The Rwandan Genocide is one of the darkest chapters in human history. Over the span of 100 days, an estimated 800,000 people died. At least a million more were displaced. During and shortly after the tragedy, the United States admitted a limited number of refugees from Rwanda with priority given to those who were in the most danger. Among those admitted were Gervais ("Ken") Ngombwa and purported members of his family. The government alleged and proved at trial that his admission, status, and eventual naturalization were based on material falsehoods. And at sentencing, the government proved to the district court's 1 satisfaction that the falsehoods were used to conceal a darker secret: Ngombwa's participation in the Rwandan Genocide.

Ngombwa now alleges he deserves a new trial because his representation was constitutionally deficient and, failing that, he should be re-sentenced because the district court made several errors at sentencing. We disagree and affirm his conviction and sentence.

I.

During the Rwandan Genocide, life and death were tied to one's ethnic background. The decisive fault line was between the Hutus and Tutsis. The former made up 85% of the population; the latter made up roughly 14%.

Relations between the two ethnic groups have historically been fraught. In 1993, major parties representing both ethnicities entered into a power-sharing agreement known as the Arusha Accords. The leader of this power-sharing government, a Hutu, was assassinated in April of 1994. Extremist Hutu groups, already unhappy with the Arusha Accords, seized on this moment. They unleashed a wave of mass murder and violence against both Tutsis and those Hutus who were thought to sympathize with the Tutsis in what is now known as the Rwandan Genocide.

Against this backdrop, Ngombwa, a Hutu, sought refuge in the United States. In June 1994, two months after the outbreak of the Genocide, Ngombwa left Rwanda for a refugee camp in Tanzania. With him were his purported wife, Antoinette Mukakabanda (a Tutsi), and seven minors-only some of whom were his biological children. 2 Ngombwa was flagged for potential resettlement because he was believed to be in a mixed marriage. And in 1998, he began a series of interviews with UN officials, aid workers, and, eventually, a U.S. immigration attache, to assess whether he (and his alleged family) should be granted refugee status in the United States.

Over the course of these interviews, Ngombwa told a number of lies. These falsehoods were designed to enhance the perceived risks if he were to stay in the region. See Sent. Mem. 10-13. For example, he lied that he and Mukakabanda (and her mother) were arrested and beaten in the Rwandan capital. One lie, though, was particularly egregious-and particularly helpful to his claim. At trial, the U.S. immigration attache who interviewed him, Joe Martin, testified that Ngombwa's assertion that he was the brother of a prominent moderate Hutu politician, Faustin Twagiramungu-who was perceived as sympathetic to Tutsis-proved decisive to his refugee claim. On the basis of this false statement and many others, Ngombwa, Mukakabanda, and the seven minors (all of whom were thought to be their children), were resettled in the United States at the end of 1998.

Ngombwa eventually became a citizen of the United States in 2004. When he applied to become a permanent resident in 2001 and a citizen in 2003, he affirmed, among other things, that he never obtained an immigration benefit or entered the United States through fraud or willful misrepresentation. His purported children also gained citizenship through derivative applications. Ngombwa's naturalization was not to be the last time he dealt with immigration authorities, however.

Many years after Ngombwa's naturalization, Department of Homeland Security ("DHS") investigators received a tip. It was not a run-of-the mill tip nor was it from a run-of-the-mill source. Prosecutors in Rwanda informed DHS that they had credible information that a perpetrator of the Rwandan Genocide was residing in the United States. That tip sparked an investigation-marked by numerous interviews in Rwanda over the span of three years and extensive document review-that led to Ngombwa's doorstep in Cedar Rapids, Iowa.

In 2014, Ngombwa was interviewed twice by DHS investigators. By this time, investigators had learned that Ngombwa had twice been convicted in absentia by Rwandan tribal courts (known as "GACACA courts") for participation in the Rwandan Genocide. He had also been named in an indictment in the International Criminal Tribunal for Rwanda as a participant in the Genocide. And around the time DHS first spoke to him in June 2014, he was indicted by Rwandan prosecutors for genocide. During his interviews with DHS investigators, Ngombwa recanted many things he initially told U.S. and aid officials in the Tanzanian refugee camp. Curiously, he maintained that he never said he was related to the moderate Hutu politician Twagiramungu-a fact U.S. immigration officials found critical to his refugee claim.

