specially concurring in part and dissenting in part:
I concur generally in parts I-V and VII-VIII of the majority’s opinion, but respectfully dissent from part VI. I would hold (a) that Payne was not a reasonably avail*386able source for at least some of the information sought by the IRS in this case, and (b) that it is always “necessary” for IRS Special Agents to identify themselves as members of the Criminal Investigative Division when conducting in-person third-party interviews.
I
I agree with the majority that the taxpayer under investigation might sometimes be an “otherwise reasonably available” source for at least some information sought by the IRS. The taxpayer might be able to provide exculpatory documents with independent indicia of reliability, such as bank records, cancelled checks, notarized copies of contracts, deeds, or credit card receipts that can help show that he has not committed a crime. These documents might alleviate the need for the IRS to contact at least some third parties in determining that a taxpayer has not cheated on his taxes.1
But I do not agree that Payne was a reasonably available source of information for all of the information sought by the IRS here. The majority overlooks this issue: it devotes much attention to the general question of whether taxpayers can ever be reasonably available sources of information, but little to the facts of this case. The majority’s only analysis of this issue consists in a citation to a supposed factual finding by the district court that Payne was a reasonably available source for the information sought. In fact, the district court made no such finding. Although the tenor of the district court’s opinion might in places imply that Payne was a reasonably available source of information, the court made no explicit finding to that effect.2
As the majority itself describes the requirements of suits based on 26 U.S.C. § 6103(k)(6), three inquiries are involved: (1) whether the information sought is related to enforcing the tax code, (2) whether the information was not “otherwise reasonably available,” and (3) whether the disclosure of return information was “necessary” in order to obtain the information sought. No one disputes that the first criterion is satisfied here. The district court and the majority both seem to skip over the second step, and proceed directly to a determination of “necessity.”
The questions of whether the information was “otherwise reasonably available” and whether the disclosures were “necessary” are “interdependent.” Barrett v. United States, 795 F.2d 446, 449 (1986) (hereinafter “Barrett I”). We are concerned in this case with the necessity of three types of disclosures by the IRS: the disclosure of the identity of the taxpayer, *387the disclosure that the taxpayer is under some kind of investigation, and the disclosure that the investigation into the taxpayer is criminal in nature. It is not possible to identify which of these three types of disclosures by Batista were “necessary” •without determining what information was “reasonably available” from Payne. If Payne was not a reasonably available source of some information, then it was okay for Batista to try to get that information from other sources. If it was lawful for Batista to contact third parties, it was also by that fact necessary for him to disclose two kinds of return information: the identity of the taxpayer (Payne) and the fact that the taxpayer was under investigation. Batista could hot solicit information about Payne without identifying him. And the very fact that the IRS is asking questions about someone will tip off the third party that the taxpayer is under some kind of investigation. Therefore, if Payne was not a reasonably available source of information, then the only unnecessary disclosure by Batista would be (in the majority’s view) the disclosure that the investigation was criminal in nature.
But if Payne was a reasonably available source of information, then all of the disclosures were unnecessary: the identity of the taxpayer, the fact of an investigation, and the fact that the investigation was criminal. If Payne was a reasonably available source of information, then Batista could have avoided all of the disclosures to third parties by securing the information from Payne himself.
In Barrett v. United States, 100 F.3d 35, 39-41 (1996) (Barrett III), we explained that the IRS is liable only for the damages occasioned by unlawful disclosures — if some disclosures were permitted, and others not, then the IRS has to pay damages only for the disclosures that were not permitted. Id. at 40. In Barrett III, the IRS had mailed a number of circular letters to patients seeking information about a plastic surgeon. We explained that, because the information sought was not reasonably available from the doctor, there was nothr ing wrong with the mere fact that the IRS sent the circular letters. That is, because the information was not reasonably available from the taxpayer, it was not unlawful for the IRS to disclose the identity of the taxpayer and the fact that he was under some kind of investigation. The only unlawful disclosure was the statement in the body of the letter that the taxpayer was under criminal investigation. Id. at 40 n. 6. The doctor was thus required to show that his damages were caused — not by the lawful disclosure of his identity and the fact of the investigation — but by the unlawful disclosure of the criminal nature of the investigation. Id. at 39-40. If the doctor’s medical practice was damaged because patients were worried about their privacy, about the fact of their having had plastic surgery becoming public during an investigation, then the damage would have been done just as effectively by the permitted disclosure that the doctor was under some kind of investigation as the impermissible one that the investigation was criminal. We therefore required the doctor to show that the patients were worried about the criminality of his behavior, not just their loss of privacy.
