United States v. Earl Hall, III

PRECEDENTIAL UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT _______________________ No. 20-2268 _______________________ UNITED STATES OF AMERICA v. EARL LAFAYETTE HALL, III, Appellant _______________________ On Appeal from the United States District Court for the Middle District of Pennsylvania District Court No. 1-16-cr-00050-001 District Judge: Honorable Sylvia H. Rambo __________________________ Submitted Under Third Circuit L.A.R. 34.1 (a) January 20, 2022 Before: JORDAN, RESTREPO, and SMITH, Circuit Judges (Filed March 14, 2022) Ronald A. Krauss Quin M. Sorenson Frederick W. Ulrich Office of Federal Public Defender 100 Chestnut Street Suite 306 Harrisburg, PA 17101 Counsel for Appellant Stephen R. Cerutti, II Kim D. Daniel Scott R. Ford Office of United States Attorney Middle District of Pennsylvania 228 Walnut Street, P.O. Box 11754 220 Federal Building and Courthouse Harrisburg, PA 17108 Jenny P. Roberts Office of United States Attorney 235 North Washington Avenue P.O. Box 309, Suite 311 Scranton, PA 18503 Counsel for Appellee __________________________ OPINION OF THE COURT __________________________ 2 SMITH, Circuit Judge. Earl Hall challenges three pieces of evidence admitted during his criminal trial: (1) testimony from his former probation officer identifying the voice on recorded phone calls as Hall’s; (2) a recording of Hall’s post-arrest interview; and (3) bank records obtained without a warrant. Hall contends that his conviction must be vacated because the District Court committed constitutional or other error in admitting each piece of evidence. The District Court did not err, so we will affirm Hall’s conviction.1 In so doing, we expound, in particular, on the due 1 The District Court had jurisdiction pursuant to 18 U.S.C. § 3231, and Hall timely appealed from the District Court’s entry of the judgment of conviction. We have jurisdiction over Hall’s appeal pursuant to 28 U.S.C. § 1291. We review decisions to admit evidence over an objection for abuse of discretion. In so doing, we review factual findings made in support of an evidentiary ruling for clear error. We apply de novo review to legal questions implicated in a decision to admit evidence—including, for example, whether admitting identification evidence would violate the defendant’s due process rights. United States v. Brownlee, 454 F.3d 131, 137 (3d Cir. 2006). We review denials of motions to suppress “for clear error as to the underlying factual findings and exercise plenary review over the District Court’s application of the law to those facts.” United States v. Vastardis, 19 F.4th 573, 580 (3d Cir. 2021). 3 process and Federal Rules of Evidence standards governing the admission of voice identification evidence. I In connection with an alleged scheme to file false unemployment claims with the Unemployment Compensation for Ex-Service Members Program, Hall and his wife, Renita Blunt, were charged with several counts of mail fraud, money laundering, and aggravated identity theft, as well as one count of conspiracy to commit mail fraud and one count of conspiracy to commit money laundering. The government sought to prove Hall’s involvement in the alleged scheme by using recordings of telephone calls made from Blunt’s cell phone to several unemployment compensation offices. Hall and Blunt were tried jointly, and at trial, Blunt testified that it was Hall who made all but one of those calls. United States v. Blunt, 930 F.3d 119, 123–24 (3d Cir. 2019). Hall was convicted on all but one of the counts submitted to the jury.2 We vacated Hall’s conviction, however, recognizing that spousal privilege grounds raised before trial should have led to the District Court’s severing of Hall’s prosecution from his wife’s. Id. at 127. Prior to his new trial, which resulted in the conviction before us in this appeal, Hall objected to the admission of three pieces of evidence. First, Hall challenged the admission of testimony from his former probation officer, Edgar Leon, who testified at Hall’s first trial that it was Hall’s voice on recorded phone calls with unemployment compensation offices. According to Hall, Leon’s testimony was unreliable because it 2 Hall was acquitted on one count of aggravated identity theft. 