In October 2014, shortly after the last time Ngombwa spoke with DHS investigators, a grand jury returned a four-count indictment against Ngombwa. Three counts alleged Ngombwa's unlawful procurement of naturalization and conspiracy to commit the same. See 18 U.S.C. §§ 1425 , 371. The last count charged him with falsely stating to investigators that he never claimed he was related to Twagiramungu. See id. § 1001(a)(2). Ngombwa went to trial, where it was agreed that no reference was to be made to evidence of Ngombwa's personal participation in the Genocide. He was convicted on all four counts. Post trial, he filed a motion for a new trial on the basis of ineffective assistance of counsel ("IAC"). The district court denied that motion, and Ngombwa proceeded to sentencing.

The district court granted Ngombwa's request to merge two counts of naturalization fraud for sentencing purposes, which meant that he would be sentenced on one count of naturalization fraud, one count of conspiracy to do the same, and one false-statement count. At sentencing, the district court heard evidence about Ngombwa's participation in the Rwandan Genocide. Largely driven by this evidence, Ngombwa's Guideline range was set at 100-125 months imprisonment. The district court eventually imposed an above-Guidelines sentence of 180 months.

Ngombwa now appeals the denial of his motion for a new trial and his sentence. We examine each in turn.

II.

Our first task is to ensure that review of Ngombwa's IAC claim is proper at this stage because "[g]enerally [ ] [IAC] claims are better left for post-conviction proceedings." United States v. Long , 721 F.3d 920 , 926 (8th Cir. 2013) (internal quotation marks omitted). We address IAC "claims on direct appeal only where the record has been fully developed, where not to act would amount to a plain miscarriage of justice, or where counsel's error is readily apparent." United States v. Hubbard , 638 F.3d 866 , 869 (8th Cir. 2011) (internal quotation marks omitted).

In this case, "the district court held an evidentiary hearing at which it allowed [the defendant] to present evidence regarding the alleged ineffective assistance of counsel." See United States v. Orr , 636 F.3d 944 , 950 (8th Cir. 2011) (alteration in original) (internal quotation marks omitted). Ngombwa's claim relates to trial counsel's investigation of certain witnesses. The district court allowed Ngombwa to present evidence, through live testimony and affidavit, on how those witnesses would have bolstered his defense. And trial counsel submitted multiple affidavits in addition to testifying at the evidentiary hearing. Given this, we find that "the trial court developed a record sufficient to examine counsel's performance" on direct review. Hubbard, 638 F.3d at 869 . The parties agree as well. See Orr , 636 F.3d at 950 (finding "additional justification for reviewing ineffective-assistance claims on direct appeal when the parties concur that the record is fully developed and thus ripe for review"). Thus, we proceed and review the district court's decision to deny a new trial on IAC grounds for abuse of discretion. United States v. Thompson , 690 F.3d 977 , 992 (8th Cir. 2012).

To succeed on an IAC claim, a party "must show: (1) trial counsel's performance was so deficient as to fall below an objective standard of the customary skill and diligence displayed by a reasonably competent attorney, and (2) trial counsel's deficient performance prejudiced the defense." Long v. United States , 875 F.3d 411 , 413 (8th Cir. 2017) (internal quotation marks omitted). We find that Ngombwa has failed to make the requisite showing on the first prong-he has not shown that his trial counsel's performance was constitutionally deficient.

On appeal, Ngombwa's claim is narrow: he believes his counsel was ineffective for failing to contact and interview five of his family members. "In any ineffectiveness case, a particular decision not to investigate must be directly assessed for reasonableness in all the circumstances, applying a heavy measure of deference to counsel's judgments." Strickland v. Washington , 466 U.S. 668 , 691, 104 S.Ct. 2052 , 80 L.Ed.2d 674 (1984). And so even if the record suggests that counsel "could well have made a more thorough investigation ... in considering claims of [IAC] we address not what is prudent or appropriate, but only what is constitutionally compelled." Burger v. Kemp , 483 U.S. 776 , 794, 107 S.Ct. 3114 , 97 L.Ed.2d 638 (1987) (internal quotation marks omitted). Put simply, there only needs to be "a reasonable basis for [counsel's] strategic decision" not to speak with Ngombwa's family members. Id. at 795 , 107 S.Ct. 3114 .

