This case is now on appeal for the third time. TFWS, Inc., a large liquor retailer in Maryland, is suing the State Comptroller of Maryland, seeking a declaration that certain Maryland statutes and regulations governing the wholesale pricing of liquor and wine violate the Sherman Act. The Comptroller asserts that the Twenty-first Amendment shields the Maryland regime from federal antitrust scrutiny. We have already concluded that the regulations violate the Sherman Act, and the remaining issue is whether the Comptroller has a valid Twenty-first Amendment defense. The last time this case was before us, we reversed the district court’s order awarding summary judgment to the Comptroller, an order based on the district court’s conclusion that Maryland’s Twenty-first Amendment interest in promoting temperance outweighs the federal interest in promoting competition under the Sherman Act. We concluded that summary judgment was inappropriate because there existed disputed factual issues about the effectiveness of the Maryland regulations in promoting temperance. On remand the district court held a bench trial and awarded judgment to TFWS after finding that the regulations do not promote temperance because they do not raise liquor and wine prices in Maryland. (This result would leave Maryland without a Twenty-first Amendment interest.) The district court’s finding that the challenged regulations do not raise liquor and wine prices in Maryland is based on a comparison of prices in Maryland and Delaware. Because the district court failed to take into account whether the difference in the two states’ excise tax rates affects the price comparison analysis, we cannot conclude *332that the district court’s determination is free of clear error. We therefore vacate the award of judgment to TFWS and once again remand for further proceedings.
I.
The two challenged Maryland liquor regulations are explained in some detail in our first opinion, TFWS, Inc. v. Schaefer, 242 F.3d 198, 202-03 (4th Cir.2001) (TFWS I), so we will provide only a brief summary here. The first regulation, the post- and-hold regulation, establishes how and when liquor wholesalers may alter their prices. See Md. Ann.Code art. 2B, § 12-103(c). The second regulation, the volume discount ban regulation, requires a wholesaler to offer every retailer the same price for a particular product. Md. Ann.Code art. 2B, § 12-102(a). One effect is that wholesalers cannot offer discounts to larger retailers for purchasing large volumes because discounts of any kind are prohibited. Id.
In TFWS I we affirmed the district court’s determination that both regulations violate federal antitrust law because they constitute per se violations of § 1 of the Sherman Act. 242 F.3d at 210. We reversed, however, the district court’s dismissal of TFWS’s complaint on Twenty-first Amendment grounds. The district court had determined on its own motion that despite their anti-competitive effect the regulations were nonetheless valid under the liquor control powers reserved to the states under the Twenty-first Amendment. Because neither side had an opportunity to address the Twenty-first Amendment issue, we vacated the order of dismissal and remanded the case. We provided the following instructions to the district court:
On remand Maryland should be given the opportunity to assert and substantiate its Twenty-first Amendment defense, and TFWS should be permitted to respond. The analysis the district court should undertake in analyzing Maryland’s interest and then balancing it against the federal interest is straightforward. First, the court should examine the expressed state interest and the closeness of that interest to those protected by the Twenty-first Amendment. We acknowledge that little analysis is needed on this point. Temperance is the avowed goal of the Maryland regulatory scheme, and the Twenty-first Amendment definitely allows a state to promote temperance. Second, the court should examine whether, and to what extent, the regulatory scheme serves its stated purpose in promoting temperance. Simply put, is the scheme effective? Again, the answer to this question may ultimately rest upon findings and conclusions having a largely factual component. Finally, the court should balance the state’s interest in temperance (to the extent that interest is actually furthered by the regulatory scheme) against the federal interest in promoting competition under the Sherman Act.
TFWS I, 242 F.3d at 213 (internal quotation marks and citation omitted).
On the first remand both sides moved for summary judgment after discovery. The district court awarded summary judgment to the Comptroller, concluding that (1) the Maryland regulations were effective in promoting temperance and (2) Maryland’s interest in promoting temperance outweighed the federal interest in promoting competition. We reversed because “[t]he district court arrived at its conclusion that the Maryland regulations were effective in promoting temperance by weighing conflicting evidence” at the summary judgment stage. TFWS, Inc. v. Schaefer, 325 F.3d 234, 241 (4th Cir.2003) (TFWS II). Because “a district court *333may not resolve conflicts in the evidence on summary judgment motions,” we vacated the order awarding summary judgment and remanded for trial on “the question of whether, and to what extent, Maryland’s regulatory scheme is effective in promoting temperance.” Id. at 241^42 (internal quotation marks omitted).
On remand the district court conducted a bench trial, at which the Comptroller sought to prove in two steps that the regulations are effective in promoting temperance. First, the Comptroller attempted to establish that the challenged regulations increase retail liquor and wine prices in Maryland. Second, the Comptroller attempted to establish that the higher prices are effective in reducing consumption of liquor and wine in Maryland. The district court concluded, after considering all of the evidence, that the challenged regulations “do not increase Maryland liquor prices,” and, as a result, the regulations are not effective in promoting temperance. J.A. 1249. Because Maryland has no Twenty-first Amendment interest in the regulations, the district court reasoned, there is nothing to balance against the federal interest reflected in the Sherman Act. The district court enjoined the Comptroller from enforcing the regulations, and the Comptroller now appeals.
