Before HAYNES and OLDHAM, Circuit Judges, and HANEN,* District Judge. PER CURIAM: Jesus Rodriguez-Peña appeals his 44-month sentence for illegal reentry under 8 U.S.C. § 1326(a)–(b). He argues the district court plainly erred in calculating his Guidelines range. The Government concedes the calculation error and challenges only whether the error was prejudicial and requires correction. On the facts of this case, we vacate and remand for resentencing. On May 5, 2018, agents from Customs and Border Protection encountered Rodriguez-Peña near Penitas, Texas. Rodriguez-Peña had been deported or removed from the United States in 2002, in 2007, and most recently on January 24, 2017. So the Government charged him with illegally reentering the country after having been previously removed. See 8 U.S.C. § 1326(a)–(b). Rodriguez-Peña pleaded guilty on June 28, 2018. Before the sentencing hearing, the probation officer prepared a Presentence Report (“PSR”). The PSR assigned an offense level of 17 and a criminal history category of III. That produced a recommended Guidelines range of 30–37 months.† The PSR’s criminal-history calculation was premised on two offenses. First, in 2003, Rodriguez-Peña pleaded guilty to the felony of indecency with a child involving sexual contact. See TEX. PENAL CODE § 21.11. The child involved in that incident was Rodriguez-Peña’s 14-year-old cousin. After serving a 42-month sentence for that offense, Rodriguez-Peña was removed in 2007. He returned and eventually pleaded guilty to illegal reentry. That was his second relevant offense. He then served a 41-month sentence and was removed again on January 24, 2017. The PSR counted each of those offenses—indecency with a child and illegal reentry—for three points each, leading to a total of six criminal history points. That placed Rodriguez-Peña in criminal history category III. At sentencing, the judge emphasized the need for Rodriguez-Peña to not illegally reenter the country again. The judge further explained that since a 41-month sentence had proved insufficient to deter Rodriguez-Peña, “a graduated sentence, something bigger than [41 months]” would likely be appropriate. So, although the judge adopted the PSR’s findings, he also determined that criminal history category III substantially underrepresented the seriousness of Rodriguez-Peña’s prior criminal conduct and did not reflect “ the likelihood of recidivism.” The judge then opted for criminal history category IV, which had a range of 37–46 months. He sentenced Rodriguez-Peña to 44 months in prison. Rodriguez-Peña did not object then. He now appeals. As Rodriguez-Peña concedes, our review is for plain error. Plain error requires a defendant to show: “(1) that the district court committed an error (2) that is plain and (3) affects his substantial rights and (4) that failure to correct the error would seriously affect the fairness, integrity or public reputation of judicial proceedings.” United States v. Sanchez-Hernandez, 931 F.3d 408, 410 (5th Cir. 2019) (internal quotation marks omitted). As the Government concedes, the first two prongs of plain error are met. At issue, then, are prongs three and four. In most cases where prong three is satisfied, this court “must ‘exercise o[ur] discretion’ to remand.” United States v. del Carpio Frescas, 932 F.3d 324, 333 (5th Cir. 2019) (quoting Rosales-Mireles v. United States, 138 S. Ct. 1897, 1909 (2018)). That is because usually “a plain Guidelines error that affects substantial rights” will also satisfy the fourth prong of plain-error review. Rosales-Mireles, 138 S. Ct. at 1908. At prong three, Rodriguez-Peña argues that without the plain error in his Guidelines calculation, his “departure range” would have dropped to 15–21 months. We note the district court appeared to base its sentence in large part on the apparent insufficiency of Rodriguez-Peña’s previous 41-month sentence. The district judge said, “you already got a 41-month sentence in here, you’ve committed the same crime over again and—you know, likely I should give you a graduated sentence, something bigger than that to prevent you from coming back next time, when you didn’t stay out even a year.” (Emphasis added). In that sense, this case is much like Sanchez-Hernandez, in which this court affirmed the same district judge’s decision to impose “graduated punishment” on a man who sexually abused a child and had a history of illegal reentry. See 931 F.3d at 411–12. Unlike in Sanchez-Hernandez, however, the district judge in this case did mention the incorrect Guidelines range in explaining the sentencing decision. The judge