Here, there was more than a "reasonable basis" for trial counsel's decision. Counsel decided not to investigate the family members of Ngombwa because "he determine[d] that any potential information an investigation might uncover would have limited"-indeed, in this case, detrimental-"value [and] could be easily attacked on cross-examination." United States v. Mohammed , 863 F.3d 885 , 890 (D.C. Cir. 2017) (internal quotation marks omitted). As the district court noted in its findings, interviews of other family members led to the discovery that some of Ngombwa's purported children were not his biological children as he had claimed during his initial resettlement-a fact that was not known to the government until a week before trial. This was yet another falsehood Ngombwa used in the naturalization process. It was a harsh inculpatory fact for Ngombwa and strong evidence for the government. As counsel explained, he made a strategic decision not to speak with certain family members in order to prevent the government from learning this inculpatory fact and potentially others. 3 This is not a case, then, where "[c]ounsel's failure to place calls or otherwise reach out to potential witnesses cannot be traced to any strategic decision." Mohammed , 863 F.3d at 891 . Instead, " 'reasonable professional judgments support the limitations on [counsel's] investigation.' " Wiggins v. Smith , 539 U.S. 510 , 533, 123 S.Ct. 2527 , 156 L.Ed.2d 471 (2003) (quoting Strickland , 466 U.S. at 691 , 104 S.Ct. 2052 ).

In sum, the district court did not abuse its discretion in denying a new trial for Ngombwa.

III.

As to Ngombwa's sentence, he argues that we should remand for four distinct reasons, and we consider each argument separately.

A.

Ngombwa's first argument is a procedural one with many moving parts. The first piece is that Ngombwa believes his charges, which capture conduct ranging from 1998 to 2014, were improperly grouped by the district court for sentencing. The grouping rules set forth by the Guidelines instruct sentencing courts to bundle "[a]ll counts involving substantially the same harm" in order to calculate a single offense level. See USSG § 3D1.2. Given the consequences of this-normally a lower overall offense level-often times a district court's decision not to group is challenged. See, e.g. , United States v. Espinosa , 539 F.3d 926 , 929 (8th Cir. 2008).

Here, however, Ngombwa challenges the grouping of his counts because it is the first step towards the application of another part of the Guidelines he finds objectionable: the one-book rule. The one-book rule holds that one version of the Guidelines should be used when sentencing multiple counts spanning many years. See USSG § 1B1.11. Practically speaking, this is usually the edition of the Guidelines that is in effect at the time of sentencing. 4 Our precedent has approved of the one-book rule only when applied to grouped offenses. See United States v. Anderson , 570 F.3d 1025 , 1033 (8th Cir. 2009) ; but see USSG § 1B1.11, comment. (n.2)(one-book rule should be applied "regardless of whether the offenses of conviction are the type in which conduct is grouped"). 5 Thus, Ngombwa challenges the grouping of his offenses as a way to wriggle free of the one-book rule which, as described below, increases his offense level. The second piece of his argument is a more direct challenge to the one-book rule. Even if his counts were properly grouped, Ngombwa believes the rule, as applied to him, violates the Constitution's Ex Post Facto Clause.

We first examine his grouping argument before moving to his constitutional challenge. Factual findings made by the district court are reviewed for clear error while application of the Guidelines-including which version should be applied-is something reviewed de novo. Anderson , 570 F.3d at 1033 .

1.

