA rate; the only constraint is that if the prior rate on which the BIA is based should be altered in the course of administrative or judicial review before the BIA rate becomes final, the BIA rate must be adjusted accordingly. The adjustment typically should involve no more than a simple change in the duty rate, not the sort of procedure that is apt to add significantly to the administrative burden on Commerce or the complexity of the administrative review process.
We therefore conclude that Commerce committed legal error when it refused to revise its BIA determination in this case, and we decline to defer to Commerce’s BIA procedures in that respect. See Borlem, 913 F.2d at 937 (noting that although this court has an “obligation to give substantial weight to an agency’s interpretation of a statute it administers,” it cannot “defer to an interpretation of the law where there are compelling indications that that interpretation is incorrect”). Accordingly, we hold that Commerce must revise the challenged 1990-91 BIA rate, which was based on a prior rate that was invalidated while the BIA determination was still under review.
REVERSED AND REMANDED.