noted that he had considered the appropriate sentencing factors and found “that a sentence within these guidelines satisfies them . . . .” Moreover, the district judge indicated that he “depart[ed] from the guideline range for one or more reasons provided in the Guidelines Manual.” So the Guidelines apparently played a more significant role here than in Sanchez-Hernandez. Cf. United States v. Wikkerink, 841 F.3d 327, 338 (5th Cir. 2016) (“[I]n the normal course, a non-Guideline sentence still uses the Guidelines range as a reference point.”). And “[e]ven if the sentencing judge sees a reason to vary from the Guidelines, if the judge uses the sentencing range as the beginning point to explain the decision to deviate from it, then the Guidelines are in a real sense the basis for the sentence.” Peugh v. United States, 569 U.S. 530, 542 (2013) (internal quotation marks omitted). The Supreme Court has also said that if “the record is silent as to what the district court might have done had it considered the correct Guidelines range, the court’s reliance on an incorrect range in most instances will suffice to show an effect on the defendant’s substantial rights.” Molina-Martinez v. United States, 136 S. Ct. 1338, 1347 (2016). That is not true in all cases. Id. at 1346. The Molina-Martinez Court was quite clear that “appellate courts retain broad discretion in determining whether a remand for resentencing is necessary.” Id. at 1348. As an example, the Court pointed to “mechanisms short of a full remand to determine whether a district court in fact would have imposed a different sentence absent the error.” Id. (citing United States v. Currie, 739 F.3d 960, 967 (7th Cir. 2014)). We asked the parties to prepare to discuss the possibility of a limited remand in this case. At oral argument, the Government rejected that option. So we decline to apply it here. On the facts of this case and under current Supreme Court precedent, we hold that Rodriguez-Peña has met prongs three and four of plain error review. See del Carpio Frescas, 932 F.3d at 333. * * * We VACATE the sentence and REMAND to allow the district court to resentence Rodriguez-Peña in accordance with this opinion. Nothing in this opinion precludes the district court from exercising its discretion to depart from the Guidelines and choose any sentence permitted by 18 U.S.C. § 3553. ANDREW S. OLDHAM, Circuit Judge, concurring: Another day, another “plain error” vacatur. All because of an error that the defendant failed to notice in the district court but now contends is “plain” and obvious. What’s not obvious is whether the error prejudiced Rodriguez- Peña in any way. The Supreme Court told us that, in cases like this one, we should explore remedies that are less severe than a full-blown vacatur and resentencing. Molina-Martinez v. United States, 136 S. Ct. 1338, 1348–49 (2016). The “limited remand” is one such remedy. See id. But in this case, the Government disavowed it. In an appropriate case, the limited remand strikes me as a wise solution to the plain-error problem. I. Let’s start with plain error. To meet that standard, a “defendant must show (1) that the district court committed an error (2) that is plain and (3) affects his substantial rights and (4) that failure to correct the error would seriously affect the fairness, integrity or public reputation of judicial proceedings.” United States v. Sanchez-Hernandez, 931 F.3d 408, 410 (5th Cir. 2019) (quotation omitted). The first three prongs of the plain-error standard come from the text of Rule 52(b) of the Federal Rules of Criminal Procedure. It says: “A plain error that affects substantial rights may be considered even though it was not brought to the court’s attention.” FED. R. CRIM. P. 52(b); see also United States v. del Carpio Frescas, 932 F.3d 324, 343 (5th Cir. 2019) (Oldham, J., concurring). The textual hook for the fourth prong is the word “may.” See del Carpio Frescas, 932 F.3d at 343 (Oldham, J., concurring). That’s the stuff of discretion: The court “may . . . consider[]” the forfeited error. See Henderson v. United States, 568 U.S. 266, 282 n.1 (2013) (Scalia, J., dissenting); cf. Kingdomware Techs., Inc. v. United States, 136 S. Ct. 1969, 1977 (2016) (noting that “may” connotes discretion). The Supreme Court gave us guidance for the exercise of that discretion in United States v. Olano, 507 U.S. 725 (1993). There, the Court explained that Rule 52(b) “leaves the decision to correct the forfeited error within the sound discretion of the court of appeals, and the court should not exercise that discretion unless the error seriously