As applicable here, the Guidelines define "counts involving substantially the same harm"-e.g., those counts which should be grouped-in part as those that "involve the same victim." USSG § 3D1.2(b). Ngombwa believes his counts were improperly grouped because his conduct affected different "victims." He draws support from the Guidelines application notes. The application notes to § 3D1.2 state that certain crimes have "no identifiable victims." Id. , comment. (n.2). For those crimes, including "immigration offenses," "the 'victim' ... is the societal interest that is harmed." Id. Ngombwa believes that his false statement count affected a specific victim, DHS, and his substantive naturalization fraud counts affected a broader "societal interest." The district court disagreed and found that all of his counts harmed DHS.

Ngombwa's argument is unconvincing. The lie he told DHS investigators in 2014 was meant to prevent them from enforcing the immigration laws. He lied specifically about the most crucial part of his resettlement application: his purported relationship with Twagiramungu. His lie to investigators was simply another act in the same play-one that continued to "corrupt[ ] the established processes of immigration." United States v. Worku , 800 F.3d 1195 , 1207 (10th Cir. 2015). In this case, at least, "[s]ubmitting false statements to the government, in an effort to circumvent a regulation promulgated for the benefit of society at large, is akin to the victimless crime[ ] of obstructing justice." United States v. Bordinaro , 777 F.Supp. 1229 , 1233 n.2 (E.D. Pa. 1991), aff'd , 970 F.2d 900 (3d Cir. 1992) (unpublished table decision).

Ngombwa responds by citing cases which suggest violations of 18 U.S.C. § 1001 are not victimless. These violations, he argues, affect specific government agencies. See, e.g , United States v. Shah , 44 F.3d 285 , 289 (5th Cir. 1995) (affirming false statement conviction for submitting manipulated bid on government contract). At one time, this sort of harm affecting the operation of government may have been the exclusive domain of the statute. See Lisa Kern Griffin, Criminal Lying, Prosecutorial Power, and Social Meaning , 97 Cal. L. Rev. 1515 , 1536 (2009) (" Section 1001 originated with concerns about government property and was passed as a response to Civil War procurement fraud."). Congressional revision, however, has expanded the scope of the statute beyond "only those falsehoods that pervert governmental functions." Brogan v. United States , 522 U.S. 398 , 402, 118 S.Ct. 805 , 139 L.Ed.2d 830 (1998).

Without a doubt, 18 U.S.C. § 1001 still covers its historical roots. But, there is a difference between the use of "false information to pursue a governmental privilege such as employment" or, like in Shah , to "receiv[e] a pecuniary benefit" from the government-instances where the government agency is squarely the victim-and this case. Michael Nagelberg et. al., False Statements and False Claims , 54 Am. Crim. L. Rev. 1273 , 1276 (2017). Here, Ngombwa's false statement was used to "frustrate lawful regulation." Id. In other words, his false statement undermined the very "societal interests" that the immigration regulations protect. See USSG § 3D1.2, comment. (n.2). Thus, under the Guidelines, his counts were properly grouped. 6 See also United States v. Pagan-Ferrer , 736 F.3d 573 , 599 (1st Cir. 2013) (affirming, without comment, district court's grouping of physical assault counts with later false statement and obstruction charges which related to investigation of underlying physical assault).

2.

Next, Ngombwa argues that application of the one-book rule to his grouped offenses violates the Constitution's Ex Post Facto Clause. Ngombwa's naturalization fraud counts cover conduct up to 2006. His false statement count, however, occurred in October 2014. So, under the one-book rule, the district court used the 2015 version of the Guidelines, the version in effect at sentencing, for all counts. 7 Under the 2015 Guidelines, the district court applied USSG § 2L2.2(b)(4)(B)(ii) -a section which was added to the Guidelines in 2012 "to increase the [G]uideline range for offenders violating immigration laws to conceal the violation of human rights." Worku , 800 F.3d at 1208 . Application of the enhancement raised Ngombwa's base offense level from 8 to 25. Ngombwa argues that this is an Ex Post Facto violation because if the false statement count "did not exist, [he] would have to be sentenced under the 2006 version of the Guidelines that does not contain the 17 level increase set forth in [USSG] § 2L2.2(b)(4)(B)(ii)."