II.
The Comptroller argues that the district court erred in finding that the challenged regulations do not raise liquor and wine prices in Maryland. We review for clear error the factual findings of a district court sitting without a jury. Fed.R.Civ.P. 52(a). “A finding is ‘clearly erroneous’ when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.” United States v. U.S. Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948). “If upon ... review, we think that the findings of the judge below were clearly erroneous, i.e. that he misapprehended the evidence or went against the clear weight thereof, it is our duty to say so and reverse the decision.” United States v. One 1955 Mercury Sedan, 242 F.2d 429, 430 (4th Cir.1957). This deference to the district court’s factual findings applies even when they “do not rest on credibility determinations, but are based instead on physical or documentary evidence.” Anderson v. City of Bessemer City, 470 U.S. 564, 574, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985).
In finding that the challenged regulations do not raise liquor and wine prices in Maryland, the district court relied exclusively on evidence comparing liquor and wine prices in Maryland to liquor and wine prices in Delaware, a state that repealed comparable regulations over a decade ago. The evidence generally showed that Maryland prices are lower than (or at least the same as) Delaware prices, which indicates that the challenged regulations do not raise prices in Maryland and thus do not promote temperance. The district court relied heavily on two of TFWS’s exhibits, Exhibits 93 and 94, which compare Maryland wholesale case prices to Delaware wholesale case prices. (Wholesale prices are presumably indicative of retail prices.) Exhibit 93 compares the lowest prices in 2003 for 2,637 liquor and wine products. Prices were based on data provided by TFWS’s stores in Maryland and Delaware. Exhibit 93 shows that when quantity discounts were taken into account in Delaware, the lowest Maryland price was lower than the lowest Delaware price for 54 percent of the surveyed products. Exhibit 94 compares the lowest prices of forty liquor and wine products over a seven-month period. Prices were taken from the Dela*334ware Beverage Monthly and the Maryland Beverage Journal. Exhibit 94 shows that when quantity discounts were taken into account, the lowest Maryland price was lower than the lowest Delaware price for 67.5 percent of the surveyed products. The district court also relied on anecdotal evidence that Maryland retail prices are widely known to be lower than Delaware retail prices and that many Delaware residents cross into Maryland to buy their liquor and wine.
According to the Comptroller, the district court’s reliance on Exhibits 93 and 94 and the anecdotal evidence is misplaced because this evidence fails to take into account the difference in excise taxes imposed by Maryland and Delaware. The Comptroller asserts, as he did before the district court, that any comparison of prices must adjust for excise taxes because the significant difference in excise taxes distorts the results (and thus the probative value) of the comparisons.
We conclude that the district court’s finding that the regulations do not raise prices in Maryland is open to clear error because the court’s analysis is incomplete. The district court is correct that there is ample evidence in the record that Maryland wholesale and retail prices are the same as, or lower than, Delaware wholesale and retail prices. But as the district court itself pointed out, “both parties used Delaware liquor prices as a control group to gauge the effect of the regulations on Maryland liquor prices. ” J.A. 1246 (emphasis added). Put another way, the whole purpose of comparing Maryland and Delaware prices is to determine whether the challenged regulations actually have the effect of raising prices in Maryland. Excise taxes imposed on wholesalers, the parties agree, affect wholesale and retail prices. However, Maryland currently imposes one of the nation’s lowest wholesale excise tax rates whereas Delaware currently imposes a relatively high wholesale excise tax rate. In Maryland a $.40/gallon tax is imposed on wine and a $1.50/gallon tax is imposed on liquor. In Delaware a $.97/gallon tax is imposed on wine, a $3.75/gallon tax is imposed on liquor of twenty-five proof or greater, and a $2.50/gallon tax is imposed on liquor of less than twenty-five proof. See Federation of Tax Administrators, State Liquor Excise Tax Rates, http://www.taxadmin.org/fta/rate/liquor.html (Jan. 1, 2005); Federation of Tax Administrators, State Wine Excise Tax Rates, http://www.taxadmin.org/fta/rate/wine.html (Jan. 1, 2005). This may indicate that in order “to gauge the effect of the regulations on Maryland liquor prices,” J.A. 1246, price comparisons should be adjusted for excise taxes in order to control for the difference in excise taxes between the two states. Otherwise, the fact that Maryland prices are the same as, or lower than, Delaware prices may tell us nothing about the effect of the challenged regulations on Maryland prices. Any difference in prices could be due to excise taxes, the challenged regulations, or both.