affects the fairness, integrity or public reputation of judicial proceedings.” Olano, 507 U.S. at 732 (emphasis added) (quotation omitted). Judging by the text of Rule 52(b) and the Court’s guidance on when we “may” correct a forfeited error, you’d think a defendant trying to show plain error would have a hard time doing so. See, e.g., Puckett v. United States, 556 U.S. 129, 135 (2009) (“Meeting all four prongs is difficult, as it should be.” (quotation omitted)). You might think that’s true even in a case like this one, where the Government has conceded both the error and its plainness. But it turns out that prong three—whether an error affected a defendant’s substantial rights—offers little resistance. “Our Court and others routinely conclude criminal defendants have cleared the third hurdle whenever they show a Guidelines calculation error.” Del Carpio Frescas, 932 F.3d at 342 (Oldham, J., concurring) (collecting cases).[1] Often enough, then, if a defendant gets past the first two prongs, the question at prong three—whether the error affected his substantial rights—is as good as answered. Even so, if you’d only read Olano and Puckett, you might imagine the fourth prong would do some work. At the fourth prong, Olano said we “should not” exercise our discretion to correct a forfeited error except where “the error seriously affects the fairness, integrity or public reputation of judicial proceedings.” 507 U.S. at 732 (quotation omitted). Think about the judicial proceedings in this case. After hearing from Rodriguez-Peña and his attorney at the first scheduled sentencing hearing, the court below scheduled another hearing to allow Rodriguez-Peña to conduct further preparation. Then, at the second sentencing hearing, the judge confirmed with Rodriguez-Peña that he’d read and reviewed the PSR with his attorney. Rodriguez-Peña affirmed the PSR as correct. The PSR noted that he had been deported or removed three times before. It also described his felony of indecency with a child involving sexual contact. See TEX. PENAL CODE § 21.11. The offense report prepared by the sheriff’s office alleged that, while his 14-year-old cousin was sleeping, Rodriguez-Peña began kissing her. She woke up and tried to escape but could not. He then groped her and forced her to touch his penis. Afterwards, he told her she “knew what would happen” if she reported the assault. Then he pointed to a pistol. The sentencing court considered that crime, plus the apparent insufficiency of Rodriguez-Peña’s recent 41-month sentence, and determined that criminal history category IV (with a range of 37–46 months) better suited Rodriguez-Peña. Given the court’s apparent inclination to sentence Rodriguez-Peña to something more than 41 months based on the inadequacy of his prior sentence, you might struggle to see how an error in the PSR—which he read, reviewed, and expressly affirmed—calls into question the fairness or integrity of the judicial system.[2] Likewise, when a man has been removed from the country three times and imprisoned for molesting his fourteen-year-old cousin, it’s not obvious that a sentence clearly meant to deter him from returning yet again will do much damage to the “public reputation of judicial proceedings,” Olano, 507U.S. at 732(quotation omitted), even if there’s a miscalculation of the advisory Guidelines range. But in “the ordinary case” involving a Guidelines miscalculation, we’ve been told that when the third prong is met, the fourth prong is too. See Rosales-Mireles v. United States, 138 S. Ct. 1897, 1911(2018). So under plain-error review as it stands today, the court is right to remand for resentencing under these circumstances. II. Resentencings are not as costly as new trials. Molina-Martinez, 136 S. Ct. at 1348–49. Yet it’s equally clear that resentencings are not “costless.” Id. at 1348 (quotation omitted). A. First, consider that the overwhelming majority of criminal convictions in the federal system result from guilty pleas rather than trials. See Missouri v. Frye, 566 U.S. 134,143(2012). For these defendants, the sentencing hearing is the only “trial” they’re going to get. Moreover, it’s obviously easier to win a new sentencing hearing than a new trial. So even if one resentencing is much less costly than one new trial, the sheer volume of the former can impose significant costs on the system. Cf. Ryan W. Scott, In Defense of the Finality of Criminal Sentences on Collateral Review, 4 WAKE FOREST J. L. & POL’Y 179, 201 (2014) (“[T]aking into account aggregate costs, government interests in the finality of sentences are just as strong as