To start, we have already rejected this argument in Anderson . 570 F.3d at 1033 (noting our prior holding that "the one-book rule does not violate the Ex Post Facto Clause, even when it results in a higher guidelines range, because defendants have fair warning that the revised guidelines manual will apply to grouped offenses" (internal quotation marks omitted) ). And as Ngombwa candidly agrees, nothing in the Supreme Court's holding in Peugh affects Anderson . See supra note 4. Indeed, every circuit that has examined the rule espoused by Anderson post- Peugh has agreed with this. 8 As a result, we have no occasion to revisit Anderson .

Ngombwa asks us to follow United States v. McMillian , 777 F.3d 444 (7th Cir. 2015). But, his reliance on that case is misplaced. In McMillian , Judge Posner (in dicta) urged caution in applying the one-book rule "indiscriminately." Id. at 448 . But, even so, Judge Posner recognized that the one-book rule was appropriate in cases where a later "offense ... embraced both crimes." Id. Ngombwa's case falls into this category. As we describe above, Ngombwa's false statement was designed to conceal his prior naturalization fraud and, in this way, his false statement "embraced" his prior crimes. Thus, even McMillian would not move Ngombwa's claim forward.

At bottom, the rule set forth in Anderson controls: application of the one-book rule to grouped offenses does not violate the Ex Post Facto Clause. And we have no reason to depart from that rule here. While Ngombwa may see unfairness in the one-book rule, "it was not the amendments to the Sentencing Guidelines that disadvantaged [Ngombwa], it was [his] election to continue [his] criminal activity." United States v. Kumar , 617 F.3d 612 , 628 (2d Cir. 2010) (internal quotation marks omitted).

B.

Ngombwa next contends that the district court's upward departure from Criminal History Category I to IV under USSG § 4A1.3 for underscored criminal history was improper. We review the district court's decision to depart upwards for abuse of discretion and the factual findings underpinning that decision for clear error. United States v. Peeples , 879 F.3d 282 , 287 (8th Cir. 2018).

Ngombwa's sole contention is that the district court improperly relied on the GACACA court convictions to depart upward because those courts do not afford due process and procedural rights. We need not wade into those waters, however.

To its credit, the district court independently assessed the voluminous evidence before it to make its own judgment as to whether Ngombwa participated in the Rwandan Genocide. It found "that the eyewitness reports of Defendant's acts of violence-bolstered by his convictions in two separate [GACACA] courts-constitute 'reliable information' indicating that the Defendant's classification in Criminal History Category I under-represents the severity of his criminal history." While the district court noted that the GACACA court convictions were consistent with its own independent judgment, it did not rely solely on the GACACA court convictions to find under represented criminal history.

The district court did, however, use the GACACA court convictions to assess which precise criminal history category to move up to-a task distinct from finding under represented criminal history in the first instance. The GACACA court convictions were proxies. They were an "attempt to assign hypothetical criminal history points to the conduct that did not result in convictions, and then determine what the appropriate criminal history category would be." United States v. Azure , 536 F.3d 922 , 932 (8th Cir. 2008). We find no abuse of discretion in using the GACACA convictions in this way. The district court's use of the convictions as proxies to "adequately explain why it determine[d] that intermediary [criminal history] categories fail to meet the purposes of [USSG] § 4A1.3," is perhaps better practice than simply noting that Ngombwa's "criminal history category would be insufficient because [he] had been a 'big time crook' "-a rationale we have affirmed in the past. United States v. Mees , 640 F.3d 849 , 854-55 (8th Cir. 2011). 9

In short, the district court did not abuse its discretion in using the GACACA court convictions as proxies to assess which criminal history category to move to.

C.

The final two challenges concern the evidence considered at sentencing. First, Ngombwa argues that witness statements attesting to Ngombwa's participation in the Rwandan Genocide-which were gathered by DHS investigators in Rwanda-should not have been considered at sentencing because they are "unreliable hearsay." As Ngombwa acknowledges, "the Rules of Evidence expressly do not apply to ... sentencing proceedings." United States v. Sheridan , 859 F.3d 579 , 583 (8th Cir. 2017) (alteration in original) (internal quotation marks omitted). Neither does the Confrontation Clause. United States v. Cross , 888 F.3d 985 , 993 (8th Cir. 2018). And thus, "[h]earsay evidence, even double hearsay, can be used at sentencing proceedings if it bears sufficient indicia of reliability to support its probable accuracy." United States v. Shackelford , 462 F.3d 794 , 796 (8th Cir. 2006) (internal quotation marks omitted). "[W]hether hearsay evidence is sufficiently reliable to support a sentencing decision depends on the facts of the particular case, and is committed to the sound discretion of the district court." United States v. Pratt , 553 F.3d 1165 , 1170 (8th Cir. 2009) (internal quotation marks omitted). We review only to ensure there was no abuse of that discretion. Sheridan , 859 F.3d at 584 .