4 was based on insufficient contacts with Hall, and because it was a product of impermissible suggestion and pressure from the investigating officer, Joel Parisi. Thus, Hall argued that admitting Leon’s testimony would violate his Fifth Amendment Due Process rights. Second, Hall contended that the recording of Hall’s post-arrest interview was inadmissible for its proffered purpose: allowing the jury to compare Hall’s voice on the interview with the voice on the calls to the unemployment compensation offices. Hall claimed that admitting the interview for this purpose would impermissibly task the jury with identifying the voice on the calls, in violation of Federal Rule of Evidence 901, and thereby require them to act as voice identification experts, in violation of Rules 606 and 701. Third, Hall argued that the Fourth Amendment required the government to obtain search warrants for his bank records.3 3 In passing, Hall also contends that Leon’s identifications of Hall’s voice on the recordings without Hall’s counsel present violated Hall’s Sixth Amendment right to counsel. Because Hall did not timely include his right to counsel argument in his motion to suppress, the District Court declined to reach it. United States v. Hall, No. 1-16-cr-00050-001, 2019 WL 5892776, at *2 n.1 (M.D. Pa. Nov. 12, 2019). We note that, even if the District Court had reached Hall’s right to counsel argument, it would have been foreclosed by United States v. Ash. Because Hall was not present at the voice identifications, there was no possibility that Hall would have “be[en] misled by his lack of familiarity with the law or 5 The District Court rejected all three of Hall’s evidentiary challenges. United States v. Hall, No. 1-16-cr- 00050-001, 2019 WL 5892776, at *1 (M.D. Pa. Nov. 12, 2019) (denying motions to exclude Leon’s testimony and to suppress bank records); App’x 570 (admitting post-arrest interview recording). A. Leon’s identification of Hall’s voice Prior to denying Hall’s motion to suppress Leon’s testimony, the District Court conducted a hearing during which Leon explained the basis for his identification of Hall’s voice on the recorded calls. Leon testified that he was Hall’s supervising probation officer starting in early 2012 and ending in late 2013. Leon’s supervision of Hall began with a 45- minute, in-person orientation meeting in Leon’s office. During the nearly two years that Leon supervised Hall, they met in Leon’s office “approximate[ly] 17 to 18 times.” App’x 441. Leon also visited Hall at his home about five times and spoke with Hall over the phone “very frequent[ly], either setting up an appointment or rescheduling an appointment.” App’x 441– 42. Leon described Hall’s voice as “different” and as having “this deep, rich quality to it.” App’x 442. Leon also characterized Hall’s voice as “very distinct”—one that he could “remember . . . from several meetings with him over time.” App’x 441. overpowered by his professional adversary.” 413 U.S. 300, 317 (1973). 6 Leon then testified that Parisi contacted him in mid- 2014, in connection with a criminal investigation of Hall, and that Parisi asked him whether he could identify Hall’s voice on recorded phone calls. Subsequently, the two corresponded about the recorded calls on what Leon described as a “sporadic” basis up until Hall’s first trial in 2017. App’x 442 Leon and Parisi’s conversations always concerned possible identification of the voice on recorded calls as Hall’s, and Parisi never provided Leon with recordings that were known not to contain Hall’s voice. For example, in a September 2014 email, sent with the subject line “Earl Hall,” Parisi asked Leon: “Please listen to the [attached] recordings and let me know if you believe the callers are once again Earl Hall.” App’x 195. Parisi’s request followed Leon’s “100% sure” identification of Hall’s voice on other recordings, App’x 183–84, which Leon made after being asked to do so by U.S. Probation Officer (“PO”) Cristina Figueroa. Figueroa, who took over Hall’s probation supervision from Leon, and who explained that she had “been supervising [Hall] for less time,” had not been able to “make the same conclusion” as Leon.4 Id. 4 Hall appeared to suggest that Parisi’s decision to work with Leon instead of with Figueroa was another factor calling into question the reliability of Leon’s identification. In a June 2015 exchange, Leon asked Parisi, who had been waiting for over a month for a response from Leon regarding Hall’s incarceration 7 Although Leon definitively recognized Hall’s voice on some recordings, Leon was not able to confidently identify Hall’s voice in other recordings upon first listening to them. For one recording, Leon asked Parisi to provide an audio- enhanced version. After a second listen, Leon expressed that he was “90% sure it’s Earl [Hall]” who was speaking in the recording. App’x 203. Leon later confirmed that his estimation was “still 90% certainty” after being prompted by Parisi to listen again to the recording. App’x 226. In May 2016, after a first listen during another series of recordings, Leon told Parisi that the voice in some “did not sound like Earl Hall” and that, in others, he could not “with certainty say it’s Earl Hall[’s]” voice. App’x 213. Parisi, asking Leon to “clarify. . . so there is absolutely no confusion,” wrote in response with respect to those recordings: Can you please send us a reply advising your opinion, if you have one, with respect to the identity of the purported callers in the below listed recorded conversations . . . . If any of the recordings are of such a poor audio quality that dates for a prior offense, whether Parisi had asked Figueroa in the meantime for that information. Parisi responded: “No, I didn’t ask her. I decided to ask you instead because you seem interested in helping and I thought our phone conversations were good.” App’x 209. Figueroa was not a witness at Hall’s second trial. 