OPINION

KETHLEDGE, Circuit Judge.

Richard Bistline pled guilty to knowingly possessing 305 images and 56 videos of child pornography on his computer. Many, if not a majority, of those images and videos depicted 8- to 10-year-old girls being raped by adult men. Under the Sentencing Guidelines, Bistline's recommended sentence was 63 to 78 months' imprisonment. The district court rejected that recommendation and instead sentenced Bistline to a single night's confinement in the courthouse lockup, plus ten years' supervised release. The United States contends that Bistline's sentence is substantively unreasonable, arguing that the district court improperly rejected the relevant sentencing guideline as "seriously flawed" and that Bistline's sentence fails to reflect the factors recited in the sentencing statute. We agree, and vacate his sentence.

I.

In September 2007, law-enforcement agents downloaded 12 images of child pornography from an IP address in Mount Vernon, Ohio. The IP address was Bistline's. The agents were able to download the images because Bistline had placed them in his "shared" files on a "peer to peer" internet program. A month later, agents arrived at Bistline's home to execute a search warrant there. He was outside mowing the lawn when the agents arrived. Their search revealed the images and videos described above.

Bistline later pled guilty to one count of possession of child pornography in violation of 18 U.S.C. § 2522. His guidelines range, as noted above, was 63 to 78 months' imprisonment. But the probation officer recommended a sentence of only 24 months, citing the facts that Bistline was 67 years old, had no prior criminal convictions, had suffered two strokes within the preceding 11 years, and was to some extent a caretaker for his wife. The United States opposed that recommendation in a detailed memorandum that argued in support of a sentence within the guidelines range.

The district court held a sentencing hearing. Near its outset, the court said that "the guidelines for possession of child pornography are seriously flawed" as a result of Congress's involvement in them. The court then contrasted possession of child pornography with some related offenses —such as its distribution—that the court regarded as more serious. The court also emphasized Bistline's age, health, and putative need to care for his wife. Eventually the court announced its intention not to imprison Bistline at all, but instead to confine him overnight in the courthouse lockup. The government responded with a lengthy objection detailing what it thought were the defects in the district court's reasoning under the sentencing statute. The government also asked for an opportunity to "brief the issue of disregarding the [child pornography] guidelines," since the court had not earlier informed the parties that it planned to disregard them. The court allowed the parties to brief that issue.

Two months later, the court held another sentencing hearing. It had not changed its mind in the meantime. At the hearing's close, once again over the government's detailed objections, the court imposed a sentence of overnight confinement in the courthouse lockup, to be followed by 10 years of supervised release.

The United States brought this appeal.

II.

The government argues that Bistline's sentence is substantively unreasonable. "[A] sentence may be substantively unreasonable when the district court selects the sentence arbitrarily, bases the sentence on impermissible factors, fails to consider pertinent [18 U.S.C.] § 3553(a) factors or gives an unreasonable amount of weight to any pertinent factor." United States v. Borho, 485 F.3d 904, 908 (6th Cir.2007) (internal quotation marks omitted).

Although the Sentencing Guidelines are now only advisory, they still "should be the starting point and the initial benchmark" for choosing a defendant's sentence. Gall v. United States, 552 U.S. 38, 49, 128 S.Ct. 586, 169 L.Ed.2d 445 (2007); see generally United States v. Booker, 543 U.S. 220, 246-58, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005) (striking down the mandatory guidelines regime on Sixth Amendment grounds, not on grounds of disagreement with the policies animating the guidelines regime). If the district court decides to impose a sentence outside the applicable guideline range, the court must "ensure that the justification is sufficiently compelling to support the degree of the variance. We find it uncontroversial that a major departure should be supported by a more significant justification than a minor one." Gall, 552 U.S. at 50, 128 S.Ct. 586.

Here, the district court's departure could not have been any greater, since it varied downward to an essentially non-custodial sentence. We turn to its justifications for doing so.

A.

Perhaps the keystone of the district court's reasoning was its rejection of the relevant sentencing guideline, § 2G2.2, as "seriously flawed." Our court has said that a district court may disagree with § 2G2.2 on policy grounds, just as it may any other. See United States v. Brooks, 628 F.3d 791, 799 (6th Cir.2011). But if a district court chooses to disagree with a guideline, we will "scrutinize closely" its reasons for doing so. United States v. Herrera-Zuniga, 571 F.3d 568, 585 (6th Cir.2009). We now scrutinize the district court's reasons for rejecting § 2G2.2 here.

"Congress has taken an active role" in crafting § 2G2.2. United States v. McNerney, 636 F.3d 772, 775 (6th Cir.2011). Indeed, on numerous occasions Congress has amended this guideline directly or through mandates to the Sentencing Commission. See United States v. Pugh, 515 F.3d 1179, 1197-98 (11th Cir.2008) (collecting amendments). That, in the district court's view, was the problem: the court said these amendments "were not arrived at through empirical study and data, but in some instances are a reflection of congressional mandates. And that gives the court some concern that political considerations may well have influenced the severity of these guidelines."