The same analysis applies here: it matters whether Payne was a reasonably available source of information so that we can correctly determine for which disclosures the IRS is liable. Under Barrett III, if Payne was a reasonably available source of the information sought, the IRS is liable for all of the damages Payne suffered as a result of the third party contacts. If Payne was not a reasonably available source of information, the IRS is liable only for so much of Payne’s damages as can be attributed to the disclosure that *388the investigation was criminal in nature. If the IRS’s disclosure that Payne was under criminal investigation resulted in no more damages than the lawful disclosure that Payne was under some kind of investigation, the IRS is not liable for any of Payne’s damages. It is therefore necessary to evaluate whether Payne was a reasonably available source of the information sought by Batista.
The majority’s opinion may be internally contradictory on this point. It seems to leave undisturbed a supposed finding by the district court that Payne was a reasonably available source of the information sought. But it then cites the analysis from Barrett III explained in the preceding paragraphs, and proceeds as if the only issue is the damages caused by revealing the criminal nature of the investigation: “we focus on the finding that Batista regularly revealed the criminal nature of his investigation of Payne when making third-party contacts” (emphasis in original). If none of the contacts were permissible at all, it does not matter whether the damages were caused by the clients’ concerns over privacy or over whether Payne was a tax cheat: any damage from the third-party contact is redressable.
Even assuming for the sake of argument that the district court opinion can be read as finding that Payne was a reasonably available source for all of the information sought, then this conclusion would be clearly erroneous with respect to at least some of the information. Admittedly, for some of the information, Payne might have been an otherwise reasonably available source. For example, Batista contacted the Texas Lawyers Insurance Exchange to find out about $36 in interest income reported by Payne. Batista apparently admitted on the stand that he could have gotten this information (presumably in a sufficiently probative form) from Payne himself. Batista also contacted one of Payne’s clients named Neal Talmadge. The sole information sought from this contact was apparently to confirm that Tal-madge was a client of Payne, even though Batista had already confirmed that fact from court records. In these situations, the information sought by Batista might very well have been otherwise available without contacting third parties and without disclosing any return information.
But for other types of information sought by Batista, Payne clearly was not an available source of the information. For example, Batista asked a number of people about Payne’s involvement with drug sales.3 Payne obviously could not have provided sufficiently probative information about his own involvement in illegal drug transactions. So I would find that Payne was not a reasonably available *389source for all of the information sought by Batista, and any district court finding to the contrary is clearly erroneous.
The district court did not make specific findings as to what information sought from third parties was reasonably available from Payne. These findings are necessary to a correct determination of the case, and I would remand to the district court for it to make these determinations in the first instance.
II
Assuming that Payne was not a reasonably available source for at least some of the information sought by Batista, the next issue is whether it was “necessary” for Batista to disclose during third-party interviews that he was a criminal investigator. Batista’s identification badge states, in large type, “Criminal Investigative Division;” Batista identified himself as a criminal investigator when talking to some potential witnesses; and Batista sometimes told third parties that he was conducting a criminal investigation of Payne. In Barrett v. United States, 51 F.3d 475, 478 (1995) (Barrett II), we held that the IRS needed to produce trial evidence of the necessity of disclosing the criminal nature of an investigation in the body of a circular letter. In the absence of evidence to the contrary, the district court was obligated to find that the disclosure was not necessary. Barrett III concerned mitten disclosures in a circular letter, not oral disclosures in the course of an in-person interview, as are at issue here. The majority nevertheless feels that Barrett II controls and that the IRS was obliged to present trial evidence of the necessity of Batista’s disclosures. I disagree.