8 you feel that you cannot state any opinion, please advise. You should, of course, only express an opinion as to the identity of any of the purported callers if you are “reasonably certain” that person is the caller. App’x 216–17 (emphases in original). In a subsequent email, Parisi added: “Given the potential Brady implications, we really need to clear up any remaining issues as soon as possible.” App’x 222. After Leon had not replied for several weeks, Parisi reminded him of the need for an answer. Leon eventually emailed in response: “Sorry about the delay. I am working from home tomorrow and can listen to the recordings again . . . . I can then follow along with the titles/dates of the [below listed] recordings and tell you what my impressions are accurately.” Id. After again listening to recordings about which he had initially expressed uncertainty as to the identity of the speaker, Leon identified Hall’s voice on some but not all of them. For two, Leon stated: “This is Earl Hall.” App’x 226–27. For another two, Leon reported that the speaker “sounds like Hall but the voice doesn’t sound as deep and as distinct as his previous recordings.” App’x 226. For a third set of two, Leon said: “This sounds like Earl Hall.” App’x 227. But Leon could not identify the voice on one recording: “I cannot confirm this is Earl Hall. Voice sounds disguised.” Id. Leon testified that he chose to make his identifications of Hall’s voice on his own accord and “did not feel pressured” 9 by Parisi to do so. App’x 462. He also testified that he did not feel that he was “led,” “intimidated,” or “persuaded” by Parisi to make the identifications. Leon declared that he made the identifications “to the best of my knowledge” and “to the best of my memory and belief.” App’x 463. In denying Hall’s motion to exclude Leon’s testimony, the District Court concluded that the government established “sufficient indicia of reliability in PO Leon’s voice identifications such that the testimony may be presented at trial.” Hall, 2019 WL 5892776, at *4. And in reaching its conclusion, the District Court applied the test for due process challenges to voice identification evidence set forth in Virgin Islands v. Sanes, where we extended the Supreme Court’s multi-factor due process standard for the reliability of eyewitness testimony to voice identification testimony. 57 F.3d 338, 340–41 (3d Cir. 1995) (citing Neil v. Biggers, 409 U.S. 188 (1972); Manson v. Brathwaite, 432 U.S. 98 (1977)). Those factors include: the opportunity of the witness to view the criminal at the time of the crime; the witness’ degree of attention; the accuracy of the witness’ prior description of the criminal; the level of certainty demonstrated by the witness at the time of the confrontation; and the length of time between the crime and the confrontation. Id. at 340. As the District Court acknowledged, the Biggers factors we applied in Sanes are not a perfect fit for Leon’s identification of Hall’s voice: “Unlike the witness in Sanes, PO Leon was not a victim of a crime and heard [Hall’s] voice during times that would require a degree of attention from PO Leon.” Hall, 2019 WL 5892776, at *4. Like the witness in 10 Sanes, though, Leon had sufficient opportunity to pay attention to Hall’s voice and was confident in his identification of Hall’s voice. Id. The District Court was not swayed by Hall’s observation that Figueroa was unable to identify Hall’s voice on the recordings. Pointing to Federal Rule of Evidence 901(b)(5), which provides that voice identification can be made via “an opinion . . . based on hearing the voice at any time under circumstances that connect it with the alleged speaker,” the District Court held that “voice identification is opinion testimony” and that Figueroa’s inability to identify Hall’s voice was simply her opinion. The Rules of Evidence did not preclude Leon from expressing a different opinion about the identity of the speaker on the recordings. Id. As to Hall’s argument that Parisi was impermissibly suggestive in how he asked Leon whether he recognized Hall’s voice on the recordings, the District Court noted only that “neither party has provided case law” indicating that a voice identification process must be conducted with “open-ended question[s]” about the identity of the speaker.5 Id. It focused more on the reliability of Leon’s voice identification testimony. And because it viewed Leon’s identification of Hall’s voice as sufficiently reliable, the District Court concluded that “the jury can make the appropriate determination as to credibility, weight, and reliability of PO Leon’s testimony.” Id. 5 The District Court did not otherwise analyze whether the identification process was suggestive. Hall, 2019 WL 5892776, at *4. 