The court's concern about "congressional mandates" was misguided. "In our system, so far at least as concerns the federal powers, defining crimes and fixing penalties are legislative ... functions." United States v. Evans, 333 U.S. 483, 486, 68 S.Ct. 634, 92 L.Ed. 823 (1948). To some extent, Congress has delegated this power to the Sentencing Commission. See generally 28 U.S.C. § 994. But "[t]he Sentencing Commission unquestionably is a peculiar institution within the framework of our Government." Mistretta v. United States, 488 U.S. 361, 384, 109 S.Ct. 647, 102 L.Ed.2d 714 (1989). So peculiar, in fact, that the Commission is something of a constitutional centaur, half judicial and half legislative, "an unusual hybrid in structure and authority[.]" Id. at 412, 109 S.Ct. 647. For although located in the judicial branch, "the Commission is not a court [and] does not exercise judicial power[.]" Id. at 393, 109 S.Ct. 647. Instead its power is "quasi-legislative[,]" id., which of course is an odd power for a body within the judicial branch to have. The Supreme Court ultimately permitted this arrangement; but it did so only after acknowledging that "the unique composition and responsibilities of the Sentencing Commission give rise to serious concerns about a disruption of the appropriate balance of governmental power among the coordinate [b]ranches[.]" Id. at 384, 109 S.Ct. 647.

For our purposes, however, the relevant point is one of constitutional context: namely, that the Constitution merely tolerates, rather than compels, Congress's limited delegation of power to the Commission. And that context, we think, puts in a different light the various complaints, both within and without the judiciary, that Congress has encroached too much on the Commission's authority with respect to sentencing policy. That is like saying a Senator has encroached upon the authority of her chief of staff, or a federal judge upon that of his law clerk. And thus we disagree with the complaint—to the extent it is one—that Congress's amendments to § 2G2.2 "evince a `blatant' disregard for the Commission and are `the most significant effort to marginalize the role of the Sentencing Commission in the federal sentencing process since the Commission was created by Congress[.]'" United States v. Dorvee, 616 F.3d 174, 185 (2d Cir.2010) (quoting Vinegrad, The New Federal Sentencing Law, 15 Fed. Sent. R. 310, 315 (June 2003)). Congress can marginalize the Commission all it wants: Congress created it. Indeed it is normally a constitutional virtue, rather than vice, that Congress exercises its power directly, rather than hand it off to an unelected commission. The Constitution is fundamentally a democratic document, not a technocratic one.

The point more specifically is that Congress delegated to the Commission a limited measure of its power to set sentencing policy, and retained for itself the remainder. It is not the judiciary's province to say that Congress should have delegated still more—especially to another body within the judicial branch. We think it follows that a district court cannot reasonably reject § 2G2.2—or any other guidelines provision—merely on the ground that Congress exercised, rather than delegated, its power to set the policies reflected therein. That is not to say that a district court must agree with a guideline in which Congress has played a direct role. It is only to say that the fact of Congress's role in amending a guideline is not itself a valid reason to disagree with the guideline. The district court cited that reason here; and that reason cannot support its decision to reject § 2G2.2.

Neither can the district court's next reason, which was that "political considerations may well have influenced" the content of § 2G2.2, given Congress's involvement in amending it. This reason is merely a proxy for the previous one. To disable a political branch from acting on political considerations (which, less pejoratively, are oftentimes democratic considerations) is to disable it from acting at all. The district court's reasoning would have that effect here; and thus it would force Congress to yield still more power to the Commission. Suffice it to say that the courts cannot bar Congress from acting on political considerations, any more than Congress can bar the courts from acting on legal ones. Each branch is entitled to act according to its nature. That Congress did so here is no reason for a district court to reject § 2G2.2.

Equally without merit is the district court's criticism that "the most recent changes to [§ 2G2.2] apparently came from two lawyers in the Justice Department who persuaded a novice congressman to add them to the popular Amber Alert bill." As an initial matter, this kind of urban mythology is no reason to reject a guideline on policy grounds. And the criticism is just as misguided when considered on its own terms. The only antecedent circumstances relevant to the validity of Congress's directives with respect to § 2G2.2 are those spelled out in the Constitution itself: bicameralism and presentment. See U.S. Const. Art. I, sec. 7. Here, the directives met those requirements, which means they became law not because they were approved by a novice congressman, but because they were part of legislation approved by both Houses of Congress and then signed by the President. What came before then is no business of the courts. As noted above, a district court may still disagree with the policies embodied in those directives; but to survive the close scrutiny that follows, the court must explain its disagreement in terms that are persuasive on policy grounds, not political ones.

The district court's remaining criticism of § 2G2.2 is that it was "not arrived at through the ordinary deliberations of the Sentencing Commission," and specifically that it was "not arrived at through empirical study and data[.]" This reasoning tries to draft behind the Supreme Court's reasoning in Kimbrough v. United States, 552 U.S. 85, 128 S.Ct. 558, 169 L.Ed.2d 481 (2007). It is therefore worth comparing that case to this one.

