dissenting in part and concurring in part.
The majority finds no clear error regarding either of the two district court rulings challenged by Davenport on appeal. Because I disagree with the majority’s analysis and resolution of the first issue (drug quantity), but not of the second (firearm possession), I dissent in part and concur in part.
Gary Davenport pled guilty to three offenses: (1) possession of five grams or more of methamphetamine, (2) attempted manufacture of a detectable amount of methamphetamine, and (3) failure to appear for trial while out on bond. Regarding his sentence, Davenport’s plea agreement recommended holding him responsible for 6.9 grams of methamphetamine, which translated into an offense level of 25 and a recommended Sentencing Guidelines range of 70 to 87 months in prison. The Presentence Report (PSR), on the other hand, recommended holding Davenport responsible for 1,365.1 grams of methamphetamine, or approximately 200 times the amount to which he stipulated in his plea agreement. This higher amount, together with other factors, translated into an offense level of 37 and a recommended Guidelines range of 262 to 327 months in prison. A downward *383departure reduced the range to 151 to 188 months in prison. After explaining that the sentencing recommendations in the plea agreement lacked binding effect, the district court adopted the exponentially higher drug-quantity recommendation of the PSR and sentenced Davenport to two concurrent sentences of 130 months’ imprisonment, as well as an additional 30 months to run consecutively for his failure to appear at trial while out on bond, for a total of 160 months in prison.
On appeal, Davenport argues that the district court committed clear error by (1) attributing 1,365.1 grams of methamphetamine to him, despite allegedly insufficient evidence to establish that he participated in the theft or concealment of that quantity of the drug, and (2) adding two offense levels under U.S. S.G. § 2Dl.l(b)(l), despite allegedly insufficient evidence to support a finding that Davenport possessed a firearm during the relevant offense conduct. I set forth below the reasons why I believe that the first of these two rulings warrants vacatur.
I. ANALYSIS
Even if, as the majority opinion emphasizes, witness-credibility assessments by district courts are very rarely found to be clearly erroneous, they are subject to question in a close case where the court totally fails to explain its reasons for favoring one witness over another, objectively more reliable witness. Cf. Anderson v. Bessemer City, 470 U.S. 564, 575, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985) (“[Fjactors other than demeanor and inflection go into the decision whether or not to believe a witness. Documents or objective evidence may contradict the witness’ story; or the story itself may be so internally inconsistent or implausible on its face that a reasonable factfinder would not credit it.”). Both parties and the majority agree that the district court’s findings regarding the amount of methamphetamine attributable to Davenport must be supported by a preponderance of the evidence. The majority opinion then lists the evidence that supported the district court’s decision to attribute the entire amount of methamphetamine to Davenport, and concludes that this evidence in fact constituted a preponderance of all of the evidence in the case.
The common definition of the phrase “preponderance of the evidence,” as found in law treatises and standard jury instructions, is evidence that is of greater weight, on balance, than that offered in opposition to it. See, e.g., 32A C.J.S. Evidence § 1312 (2006). Despite lacking the district court’s advantage of personally observing the witnesses, I believe that the evidence supporting Davenport’s position outweighs the evidence supporting the district court’s attribution of the entire quantity of seized methamphetamine to him. Davenport’s position was also the government’s position at all times prior to this appeal. Although I am not quite prepared to conclude that the district court committed clear error in making this attribution, I do believe that the court clearly erred in failing to articulate why it reached the dispositive credibility determination that it did. I would accordingly vacate, rather than outright reverse, the district court’s ruling on this issue.
A. The evidence that supports Davenport
My review of the briefs and the record reveals that the following pieces of evidence support Davenport’s position that he should have been held accountable for only 6.9 grams of methamphetamine, as stipulated in his plea agreement:
(1) the plea agreement itself;
(2) the government’s objections to the PSR;
*384(3) the post-interview report compiled by DEA Agent Rick Johnson; and,
(4) the testimony of Agent Johnson.