11 B. The recording of Hall’s post-arrest interview The District Court admitted the recording of Hall’s post-arrest interview over Hall’s objection. Hall renewed his objection at trial stating the same grounds. The government countered that the jurors were “free to listen to the voice of Earl Hall” in the post-arrest interview recording “and decide for themselves whether or not [Hall was] the person” who called in to the unemployment compensation offices. App’x 658. The District Court overruled Hall’s objection. C. The bank records The District Court also rejected Hall’s motion to suppress the bank records that were obtained by subpoena and without a warrant. It concluded that no warrant was required because the third-party doctrine as articulated by United States v. Miller—which provides that there is no Fourth Amendment- protected privacy interest in “information voluntarily conveyed to the banks,” 425 U.S. 435, 442 (1976)—squarely applies to bank records. Hall, 2019 WL 5892776, at *5 (alluding to Miller). The District Court reasoned that the Supreme Court’s more recent decision in Carpenter v. United States did not undermine Miller. Rather, the Supreme Court explicitly held in Carpenter that its decision there did not “disturb the application of . . . Miller.” Hall, 2019 WL 5892776, at *5 (quoting Carpenter, 138 S. Ct. 2206, 2220 (2018)). * * * As the defense anticipated, Leon testified at trial concerning his identification of Hall’s voice on the recorded calls, recounting his contacts with Hall during his time as 12 Hall’s PO. Leon said that Hall had a “deep resonating voice” that was “very distinct,” and which Leon became “accustomed to” while acting as his supervising probation officer. Leon then explained that he had been “very confident” in his previous identifications of Hall’s voice on the recorded calls to unemployment compensation offices, except for “one or two” where he was not sure. Leon went on to testify that he had listened to the recordings on “multiple occasions.” App’x 678. On cross examination, Leon acknowledged that he and Figueroa, who could not identify the recorded voice as Hall’s, had a similar number of contacts with Hall. Leon also admitted he was told by Parisi that Parisi preferred working with him over working with Figueroa, and that Parisi asked specifically (in connection with the investigation of Hall) whether he could identify Hall’s voice on the recordings. Leon conceded that, with some of the recordings, he could not definitively identify Hall’s voice until being prompted to listen to them again. The jury found Hall guilty of every submitted count of mail fraud, money laundering, aggravated identity theft, and conspiracy to commit mail fraud and to commit money laundering. Hall filed this timely appeal. II Hall first contends that the District Court’s admission of Leon’s testimony identifying Hall’s voice on the recorded phone calls violated his Fifth Amendment Due Process rights. We disagree. Because Leon’s identification of Hall’s voice was sufficiently reliable in light of the substantial period of time Leon spent with Hall, both in person and over the phone, 13 allowing the jury to hear Leon’s identification testimony did not offend due process. Constitutional protections protecting a criminal defendant “against a conviction based on evidence of questionable reliability” generally do not prohibit the introduction of the evidence. Perry v. New Hampshire, 565 U.S. 228, 237 (2012). But in a criminal trial, there are some circumstances in which allowing a jury to consider certain evidence can be so unfairly prejudicial to the defendant that admitting the evidence would violate the defendant’s due process rights. In Foster v. California, for example, the Supreme Court determined that a police lineup was conducted in such a suggestive manner that it was “virtually inevitable” that the eyewitness would identify the criminal defendant. 394 U.S. 440, 443 (1969). Because the procedure “so undermined the reliability of the eyewitness identification,” admission of the resulting identification violated the defendant’s due process rights. Id. (emphasis added). Following a line of eyewitness identification cases, the Supreme Court in Biggers clarified that the admission of identification evidence offends due process only if the evidence meets two criteria. First, the evidence must be “so impermissibly suggestive as to give rise to a very substantial likelihood of irreparable misidentification.” 