When hearsay used at sentencing is corroborated, we have traditionally found such evidence to be sufficiently reliable. See, e.g. , United States v. Garcia , 774 F.3d 472 , 475 (8th Cir. 2014) (per curiam) (finding hearsay statements used at sentencing to be reliable because they "were consistent with other witness testimony"). 10 The district court here noted that many of the witness statements gathered by DHS investigators corroborated each other. 11 In some instances, the investigators were able to assess the veracity of these witnesses in other ways. At sentencing, the lead investigator, Frank Hunter, stated that he was able to corroborate details in witness accounts with third-party sources, including independent reports prepared by the State Department. Investigators were also able to personally corroborate certain details. For example, one witness told investigators that she remembered being in a particular church during an act of mass murder. She told investigators that, during the attack, an older woman taking refuge in the church stood up to call out to a statue of the Virgin Mary. A picture of the Virgin Mary statue from that church, taken by investigators, was presented during sentencing.

Ngombwa does not contest the district court's findings on the level of corroboration buttressing these witness statements. We reject the notion, then, that "there are no guarantees as to the reliability ... of the [witness statements]." The witness statements in this case were sufficiently corroborated by each other, documentary evidence, and the investigators' own personal observations. See United States v. Grandon , 714 F.3d 1093 , 1097 (8th Cir. 2013) (finding consistency between three witnesses was sufficient to show "corroboration [which] suggests [their] hearsay testimony ... was reliable"). Despite Ngombwa's protestations to the contrary, we find that reliance on these statements did not cause defendant to be sentenced on "misinformation of constitutional magnitude." United States v. Wise , 976 F.2d 393 , 402 (8th Cir. 1992) (en banc) (internal quotation marks omitted).

D.

The last challenge to Ngombwa's sentence concerns the testimony of an expert via videolink from the United Kingdom at Ngombwa's sentencing proceedings. Ngombwa believes there is no way to enforce a binding oath on a non-citizen not present in the United States and thus no way to guarantee that the expert was telling the truth. As we have noted above, sentencing differs from trial: "a sentencing judge may appropriately conduct an inquiry broad in scope, largely unlimited either as to the kind of information [she] may consider, or the source from which it may come." Cross , 888 F.3d at 992 (internal quotation marks omitted). Here, Ngombwa would have us impose a higher bar for foreign witnesses at sentencing than at trial. Federal Rule of Criminal Procedure 15 allows for depositions of foreign witnesses to be used as substantive evidence at trial. And federal courts have affirmed the use of foreign depositions under this rule as substantive evidence at trial-even when the deponent was not sworn in. 12 We decline Ngombwa's invitation to place restrictions on the "largely unlimited" pool of sources a court may consider at sentencing. Cross , 888 F.3d at 992 (internal quotation marks omitted).

E.

Finally, while the district court did not procedurally err in determining Ngombwa's Guidelines range, even if it did, the district court made clear that it would have handed down the same sentence after consideration of the 18 U.S.C. § 3553 (a) factors. See Sent. Tr. 443-46. And so no "reasonable probability of prejudice" warranting remand exists even if the district court committed procedural error. See Molina-Martinez v. United States , --- U.S. ----, 136 S.Ct. 1338 , 1346, 194 L.Ed.2d 444 (2016) (no prejudice where "the district court thought the sentence it chose was appropriate irrespective of the Guidelines range"). Any such error would have been harmless.

IV.

For the foregoing reasons, we affirm Ngombwa's conviction and his sentence.