1 & 2. Davenport’s plea agreement and the government’s objection to the PSR
As previously noted, the plea agreement recommended holding Davenport responsible for only 6.9 grams of methamphetamine, in stark contrast to the 1,365.1 grams attributed to him by the PSR. The government not only was a party to the plea agreement, but forcefully objected to the PSR in a letter submitted to the probation officer:
Specifically, the United States objects to the increased drug amount for this defendant. The United States adopts the comments made at the time of the entry of the guilty plea as to this defendant. In summary, the statements of all three defendants regarding Davenport’s conduct are inconsistent to say the least. Although all three defendants are consistent as to the source of the methamphetamine, there are inconsistent statements as to whether this defendant actually broke into the truck with Wilson. No witness has stated that Davenport knew how much methamphetamine was stolen from the truck over and above what was located at the camper trailer. There is no proof that the defendant participated in hiding the stolen methamphetamine. As stated in the guilty plea, the United States does not believe that the drug amount over and above the Plea Agreement could be proven at trial. Furthermore, as the United States agreed to a certain drug amount in the Plea Agreement any amount over and above that is objected to by the United States.
3 & 4. Agent Johnson’s report and testimony
DEA Agents Rick Johnson and David Gray interviewed Davenport immediately following his arrest. Agent Johnson conducted the interview, took notes, and alone compiled the post-interview DEA report. (Although Agent Johnson’s report is not in the record, it is repeatedly referred to by the parties and the court, without objection, during the several hearings transcribed in the Joint Appendix.) Johnson’s report quoted Davenport as stating, among other things, that “Wilson had told [Davenport that] Shawn Noe and Mike Cain assisted him, and him there would be Wilson, in stealing the methamphetamines from Walters.” Nowhere in the report is there an admission by Davenport that he himself was involved in the theft.
Agent Johnson also testified at the sentencing hearing. Contrary to Agent Gray’s testimony, discussed below, Johnson could not recall “Mr. Davenport ever acknowledging that he was present with Chris Wilson during the theft of the methamphetamine.” The remainder of Johnson’s testimony confirmed that his recollection of Davenport’s interview was consistent in all relevant respects with the report that Johnson compiled immediately after the interview was complete.
B. The evidence that supports the district court’s finding
The majority opinion lists the following four pieces of testimonial evidence “tending to inculpate” Davenport and to support the district court’s finding:
(1) Davenport’s purported post-arrest admission that he was in the truck and helped Wilson steal the methamphetamine;
(2) the testimony of DEA Agent David Gray;
*385(3) the testimony of codefendant Abner; and,
(4) the testimony of codefendant Wilson. I set forth below my doubts as to the weight to be given each of these pieces of evidence.
1. Davenport’s purported post-arrest admission
Regarding “the testimony of ... Davenport himself’ “tending to inculpate him” (Maj. Op. at 379), I assume as an initial matter that the majority is referring to paragraph 25 of the PSR, which reads as follows:
On November 4, 2004, Davenport was arrested by DEA agents in Madison County. Once in DEA custody, Davenport consented to being interviewed. According to Davenport, he and Wilson traveled together to the trailer of Bill Walters’ girlfriend, where Walters’ truck was parked. Davenport reported that he remained in the vehicle while Wilson stole the methamphetamine from Walters’ truck.
But in listing this alleged “admission” as a separate piece of evidence supporting the district court’s finding, the majority opinion misconstrues the record and improperly engages in double counting. This so-called admission by Davenport was Agent Gray’s testimony, not a separate and independent statement by Davenport himself. Agent Johnson’s post-interview report, discussed above, contained no such admission by Davenport; the purported admission was instead purely Agent Gray’s recollection. In other words, items (1) and (2) listed above — Davenport’s “admission” and Agent Gray’s testimony — are one and the same single piece of evidence, not two separate pieces of evidence corroborating the same alleged fact.
The majority responds to this criticism by noting that “Agent Gray’s recollection of Davenport’s alleged admission is not a separate piece of evidence from that admission.” (Maj. Op. at 377 n. 2) With all due respect, the true situation is exactly the opposite. Under the majority’s version, Davenport’s admission is an established fact that Agent Gray’s testimony simply corroborates. But the record makes clear that Davenport’s purported admission comes only through Agent Gray’s testimony. It does not otherwise exist.
2. Agent Gray’s testimony
Agent Gray recalled Davenport saying during his post-arrest interview that “himself and Mr. Wilson had stolen the methamphetamine,” but not, as Agent Johnson noted in the written report compiled immediately after the interview, that ‘Wilson told [Davenport that] Shawn Noe and Mike Cain assisted him in stealing the methamphetamine.” Curiously, although Gray admitted that “I don’t remember word-for-word everything [Davenport] said,” Gray claimed to be “100 percent certain that [Davenport] said to [me] that day, that he was there when the meth was stolen.” Making his selective memory even more suspect, Gray further conceded that this highly inculpatory statement that he recalled Davenport making “[w]ould ... be something that you guys in the DEA would try to include in your report.” But it was Davenport’s statement impheating Noe and Cain, not any alleged confession by Davenport, that Agent Johnson recorded in his post-interview report.