409 U.S. at 196– 97 (emphasis added). Second, if the evidence is impermissibly suggestive, it must also be unreliable. Id. at 198–200. The two criteria work in tandem, even though the admissibility of identification evidence often hinges on reliability. Brownlee, 454 F.3d at 139 (“reliability is the linchpin in determining the admissibility of identification testimony” (quoting Brathwaite, 432 U.S. at 114) (cleaned up)). Put differently, due process 14 requires the suppression of an identification only if it was obtained pursuant to a suggestive process that in turn raises serious questions about the reliability of the resulting identification. Cf. Perry, 565 U.S. at 248 (holding that courts need not inquire into the reliability of an eyewitness identification when it is not procured “under unnecessarily suggestive circumstances”). Biggers set forth a multi-factor test for the reliability of eyewitness testimony that we subsequently extended to voice identification testimony in Sanes as a “source of guidance.” The factors are: (1) the opportunity for the witness to view the perpetrator at the time of the crime; (2) the witness’s degree of attention; (3) the accuracy of the witness’s prior description of the perpetrator; (4) the witness’s level of certainty at the time of the identification; and (5) the length of time between the crime and the confrontation. Sanes, 57 F.3d at 340 (citing Biggers, 409 U.S. at 198–200). In Sanes, the victim of the charged offenses became familiar with the defendant’s voice during two alleged attacks by the defendant. Id. at 341. Fifteen days after the second attack, the victim identified the voice sample containing the defendant’s voice from an array of samples of different voices, each with “unique factors.” Id. at 340–41. In making the identification, the victim was “certain” that the voice in the sample was the voice of her attacker. Id. at 340. We concluded in Sanes that admitting the victim’s identification of the defendant’s voice did not offend due process, as the identification was neither unreliable nor procured pursuant to a suggestive process. First, we found the identification not to have been procured by a suggestive process because the victim had picked out the defendant’s 15 voice sample from an array of samples bearing distinguishing characteristics. Id. at 340. Second, the identification was reliable because the victim “listened to her attacker for a considerable period of time” over the course of the two attacks and also “engaged him in conversation in the hope that she could identify his voice.” Id. at 341. She was also confident in her identification and made it shortly after the second attack. Id. at 340–41. Hall, invoking Sanes, argues both that the identification process was impermissibly suggestive and that Leon’s identification of the recorded voice as Hall’s was not sufficiently reliable. Whereas the voice identification in Sanes followed a procedure where the witness identified the defendant’s voice from an array of unidentified voices, here, Parisi asked Leon—specifically in connection with the investigation of Hall—whether Leon recognized Hall’s voice on recordings obtained through the investigation. Hall also contends that Leon was insufficiently familiar with Hall’s voice, characterizing Leon’s supervision of him as only monthly and usually conducted by telephone. And Hall notes that there was no proof offered at trial indicating that Leon could distinguish Hall’s voice from others. He argues that Leon’s inability to definitively identify a voice as Hall’s on some recordings after a first listen suggests that Leon was not sufficiently familiar with his voice to be capable of providing a reliable identification. Yet even assuming that the identification procedure was suggestive, the due process challenge to the admission of Leon’s testimony fails because Leon’s identification of Hall’s voice on the recordings was sufficiently reliable. In reaching this conclusion, we note that the Biggers multi-factor test we 16 adopted in Sanes for voice identifications contemplated the reliability of a perpetrator’s identification as being made by the victim of the offense. Sanes, 57 F.3d at 340 (concerning victim’s identification of the defendant’s voice); Biggers, 409 U.S. at 200–01 (concerning victim’s identification of the defendant–petitioner’s face). Crime victim identifications of defendants present special reliability considerations, as the nature and circumstances of the crime may affect the accuracy of the victim’s