interests in convictions.”). Second, think about what goes into a new sentencing hearing. With each resentencing, a busy district court must do its hearing all over again. For the Government, there’s also the added cost and complexity of transporting a prisoner, who may be serving his sentence somewhere clear across the country, back to the sentencing court. See FED. R. CRIM. P. 43(a)(3) (requiring the defendant’s presence at sentencing). The probation office might have to prepare a new PSR. The lawyers have another hearing. And what about witnesses? Maybe it’s no big deal for an expert to testify again. But it might be a very big deal to ask a victim to testify again. And even for victims who don’t have to testify, just the uncertainty of a resentencing can impose very real “human costs.” United States v. Lewis, 823 F.3d 1075, 1081 (7th Cir. 2016). Thus appellate courts do well to “keep in mind the costs of remands for resentencing . . . .” Ibid. Molina-Martinez invites us to use limited remands to avoid some of those costs. Indeed, it was the Government’s “concern over the judicial resources needed for the resentencing proceedings” that led the Court to discuss limited remands as an example of the “broad discretion” retained by appellate courts in carrying out plain-error review. Molina-Martinez, 136 S. Ct. at 1348. That’s why the Molina-Martinez Court said we could order a “limited remand so that the district judge [could] consider, and state on the record, whether she would have imposed the same sentence . . . .” 136 S. Ct. at 1348 (quoting United States v. Currie, 739 F.3d 960, 967 (7th Cir. 2014)). Such a remand, the Court noted, would help appellate courts “determine whether a district court in fact would have imposed a different sentence absent the error.” Ibid. A limited remand can therefore save the system the expense of a needless remand for resentencing. B. Limited remands can increase the accuracy of our plain-error review, too. Take, for example, the way the Seventh Circuit used limited remands after United States v. Booker, 543 U.S. 220 (2005). Recall that in that case, the Supreme Court held that the Guidelines are advisory rather than mandatory. See id. at 246. Afterwards, courts grappled with “the application of the plain-error doctrine to appeals from sentences rendered under the federal sentencing guidelines” before Booker. United States v. Paladino, 401 F.3d 471, 474 (7th Cir. 2005) (Posner, J.). Booker presented reviewing courts with a hard counterfactual. District courts had imposed certain sentences thinking the Guidelines were mandatory—how could reviewing courts know whether those district courts would have done the same under advisory Guidelines? Judge Posner’s solution to that conundrum emphasized both efficiency and accuracy. After Booker, “[t]he only practical way (and it happens also to be the shortest, the easiest, the quickest, and the surest way) to determine whether the kind of plain error argued in these cases has actually occurred is to ask the district judge.” Paladino, 401 F.3d at 483. This approach constituted a middle ground between presuming prejudice and requiring defendants to prove the impossible. Id. at 484–85. A limited remand would allow the sentencing judge to “dispel[] the epistemic fog” of plain-error review. Id. at 484. So, “in Booker cases in which it is difficult for us to determine whether the error was prejudicial,” the court would “retain[] jurisdiction of the appeal [and] order a limited remand to permit the sentencing judge to determine whether he would (if required to resentence) reimpose his original sentence.” Ibid. If the judge affirmed that he would have imposed the same sentence regardless of the nature of the Guidelines, then the appellate court would subsequently affirm any reasonable sentence. Ibid. Contrariwise, if the judge stated he would have imposed a different sentence if he had known the Guidelines were advisory, the appellate court would “vacate the original sentence and remand for resentencing.” Ibid. The Paladino remand had become well-established by the time of Currie, the Seventh Circuit decision cited by the Molina-Martinez Court. See Currie, 739 F.3d at 964–65. In Currie, the Seventh Circuit determined that the district court misunderstood the limits of its sentencing authority. Id. at 965. To figure out whether that mistake had affected the defendant’s substantial rights, the Seventh Circuit first scoured the record. Ibid. But reviewing the record didn’t answer the question. Instead, the court discovered that “competing inferences [could] be drawn from the record