The testimony of Agent Gray is questionable for several additional reasons. To be sure, Agent Gray, along with Agent Johnson, was present during the post-arrest interview of Davenport. As Davenport notes in his brief and as neither the government nor the majority disputes, however, Agent Gray was mostly an ob*386server. He did not conduct the interview, take any notes, or assist in the preparation of the post-interview report. If this were all, I might be at least somewhat more inclined to agree with the majority that the district court’s decision to credit Gray’s testimony instead of Johnson’s is effectively immune from review. There is, however, more.
The majority opinion omits the crucial fact that the district court, not the government, called Gray as a witness during Davenport’s sentencing hearing. I do not question the propriety of the district court’s having called Gray to testify. To be sure, trial courts in criminal cases often call and are authorized to call their own witnesses. Nor do I question the district court’s having taken Gray’s testimony into account. Instead, I question the amount of weight that the district court attached to that testimony.
In the jury-trial context, judges calling their own witnesses typically instruct the jury that, in weighing the evidence, the jury is not to give the testimony of the court’s witnesses any more weight than that of the parties’ witnesses. See, e.g., Apanovitch v. Houk, 466 F.3d 460, 485 (6th Cir.2006). Strong policy rationales support this practice, which the Fourth Circuit neatly summarized some time ago:
A trial judge is not captive within the case as made by the parties. He has the authority, if not the duty, to call witnesses who possess relevant information affecting the outcome of the issues when the parties decline to call them. But the due process clause requires that a court be impartial. This impartiality is destroyed when the court assumes the role of prosecutor and undertakes to produce evidence, essential to overcome the defendant’s presumption of innocence, which the government has declined to present. Further, in this case the jury was never told why the witnesses were called as court witnesses and the jury was not instructed that these witnesses were entitled to no greater credibility because they had been called by the court. The jury, thus, may well have afforded them greater credibility than if they had been called as government witnesses. The jury’s determination of credibility of witnesses may therefore have been unfairly, albeit unintentionally, influenced and the government’s case thereby strengthened.
United States v. Karnes, 531 F.2d 214, 216-17 (4th Cir.1976).
I acknowledge that sentencing presents a different context than trial. At sentencing, for one thing, the court is free to consider otherwise inadmissible evidence such as hearsay in determining the defendant’s sentence. And a sentencing court obviously cannot instruct itself as to the amount of weight a particular piece of evidence deserves or does not deserve to be accorded. But the same policies that underlie such instructions should presumably apply.
In the present case, the district court not only called Agent Gray as its own witness, but rested its ultimate conclusion that Davenport was responsible for 1,365.1 grams of methamphetamine, as opposed to the 6.9 grams stipulated to by the government in Davenport’s plea agreement, almost entirely on Gray’s testimony. The court’s own words are telling:
Now, with respect to the pre-sentence reports, I will adopt the findings in the pre-sentence report for Mr. Davenport, as well as the additional findings that I have outlined, based on the testimony presented. In particular, I do place a lot of emphasis on Special Agent Gray.
(Emphasis added.)
This credibility determination is especially disturbing when viewed in the con*387text of Agent Johnson’s testimony, as recounted above. Agent Gray was not only the court’s own witness, but also the witness whose testimony most directly undermined Davenport’s principal witness, Agent Johnson. Absent a convincing explanation from the district court, I cannot see how Gray possibly could have won that battle in the court’s eyes.
The comparison should not have been between simply what Agent Gray said at sentencing as opposed to what Agent Johnson said at sentencing. Instead, the comparison should have pitted what Agent Gray said at sentencing against what Agent Johnson said at sentencing as well as what Agent Johnson wrote immediately following the post-arrest interview of Davenport. Johnson’s testimony at sentencing was consistent with his DEA report, which he obviously wrote before he became a witness for Davenport and thus before he could have acquired any possible motive to misrepresent the facts in Davenport’s favor. Gray himself even admitted that the statement he recalled Davenport making “[w]ould ... be something that you guys in the DEA would try to include in your report.” The district court nonetheless “placed a lot of emphasis on Special Agent Gray,” who testified based solely on his memory of words that were spoken during an interview held more than a full year earlier.