memory of the perpetrator’s characteristics. E.g., United States v. Stevens, 935 F.2d 1380, 1392 (3d Cir. 1991) (“Courts have recognized that victims, out of fear, often focus their attention on the perpetrator’s weapon, rather than his face.” (collecting cases)); Brownlee, 454 F.3d at 139–40 (similar); see generally 2019 Report of the U.S. Court of Appeals for the Third Circuit Task Force on Eyewitness Identifications, 92 TEMP. L. REV. 1, 77–93 (2019) (discussing research on the posited influence of certain “estimator variables”—such as the “weapon focus” effect, the “cross-race effect,” and the length and circumstances of the time that the witness has to observe the perpetrator—on the reliability of eyewitness identifications). By contrast, non-victim voice identification witnesses in some circumstances are “in a position to offer uniquely reliable testimony.” Brown v. Harris, 666 F.2d 782, 786 (2d Cir. 1981). Non-victim witnesses such as Leon may be among those who have familiarized themselves with the defendant’s voice under circumstances controlling for factors—such as stress—that could impair the accuracy of a voice identification. For example, the witness may have heard the defendant’s voice over the course of multiple or extended conversations. E.g., id. (witnesses had interviewed defendant– petitioner for “many hours”); United States v. Brown, 510 F.3d 17 57, 67 (1st Cir. 2007) (similar); United States v. Kim, 577 F.2d 473, 482 n.20 (9th Cir. 1978) (similar). Because non-victim witnesses may have learned a defendant’s voice through controlled circumstances, their knowledge of the defendant’s voice may be “so firm that [it] is not susceptible to suggestion.” Harris, 666 F.2d at 786; see also Kim, 577 F.2d at 483 (concluding that there was “little, if any, indication that any voice witness in this case was manipulated so that the mental image derived from the identification procedure supplanted that derived from the witness’s own experience” (cleaned up)); see generally 2019 Report of the Third Circuit Task Force, 92 TEMP. L. REV. at 17 (distinguishing identifications of perpetrators “already well known to the witness” from “when the witness and perpetrator are strangers”). Further, a voice identification may be particularly reliable if the witness has the benefit of identifying the voice on a recording of the crime itself as “memorialized on tape.” Harris, 666 F.2d at 786; compare with Sanes, 57 F.3d at 340 (involving victim’s identification of the defendant’s voice on a voice exemplar). And a voice identification may be more reliable if the witness has “the luxury of listening to the tape in an office” or a similar environment “where they can devote their full attention to [the identification].” Harris, 666 F.2d at 786; accord United States v. Recendiz, 557 F.3d 511, 528 (7th Cir. 2009); cf. Brown, 510 F.3d at 68 (concluding that a voice identification was admissible even though the circumstances there—“three men huddled together listening to the same cell phone”—were “not the best for making a voice identification”). Here, Leon was not a victim but rather Hall’s probation officer. He was familiar with Hall’s voice because of their 18 repeated conversations, by phone and in person, which took place over a period of almost two years. Leon identified Hall’s voice on the recorded calls to unemployment compensation offices over the course of several months, on his own time, with headphones, and remotely via email correspondence. Leon testified that he neither felt pressured nor was led by Parisi to make specific voice identifications. Thus, Leon’s identification of Hall’s voice was reliable enough to satisfy the requirements of due process as applied to identification evidence. Because Leon learned Hall’s voice under controlled circumstances—during his time supervising Hall as his probation officer—his voice identification at trial met the criteria for identification testimony that our sister circuits have described as “uniquely reliable.” Harris, 666 F.2d at 786; Recendiz, 557 F.3d at 528. Leon’s identification of Hall’s voice also satisfies the Sanes and Biggers standards: Leon had ample opportunity to hear and pay attention to Hall’s voice over the course of their multiple conversations.6 Sanes, 57 F.3d at 340 (including factors such as the opportunity to perceive the perpetrator and the witness’s degree of attention). Even if we assume Parisi improperly primed Leon to make voice identifications by asking him if he could identify Hall’s 6 Hall also suggests that Leon’s identifications of his voice were unreliable because they occurred too long after their contacts with one another. We disagree in light of the “degree of contact” between Hall and Leon. Brown, 510 F.3d at 67 (“Given the degree of contact between [the voice identification witnesses] and [the defendant], we accord very little weight to the fact that most of it occurred ten or twelve years prior to [the defendant’s] arrest.”). 