We interpret statutes according to the plain meaning of the words used. United States v. Ron Pair Enters., Inc., 489 U.S. 235, 242, 109 S.Ct. 1026, 103 L.Ed.2d 290 (1989). There are two senses in which the word “necessary” is commonly used. “Necessary” sometimes means “strictly essential,” or an indispensable precondition for the happening of an event. “Necessary” also sometimes means “appropriate or helpful.” Cf., e.g., McCulloch v. Maryland, 4 Wheat. 316, 17 U.S. 316, 414, 4 L.Ed. 579 (1819) (interpreting “necessary” in the Constitution’s Necessary and Proper Clause to mean the latter of these two definitions, appropriate or helpful).
The first of these definitions has a strictly empirical character: if there is empirically any possible way that an event can happen without the occurrence of the condition, the condition is not necessary. The second of these definitions has both empirical and non-empirical elements. Empirically, we want to know whether or not something is “helpful.” For example, are people really more likely to be cooperative or forthcoming in responding to a circular letter if the letter discloses the criminal nature of an investigation? Is the disclosure really “helpful” in getting meaningful replies? But there is also an element of this definition of necessary that looks to custom and usage: is the disclosure “appropriate” according to customary standards for handling such situations?
“Necessary” in § 6103 must mean the latter definition — “appropriate or helpful” — not “strictly essential.” In legal contexts, we rarely use “necessary” in the latter sense. See, e.g., Comm’r v. Heininger, 320 U.S. 467, 471, 64 S.Ct. 249, 88 L.Ed. 171 (referring to the “commonly accepted meaning” of “necessary” as “appropriate and helpful”). If we were to interpret “necessary” to mean “strictly essential,” the statute would be completely unworkable. If this were what were meant by “necessary” in the statute, the IRS could never show that a disclosure *390was necessary. All that the plaintiff would have to show is that there was some possible way — however infeasible, however impractical, and however otherwise difficult — that could have induced the witness’s cooperation without the disclosure.
Congress could not have intended to set such a high barrier. Section 6103 sets up a general rule that return information should be kept confidential. But it then sets out specific circumstances under which return information can be revealed during an investigation. It would not have made sense to have a section creating an exception from the confidentiality rules if the standard for disclosure were so high the exception could never be used. If Congress had intended to prevent the IRS from ever disclosing return information, even during investigations, it just would have left the confidentiality rule in place and not created any exceptions.
Taking “necessary” in the statute to mean “appropriate or helpful,” Barrett II does not foreclose us from holding that it is necessary for an agent to identify himself truthfully as a criminal investigator when contacting third parties. Barrett II focused on the empirical aspects of the word “necessary”: whether it really, factually, empirically produces more helpful and forthright responses to have a circular letter say that the investigation is criminal than not. There is no way to answer this question meaningfully without empirical evidence: the testimony of experienced officers, results from various circular letters that both did and did not include the disclosure, and so forth. In the absence of such evidence, the Barrett II court was not prepared to conclude that the disclosure was necessary. Significantly, for circular letters, the other aspect of the word “necessary,” the non-empirical aspect that focuses on custom and usage, is not really applicable. There are not any well-developed customs for how the police should solicit information from people in letters.
When it comes to in-person interviews by the police, however, the non-empirical aspect of the word “necessary” becomes very important: we are now concerned not only with whether such disclosures are “helpful” but also with whether they are “appropriate.” There is a strong expectation that when a police officer shows up at your door to ask you questions that he identify himself, state the agency for which he works, and show his identification. Cf. Wilson v. Arkansas, 514 U.S. 927, 115 S.Ct. 1914, 131 L.Ed.2d 976 (1995) (detailing the common law tradition of requiring the police to identify themselves and “signify the cause of [their] coming” before entering a private dwelling, and incorporating this tradition into the Fourth Amendment inquiry into the “reasonableness” of a search). A regular and commonly expected part of the procedure of being approached by the police is the showing of identification. We do not need “evidence” to help us conclude that it is appropriate for IRS Special Agents to identify themselves and show their badges when conducting interviews: unlike Barrett II, the issue is not solely empirical here, it also concerns the customs and usages of the community. Cf. Dickerson v. United States, 530 U.S. 428, 443, 120 S.Ct. 2326, 147 L.Ed.2d 405 (2000) (giving importance to the place of Miranda warnings in custom and social expectation, and noting “Miranda has become embedded in routine police practice to the point where the warnings have become part of our national culture”). In passing § 6103, Congress could not have intended to prevent IRS agents from complying with the widespread expectation that law enforcement agents show their badges and otherwise identify themselves to the people whom they interview.