3. Abner’s testimony
The majority opinion also relies on “the testimony of codefendant Abner that Davenport helped Wilson steal the meth” as one of the pieces of evidence that the district court “implicitly credited” in finding Davenport responsible for the full quantity of methamphetamine. (Maj. Op. at 377) But this alleged testimony, which does not appear in the record and for which the majority offers no citation, misstates Abner’s actual testimony. (I do not contest the separate matter of Abner’s testimony regarding Davenport’s possession of the gun.)
At the joint sentencing hearing, the district court asked Abner no fewer than four times whether “of the three defendants that are in this case, you and Mr. Wilson and Mr. Davenport, do you have personal knowledge of who was involved in the theft of the methamphetamine?” Abner repeatedly answered “no.” Even when the district court, seemingly incredulous at Abner’s answers, pressed Abner rather forcefully by asking “You just don’t know?” she replied “No, I really don’t.”
The majority opinion summarily dismisses this testimony as insignificant, noting that “the record allowed the district court to discredit Abner’s testimony that she had no personal knowledge of who stole the meth, but it did not compel the court to do so.” (Maj. Op. at 380) For the purposes of the balancing required by the preponderance-of-the-evidence standard, however, Abner’s testimony, which tips if at all in Davenport’s favor, does matter. Of the first three pieces of evidence listed by the majority opinion, only Agent Gray’s testimony supported the district court’s decision to adopt the drug-quantity recommendation in the PSR instead of that in Davenport’s plea agreement. That leaves the fourth piece of evidence — codefendant Wilson’s testimony.
4. Wilson’s testimony
At the joint sentencing hearing, Wilson testified that “me and Gary [Davenport] went and stole it” and that “!j]ust me and Gary” were present at the time. Wilson also testified in an earlier change-of-plea hearing that he remained in the vehicle while Davenport stole the methamphetamine.
*388These are the only statements, aside from Agent Gray’s testimony, that implicate Davenport in the actual theft. Putting aside the fact that Davenport was not charged with conspiracy per se, I readily acknowledge this court’s general rule that “testimonial evidence from a coconspirator may be sufficient to determine the amount of drugs for which another coconspirator should be held accountable.” United States v. Swanberg, 370 F.3d 622, 625 (6th Cir.2004) (brackets omitted). The district court in the present case, however, at least implicitly doubted the credibility of Wilson’s testimony by not mentioning it in the court’s final analysis of the evidence presented. Instead, as noted, the court stated that “I do place a lot of emphasis on Special Agent Gray.” The shakiness of Wilson’s testimony is certainly understandable in light of the fact that he was trying to pin the primary blame for the theft on Davenport, contrary to the testimony of all of the other witnesses, including Agent Gray.
In sum, the district court’s decision to attribute all 1,365.1 grams of the seized methamphetamine to Davenport relied almost exclusively on the testimony of a DEA agent (Gray), who was called as the court’s own witness and who testified about unrecorded statements allegedly made to him over a year earlier. This is hardly a “preponderance of the evidence,” especially when measured against the contrary evidence discussed in Part I.A. above. The simple fact that all of the actual methamphetamine seized appears to have derived from the same source does not, in my opinion, change this conclusion. Who actively participated in the theft of the methamphetamine — as opposed to who later possessed the methamphetamine once it had been stolen — is the issue we have to decide.
Finally, I reiterate what the majority opinion concedes: “It was the prosecution’s burden to prove the amount of meth attributable to Davenport by a preponderance of the evidence.” (Maj. Op. at 377 (citing Swanberg, 370 F.3d at 625)) (Emphasis added.) If the prosecution “proved” that all 1,365.1 grams of the stolen methamphetamine should have been attributed to Davenport, it did so only inadvertently. The majority opinion all but ignores the fact that the government objected in writing to the PSR’s recommendation to attribute the full amount of the methamphetamine seized to Davenport. Instead, the government insisted that the district court hold Davenport accountable for only 6.9 grams, noting that “the United States does not believe that the drug amount over and above the Plea Agreement could be proven at trial.” At the sentencing hearing, moreover, the prosecution called no witnesses. These facts, taken together with those set forth above, at least raise a suspicion that the district court was intent on establishing a preponderance of the evidence by its own doing. To dispel this suspicion, it should be required to clearly articulate why it weighed the evidence as it did.
II. CONCLUSION
For all of the reasons set forth above, I would VACATE the judgment of the district court and REMAND the case for a fuller explanation of its reasons for attributing the entire quantity of seized methamphetamine to Davenport.