The potential significance of excise taxes is illustrated by the following example involving a case (12 bottles) of 1.75 liter bottles of Stolichnaya 80 Vodka. According to TFWS’s Exhibit 94, the lowest published price in Maryland for the relevant seven-month period was $158.35 and the lowest Delaware price, taking into account quantity discounts, was $161.94. Thus, according to Exhibit 94, a case of Stolichnaya costs $3.59 less in Maryland and this fact supposedly indicates that the challenged regulations do not raise prices in Maryland. However, a Delaware wholesaler *335would be required to pay $20.811 in excise taxes on its case whereas a Maryland wholesaler would be required to pay only $8.322 on its case. Delaware thus imposes $12.49 more in excise taxes on a case of 1.75 liter bottles of Stolichnaya than does Maryland, and the $3.59 difference reflected in Exhibit 94 could very well be due to the difference in excise taxes. At this point, we do not know the cause of the price difference because there is nothing in the record nor any analysis by the district court that explains the extent of the impact that excise taxes have on prices or why a comparison of prices should not adjust for excise taxes.
On appeal TFWS responds to the Comptroller’s argument that the price comparisons should be adjusted for excise taxes by asserting that “[t]he price that matters to the consumer is the retail price, which includes excise taxes. It makes no sense to remove those taxes from a state-to-state comparison.” Br. for Appellee at 42. TFWS’s argument, however, does not adequately address the issue of whether a comparison of Delaware and Maryland prices must control for excise taxes in order to be indicative of the challenged regulations’ effect on Maryland prices.
We have examined all of the evidence in this case, and we are “left with the definite and firm conviction that a mistake has been committed,” U.S. Gypsum Co., 333 U.S. at 395, 68 S.Ct. 525, because the district court did not consider the issue of whether the price comparisons should be adjusted for excise taxes. We cannot conclude that the current record supports the district court’s finding that the challenged regulations do not raise liquor and wine prices in Maryland. In other words, we cannot say at this point that the price comparisons cited by the district court are sufficiently reliable to provide an indication of the effect the challenged regulations have on Maryland prices. There is no evidence in the record that explains the extent of the impact excise taxes have on prices or why a comparison of prices need not be adjusted for excise taxes. Moreover, there is no evidence showing that when excise taxes are subtracted, prices in Maryland are still lower (and thus the challenged regulations do not raise prices). In fact, we have found only evidence to the contrary, that is, evidence showing that when excise taxes are subtracted from Maryland and Delaware wholesale prices, prices in Maryland are actually higher than those in Delaware (which would appear to indicate that the challenged regulations raise prices in Maryland). Other than the above-mentioned price comparisons that include excise taxes, there is no evidence that establishes that the challenged regulations do not raise prices in Maryland. In sum, the district court’s analysis is not yet complete, and at this stage we cannot conclude that its finding that the regulations do not raise liquor and wine prices in Maryland is free of clear error. We therefore vacate the district court’s order entering judgment in favor of TFWS and remand for further proceedings.
On remand the district court should first address the issue of whether excise taxes need to be taken into account when comparing Maryland and Delaware prices. *336The district court, of course, may exercise its usual discretion in determining whether, or to what extent, the record should be reopened as a result of the excise tax issue. If the court concludes that a comparison of prices should not be adjusted for excise taxes, the court should provide its reasons, and it may then reaffirm its earlier result. If the court concludes, however, that the prices should be adjusted for excise taxes, then the court must determine the extent to which excise taxes should be taken into account. The Comptroller asserts that wholesale and retail prices reflect the entire excise tax imposed by the states and that adjusting prices requires subtracting the entire tax. However, it is unclear to us whether wholesale and retail prices reflect the entire excise tax imposed, as it is possible that the prices reflect only a portion of the excise tax. We note that the final exhibits submitted by the Comptroller, Exhibits 113 and 113A, show higher prices in Maryland when excise taxes are subtracted, but it is for the district court to determine in the first instance whether this evidence is sufficient to support a finding that the challenged regulations raise prices. Moreover, given the timing of the Comptroller’s submission of these exhibits, it may be that TFWS should have an opportunity to offer a response to these exhibits. If the price comparison evidence ultimately shows that the challenged regulations do raise Maryland prices, then the district court should determine whether the increase in prices affects consumption in that state. That determination will control whether the district court must proceed further under our instructions in TFWS I, 242 F.3d at 213.
VACATED AND REMANDED
. There are 3.785 liters in a gallon. Thus a case of 1.75 liter bottles contains 5.548 gallons of liquor ((12 x 1.75) -5- 3.785). Assuming that Stolichnaya 80 Vodka is greater than twenty-five proof, and thus that the higher Delaware rate of $3.75/gallon applies, the excise tax on a case of Stolichnaya 80 Vodka is $20.81 (5.548 gallons times $3.75/gallon).
. Applying the current Maryland tax rate results in an excise tax of $8.32 (5.548 gallons times $1.50/gallon).