as to what the sentencing judge might have done had she known” the proper sentencing range. Ibid. Facing uncertainty yet again, the Seventh Circuit reasoned that “[a]s in Paladino, a limited remand is the most prudent way to resolve all doubt on this question.” Id. at 966. That seems much simpler than ordering a full-blown do-over—especially where, as here, it’s not obvious that the district court will do anything differently the second time around. C. As tools for dispelling doubt, limited remands are quite like other tools of the federal courts. Take Pullman abstention. In the case from which that doctrine takes its name, the Supreme Court confronted two issues: one, “a substantial constitutional issue,” and another, a close question of Texas law. R.R. Comm’n of Tex. v. Pullman Co., 312 U.S. 496, 498–99 (1941). The Court thought it wiser to decide the state-law question first because it could obviate the need for a constitutional ruling. Id. at 498. Yet the Court recognized that it was not the body best suited to handle a knotty question of Texas law: “Reading the Texas statutes and the Texas decisions as outsiders without special competence in Texas law, we would have little confidence in our independent judgment regarding the application of that law to the present situation.” Id. at 499. Nor did the Court think it wise for the federal district court to answer that question. Instead, the Court “remand[ed] the cause to the district court, with directions to retain the bill pending a determination of proceedings, to be brought with reasonable promptness, in the state court in conformity with this opinion.” Id. at 501–02. Thus, in the mine-run case, “the proper course in ordering ‘Pullman abstention’ is to remand with instructions to retain jurisdiction but to stay the federal suit pending determination of the state-law questions in state court.” Harris Cty. Comm’rs Court v. Moore, 420 U.S. 77, 88 n.14 (1975). That approach helps federal courts steer clear of some unnecessary constitutional pronouncements by allowing state courts to use their expertise in answering close questions of state law. The practice of certifying questions to state supreme courts provides another analogue. That practice, the Supreme Court has observed, will “in the long run save time, energy, and resources and help[] build a cooperative judicial federalism.” Lehman Bros. v. Schein, 416 U.S. 386, 391 (1974). Certification also yields a more certain answer to questions of state law. See Doe v. Mckesson, 945 F.3d 818, 839 (5th Cir. 2019) (Willett, J., concurring in part and dissenting in part). And mechanically, it’s quite similar to a limited remand. When this court certifies a question to a state supreme court, it “stay[s] its hand” until the state court answers the state-law question. St. Joseph Abbey v. Castille, 700 F.3d 154, 169 (5th Cir. 2012). In some subset of cases, a limited remand likewise allows us to stay our hand until the sentencing court addresses a question it is uniquely well suited to answer. Prejudice from a Guidelines error is one such question. That’s for three reasons. First, just as state courts have expertise in state law, district courts possess expertise in sentencing. In Koon v. United States, the Supreme Court explained that district courts’ “day-to-day experience in criminal sentencing” gave those courts “an institutional advantage over appellate courts” in determining when a particular case justified a departure from the Guidelines. 518 U.S. 81, 98 (1996). And district courts “see so many more Guidelines cases than appellate courts do.” Ibid. So a district court will have a much greater range of comparators on which to draw in determining whether a particular sentence resulted from a miscalculation or not. Second, “[t]he sentencing judge has access to, and greater familiarity with, the individual case and the individual defendant before him than . . . the appeals court.” Rita v. United States, 551 U.S. 338, 357–58 (2007). We’re reviewing the case on a cold record. The defendant didn’t allocute in front of us. We can’t judge tone or demeanor. We’re blind to every detail not captured in the transcript. Finally, a sentencing judge is the world’s leading expert on his own thought process. And that’s the key question here: What was “driving this judge’s decision to impose this sentence for this defendant?” Sanchez-Hernandez, 931 F.3d at 411. In cases like this one, we’re guessing whether we think the Guidelines error affected Rodriguez-Peña’s substantial rights. Why guess when the district court can just tell us? III. At oral argument, we