*391For circular letters, there was not any relevant custom or usage, so the question of whether the disclosure was necessary was solely an empirical question. Trial evidence of necessity was therefore required. Here, there is a relevant custom, so it is not solely an empirical question. Given social expectations about police identifying themselves, I do not think that empirical evidence would be helpful in ascertaining whether it is appropriate for Special Agents to show their badges or otherwise identify themselves when conducting third party interviews. Barrett II is therefore distinguishable, and we should conclude that it is necessary, in the sense of appropriate, for the tax police to identify themselves when approaching third parties about a tax investigation.
The district court and the majority also premise the IRS’s liability in part on alleged ivritten disclosures made in summonses and letters of the criminal nature of the investigation. The district court found that “Batista issued a large number of administrative summonses and letters to third-parties which disclosed on their face that Payne was under criminal investigation by the IRS.” If Batista had disclosed the criminal nature of the investigation in body of letters or summonses, then Barrett II would control, and the IRS would be liable for those disclosures in the absence of trial evidence of their necessity.
But I would find that this factual finding by the district court was clearly erroneous. Batista’s letters and summonses are part of the exhibits in the record. These documents do not, in fact, disclose on their face the criminal nature of the investigation.4 All of the litigants seem to argue as if, and the majority reasons as if, the only issue is the necessity of Batista’s oral disclosures, in apparent recognition of the fact that the written documents do not contain any disclosure of the criminal nature of the investigation. To the extent that the district court premised liability on the written documents, I would reverse.
Ill
I agree with the majority that we should remand for the district court to re-evaluate the good faith issue in light of Gandy v. United States, 234 F.3d 281 (5th Cir.2000).5 I also agree that we should dis*392miss Payne’s cross-appeal, challenging the district court’s award of damages, as premature.
The majority does not make clear that the IRS, like Payne, may also raise its challenges to the district court’s damages and attorneys’ fees awards in a subsequent appeal, if a subsequent appeal proves necessary. The IRS argues that Payne did not present sufficient evidence of causation of actual damages and that the district court awarded punitive damages for actions that were not disclosures and were not unlawful. The IRS also argues that its litigating position was “substantially justified,” which would preclude the district court’s award of attorneys’ fees. If the district court finds that Batista acted in good faith, the IRS would not be liable for any damages, and we would not need to consider these issues. I assume that the majority omits mention of these claims by the IRS because the issues might be avoided by the district court’s decision on remand. I do not read the majority opinion as precluding the IRS from raising these challenges in the future if they prove necessary. With that understanding, I concur in all of the majority’s opinion but part VI.
. If the IRS determines that the taxpayer has committed a crime, it may need to obtain even these documents from third parties so that it can introduce them as evidence at trial. See, e.g., Fed. R. Evid. 802(6) & advisory committee’s note (requiring that the foundation for introducing business records ordinarily be laid by the “testimony of the custodian” of the records).
. For example, the district court found that Batista contacted third parties "without first determining whether the information was otherwise reasonably available” and that the disclosures to third parties were not necessary in light of the fact that "the information repeatedly offered by Payne was ultimately delivered to ... the United States Department of Justice.” Of course, finding that Batista made no effort to determine whether the information was reasonably available from Payne is not the same thing as finding that the information was in fact available from Payne. And the district court did not make any findings as to the content of the information eventually delivered to the Department of Justice and whether this information was the same or different from that sought by Batista from third parties.