19 voice on recorded calls, Hall has not mustered evidence sufficient to indicate that the suggested identification “supplanted that derived from [Leon’s] own experience.” Kim, 577 F.2d at 483. Because Leon’s testimony was also “an opinion . . . based on hearing the voice at any time under circumstances that connect [the voice] with the alleged speaker,”7 thereby satisfying Rule 901(b)(5)’s requirements for voice identification,8 all remaining questions about the weight and credibility of Leon’s testimony were properly placed before the jury. Recendiz, 557 F.3d at 528 (“Any remaining concerns regarding the accuracy of [the witness’s] recollection of the voice are relevant to the weight of the 7 Although here, Leon sufficiently identified Hall’s voice through “direct recognition of the person calling,” we note that the identity of the speaker on a telephone call may also be authenticated “by circumstantial evidence.” United States v. Console, 13 F.3d 641, 661 (3d Cir. 1993) (citing FED. R. EVID. 901; id. advisory committee’s note to subdivision (b), example (4)). 8 Rule 901(b)(5)’s standard aside, we held in United States v. Vento that “it is permissible to base the identification of a voice heard in intercepted conversations on relatively few conversations between [the person identifying the voice] and the accused person.” 533 F.2d 838, 865 (3d Cir. 1976). Hall appears to suggest that Vento is no longer persuasive in part because it is nearly 50 years old. But that is no basis to set aside precedent. And Vento’s reasoning remains directly on point because it tracks the language of Rule 905(b)(5), even without specifically invoking that rule. 20 testimony, not its admissibility.”). As the Supreme Court teaches us in Perry: “The Constitution . . . protects a defendant against a conviction based on evidence of questionable reliability, not by prohibiting introduction of the evidence, but by affording the defendant means to persuade the jury that the evidence should be discounted as unworthy of credit.” 565 U.S. at 237. Leon’s identification did not fall into the due process exception to this general rule. Id. (citing Dowling v. United States, 493 U.S. 342, 352 (1990)). So traditional constitutional safeguards such as “confrontation plus cross- examination” of Leon sufficed to protect Hall against the possibility of a fundamentally unfair conviction. Id. (citing Delaware v. Fensterer, 474 U.S. 15, 18–20 (1985) (per curiam)). Thus, the District Court did not err by admitting Leon’s testimony. III Hall fares no better in his second challenge relating to the recorded phone calls that were the subject of Leon’s testimony. He claims that the admission of a recording of his post-arrest interview, which the government introduced so that the jury could compare Hall’s voice in the interview with the voice in the recorded phone calls, contravened the Federal Rules of Evidence in two ways. First, Hall argues that admitting the interview recording violated Rule 901(a)’s prohibition against unidentified evidence by improperly tasking the jury with using Hall’s voice in the interview to identify the voice on the recorded phone calls. Second, Hall contends that admitting the interview recording so that the jury could compare it with the recorded phone calls inappropriately charged the jurors with acting as their own voice identification 21 expert witnesses, in violation of Rule 606(a)’s prohibition against juror testimony before other jurors at trial and of Rule 701(c)’s prohibition against lay testimony based on scientific, technical, or other specialized knowledge within the scope of Rule 702. We do not agree that admitting the interview recording violated Rule 901’s identification requirements because the government sufficiently identified Hall’s voice on the calls during pretrial proceedings using Leon’s testimony. See discussion supra. Thus, we see no issue with playing the interview recording in conjunction with the recorded calls “so that the jury could make its own aural comparisons.” United States v. Baller, 519 F.2d 463, 466–67 (4th Cir. 1975) (affirming admission of voice spectrography expert analysis “despite doubts within the scientific community about its absolute accuracy,” in part because the tapes compared in the analysis were played to the jury). In so concluding, we note that it would have been permissible under the Rules of Evidence for the government to have sought to identify Hall’s voice in the recorded calls using Hall’s voice in the recorded interview. FED. R. EVID. 901(b)(5) (providing that voices may be identified “based on hearing the voice at any time