discussed how exactly that tool might be used. The Government adamantly disclaimed the limited remand in this case on the theory that there were no procedural differences between it and a full-blown resentencing. I respectfully disagree. One major difference is that limited remands aren’t subject to Federal Rule of Criminal Procedure 43(a)’s requirement that the defendant be present. See Paladino, 401 F.3d at 484 (noting defendant’s presence is not required on limited remand); United States v. Coles, 403 F.3d 764, 770 (D.C. Cir. 2005) (same); United States v. Crosby, 397 F.3d 103, 120 (2d Cir. 2005) (similar). Instead, limited remands fall under the third exception to that rule. See Coles, 403 F.3d at 770. That exception applies to a “question of law.” FED. R. CRIM. P. 43(b)(3). And “[a]lthough the provision leaves the term ‘question of law’ undefined, the term typically refers to ‘[a]n issue to be decided by the judge, concerning the application or interpretation of the law.’” United States v. Gonzales-Flores, 701 F.3d 112, 116 (4th Cir. 2012) (quoting BLACK’S LAW DICTIONARY 1366 (9th ed. 2009)). How the sentencing judge would (or wouldn’t) have applied the correct Guidelines range is precisely that kind of question of law. We would never require a defendant to be present for our own analysis of this question, and that does not change simply because we send the question back to the sentencing court. Thus a limited remand can avoid the difficulty of transporting a defendant from prison to the sentencing court. A limited remand can also spare the court the time and cost of a hearing with lawyers present. The district court can simply enter a written order stating whether the defendant was prejudiced by any Guidelines error. See United States v. Gomez, 905 F.3d 347, 356 (5th Cir. 2018); Paladino, 401 F.3d at 484 (stating that “[u]pon reaching its decision (with or without a hearing) whether to resentence, the District Court should either place on the record a decision not to resentence, with an appropriate explanation . . . or inform this court of its desire to resentence the defendant” (quotation omitted)). That practice was well-established in the Seventh Circuit before the Supreme Court cited it with approval in Molina-Martinez. And it’s not as if limited remands are wholly foreign to our circuit. We’ve already used them to answer similar questions.[3] For example, in United States v. Gomez, this court determined that a sentencing court thought it lacked discretion in choosing a particular sentence. 905 F.3d at 353–55. The sentencing court was apparently unaware of a then-recent Supreme Court decision clarifying that the sentence was discretionary. In addressing that oversight, we discussed the Seventh Circuit’s limited-remand practice at some length. Id. at 355–56.[4] Then we chose to follow suit: “We remand this case to the district judge for the limited purpose of providing us with an answer to the following question: Do you wish to modify your original sentence in this case in light of Dean?” Id. at 356 (citing Dean v. United States, 137 S. Ct. 1170 (2017)). In much the same way, we have used a limited remand to “clarify[] the district court’s understanding of its discretion under Kimbrough and, if appropriate, its willingness to deviate from the advisory range on such grounds.” See United States v. Malone, 828 F.3d 331, 340 (5th Cir. 2016) (citing Kimbrough v. United States, 552 U.S. 85 (2007)). In that same case, we supported our use of a limited remand by citing Molina-Martinez‘s discussion of the practice. See id. at 340 n.31. And we have plenty of other examples of using limited remands outside the sentencing context. See, e.g., M.D. ex rel. Stukenberg v. Abbott, 929 F.3d 272, 283 (5th Cir. 2019) (Higginbotham, J., concurring in part and dissenting in part) (“Limited remands play a useful, but restricted, role. We grant a limited remand where we task a district court to answer a discrete question necessary for resolution of an issue before us. We retain jurisdiction to enable a return for resolution of those issues yet pending before us.”); see also id. at n.6 (collecting examples of limited remands in the Fifth Circuit). In short, the limited remand is a well-worn tool in our toolkit. * * * In many cases we can figure out the prejudice question on our own. But as between guessing whether the defendant incurred prejudice and just asking the district court the same question, the latter strikes me as better. Otherwise, we’re preferring the shadow to the form. See PLATO, THE REPUBLIC 220–23 (G.R.F. Ferrari ed., Tom Griffith trans., 2000).