. The district court apparently believed that any disclosure designed to obtain information about Payne's involvement with drugs was not "necessary,” regardless of whether the information was otherwise reasonably available, because Batista had "no rational basis” for thinking that Batista had sold illegal drugs. Our precedents preclude this line of reasoning. In Barrett I we explained that, in an action under §§ 7431 and 6103, we "do not question the right, wisdom, or necessity of a particular IRS investigation.” Barrett I, 795 F.2d at 451. The question posed by the statute is not whether the information sought is necessary, but whether the disclosures were necessary to obtain the information sought. Id. As long as the information relates to enforcing the tax laws, the IRS can investigate any crime and look for any information it so desires. If the IRS wanted to investigate Payne for unreported income from drug sales, then § 6103 poses no bar to its doing so, regardless of whether the IRS had a "rational basis,” "probable cause,” “articulable suspicion,” or any other justification for its investigation. DiAndre v. United States, 968 F.2d 1049, 1053 (10th Cir.1992) ("[S]ection 6103 does not provide a vehicle to test the probable cause or any other level of justification to investigate.”).
. Batista used a standard IRS form for the summonses. The body of the summons states:
You are hereby summoned and required to appear before David Batista or his desig-nee[J an officer of the Internal Revenue Service, to give testimony and to bring with you and to produce for examination the following books, records, papers, and other data relating to the tax liability or the collection of tax liability or for the purpose of inquiring into any offense connected with the administration or enforcement of the internal revenue laws concerning the person identified above for the periods shown. This statement does not disclose what kind of investigation the IRS is- conducting. The summonses were sometimes accompanied by attachments, listing the documents sought, which also did not disclose the criminal nature of the investigation. Batista's letters sometimes stated that he was conducting an investigation of Payne, but they did not state that the investigation was criminal.
. I note one way in which our approach to this case differs from the approach taken by the Gandy court: we decide the question of whether Batista’s oral disclosures of the criminal nature of the investigation were "necessary” before moving on to any discussion or consideration of whether the disclosures were in good faith. Gandy, like this case, concerned oral disclosures by IRS agents of the criminal nature of an investigation. The Gan-dy court decided the good faith issue first, and thereby avoided the issue of whether the disclosures were necessary. It reasoned that "we need not decide the difficult legal question of whether agents McPherson and Sander’s oral disclosures that Gandy was under criminal investigation were necessary if ... [they were acting] in good faith.” Id. at 285. The Gandy court did not consider it necessary *392to evaluate the “necessity'' of the disclosures before evaluating the good faith of the agents.
The Gandy court’s approach misunderstands the nature of the good faith defense. We have explained that the good faith defense resembles in many respects the qualified immunity for executive officials described in Harlow v. Fitzgerald, 457 U.S. 800, 102 S.Ct. 2727, 73 L.Ed.2d 396 (1982). See Huckaby v. United States, 794 F.2d 1041, 1048 (5th Cir.1986) ("pT]he good-faith [sz'c] defense in section 7431(b) should be judged by an objective standard analogous to that employed in Harlow.")', Gandy, 234 F.3d at 285. In Harlow, the Supreme Court explained that judges applying the qualified immunity test should first determine "the currently applicable law” and only then determine “whether the law was clearly established at the time an action occurred.” Harlow, 457 U.S. at 818, 102 S.Ct. 2727. The reason for this sequence is that we do not want to allow executive officials to violate the law indefinitely. We do not punish officials who make close and difficult, but wrong, decisions about the legality of their decisions. But, the first time a lawsuit is brought challenging a particular kind of official conduct, the courts are supposed to clarify the applicable state of the law so that officials do not make the same mistake in the future. The Harlow court therefore required that judges determine the "currently applicable law” before determining whether the law was "clearly established” at the time of the violations. Id. The same reasoning would seem to apply under § 7431: courts should first determine whether the IRS agents acted lawfully, and only then whether they acted in good faith.
The statute clearly contemplates that courts should follow this sequence. It permits a defense only for actions relying on "good faith, but erroneous ” interpretations of applicable law. There is no way to know whether an officer’s interpretation was "erroneous” without evaluating the current state of law. Although I joined in the opinion in Gandy, I acknowledge that our approach in that case was mistaken. In the future, I would have district courts and panels follow the lead set by the majority here — evaluating first the actual state of the law, and then any good faith defense — rather than the approach followed in Gandy.