under circumstances that connect it with the alleged speaker” (emphasis added)). We also conclude that admitting the interview recording for purposes of comparison to the voice on the recorded calls did not improperly charge the jurors with serving as expert witnesses. Hall does not identify a case, nor have we found one, that limits voice comparison testimony to the realm of expert opinion. Rather, our sister circuits have held that lay witnesses may offer voice identification testimony, e.g., United 22 States v. Mendiola, 707 F.3d 735, 739 (7th Cir. 2013),9 and the advisory committee notes to Rule 901 explicitly state that “aural voice identification is not a subject of expert testimony,” id. (quoting FED. R. EVID. 901 advisory committee’s note to subdivision (b), example (5)). Voice identification evidence, contrary to Hall’s contention, is subject to the “general rule” that “expert testimony not only is unnecessary but indeed may properly be excluded in the discretion of the trial judge.” Salem v. U.S. Lines Co., 370 U.S. 31, 35 (1962); see also FED. R. EVID. 702 advisory committee note on proposed rules (“Whether the situation is a proper one for the use of expert testimony is to be determined on the basis of assisting the trier [of fact].”). 9 See also United States v. Cambindo Valencia, 609 F.2d 603, 640 (2d Cir. 1979) (“voice identification is not generally considered to be an area where expertise is important” (citing FED. R. EVID. 901 advisory committee’s note to subdivision (b), example (5))); United States v. Lampton, 158 F.3d 251, 259 (5th Cir. 1998) (similar); United States v. Zepeda-Lopez, 478 F.3d 1213, 1220 (10th Cir. 2007) (“the defendant’s arguments that a [lay] witness’s voice identification testimony was deficient because the witness was not an expert in voice identification . . . go to the weight of the evidence” (emphasis added)); cf. United States v. Diaz-Arias, 717 F.3d 1, 14–15 (1st Cir. 2013) (seeing no issue with lay voice identification testimony and concluding that the jurors were not “misled into thinking that [the lay voice identification witness] was an expert witness”). 23 Accordingly, we hold that the District Court’s admission of the post-arrest interview recording was not contrary to the Federal Rules of Evidence. IV Hall’s final evidentiary challenge is to the admission of his bank records that were obtained pursuant to ordinary subpoenas. According to Hall, he had a Fourth Amendment- protected privacy interest in the contents of his bank records, so the government was prohibited from obtaining the records without first securing a search warrant. Because the government did not obtain the bank records pursuant to a warrant, Hall contends that the records should have been suppressed. He argues that suppression is required under Carpenter, in which the Supreme Court held that the government must obtain a warrant before seeking to obtain an individual’s cell phone location information, even though the individual “continuously reveals his location to his wireless carrier.” 138 S. Ct. at 2216–17. Hall’s Fourth Amendment suppression argument is squarely foreclosed by Miller and the third-party doctrine, which provide that there is no Fourth Amendment-protected privacy interest in bank records voluntarily conveyed to the banks. 425 U.S. at 442. The Supreme Court reaffirmed Miller as good law in Carpenter. 138 S. Ct. at 2216 (“the third-party doctrine applies to . . . bank records”); id. at 2220 (“We do not disturb the application of . . . Miller”). In the wake of Carpenter, some courts have expressed doubt that the third-party doctrine extends to certain information collected by modern technologies. E.g., United States v. Moalin, 973 F.3d 977, 989–93 (9th Cir. 2020) 24 (doubting, but not reaching, whether warrantless telephony metadata collection comported with the Fourth Amendment, as suppression was not warranted on the facts). But of course, Hall is not seeking to suppress personal information collected by technologies unanticipated by Miller. He simply challenges the warrantless seizure of bank records that do not substantively differ in character from the bank records considered by the Miller Court. Even if Hall were correct that modern realities cast doubt on the continued persuasiveness of Miller’s reasoning, we would still be bound to follow it. Agostini v. Felton, 521 U.S. 203, 237 (1997) (“the Court of Appeals should follow the case which directly controls”). After all, “only the Supreme Court may reverse its prior precedent,” and Carpenter expressly declined to overrule Miller. United States v. Moore-Bush, 963 F.3d 29, 31 (1st Cir. 2020) (discussing Carpenter’s reach), withdrawn on other grounds, 982 F.3d 50 (1st Cir. Dec. 9, 2020). V For these reasons, we will affirm Hall’s conviction. 25