Kimbrough presented the archetypal case in which a district court could disagree with a guideline on policy grounds. The case concerned the crack-cocaine guidelines, under which "a drug trafficker dealing in crack cocaine is subject to the same sentence as one dealing in 100 times more powder cocaine." 552 U.S. at 91, 128 S.Ct. 558. The Court observed that, in adopting the ratio, the Commission "did not take account of `empirical data and national experience.'" Id. at 109, 128 S.Ct. 558 (quoting United States v. Pruitt, 502 F.3d 1154, 1171 (10th Cir.2007) (McConnell, J., concurring)). Instead, the Commission had simply lifted the ratio off the rack of another, inapposite statutory provision. See Kimbrough, 552 U.S. at 96-97, 128 S.Ct. 558. (Importantly for our purposes, the Court made clear that the Commission's decision to adopt the 100-1 ratio could not be "implicitly" attributed to Congress—and thus remained the Commission's own—even though the ratio had first been set forth by Congress in the other, inapposite provision. See id. at 102, 128 S.Ct. 558 (internal punctuation omitted).) And the Court noted further that the Commission had come to regret its decision, as "it later determined that the crack/powder sentencing disparity is generally unwarranted." Id. at 97, 128 S.Ct. 558. "Given all this," the Court held, "it would not be an abuse of discretion for a district court to conclude when sentencing a particular defendant that the crack/powder disparity yields a sentence `greater than necessary' to achieve § 3553(a)'s purposes, even in a mine-run case." Id. at 110, 128 S.Ct. 558.

This case is different. It is true that the Commission did not act in its usual institutional role with respect to the relevant amendments to § 2G2.2. But that is because Congress was the relevant actor with respect to those amendments; and that puts § 2G2.2 on stronger ground than the crack-cocaine guidelines were on in Kimbrough. The reasons again reach back to the Constitution. The Commission's expertise is primarily empirical, see Kimbrough, 552 U.S. at 109, 128 S.Ct. 558; and thus, when the Commission itself makes a policy decision for reasons that lie outside its expertise, the guideline that results might be vulnerable on precisely that ground. Such was the case in Kimbrough. But Congress's power to make sentencing policy does not arise from a delegation from anyone, much less a delegation made on grounds of a particular expertise. That power instead flows directly from the Constitution, without strings other than those contained in the Constitution itself. And nothing in that document confines the exercise of Congress's sentencing power to empirical grounds alone. Of course, Congress can choose to act on empirical grounds; and when it does, a district court must contend with those grounds in explaining its disagreement (in the case of a guideline) with Congress's policy judgment. But it is also Congress's prerogative to dictate sentencing enhancements based on a retributive judgment that certain crimes are reprehensible and warrant serious punishment as a result. When a congressional directive reflects such a judgment, a district court that disagrees with the guideline that follows must contend with those grounds too. Thus, when a guideline comes bristling with Congress's own empirical and value judgments—or even just value judgments—the district court that seeks to disagree with the guideline on policy grounds faces a considerably more formidable task than the district court did in Kimbrough.

A court that disagrees with § 2G2.2 must take on this formidable task. Congress's long and repeated involvement in raising the offense levels for § 2G2.2 makes clear that the grounds of its action were not only empirical, but retributive— that they included not only deterrence, but punishment. See generally United States Sentencing Commission, The History of the Child Pornography Guidelines (Oct. 2009); id. at 54 ("Congress has demonstrated its continued interest in deterring and punishing child pornography offenses[.]"); PROTECT Act, Pub.L. No. 108-21, § 513(c) (2003) (directing the Commission "to ensure that the guidelines are adequate to deter and punish" child-pornography offenses); United States v. Morace, 594 F.3d 340, 347 (4th Cir.2010) (noting "Congress'[s] view that child pornography crimes are serious offenses deserving serious sanctions") (internal punctuation omitted).

It simply misses the point, therefore, to say that the Commission departed from its usual role in the case of § 2G2.2. Congress is the body that dictated numerous enhancements to that provision over the past two decades; and thus, with respect to those enhancements at least, it is Congress's reasons that a district court must refute before declining to apply § 2G2.2 out of hand. The district court did not seriously attempt to refute them here. Bistline's guidelines range therefore remains the starting point for his sentence. Gall, 552 U.S. at 49, 128 S.Ct. 586.

B.

We next consider whether Bistline's sentence was reasonable in light of the sentencing factors recited in § 3553(a).

1.

Prominent among the § 3553(a) factors is the "need for the sentence imposed ... to reflect the seriousness of the offense[.]" 18 U.S.C. § 3553(a)(2)(A). Here, the offense is knowing possession of child pornography in violation of 18 U.S.C. § 2252(a)(4)(B).

The district court made a number of observations with respect to the seriousness of this offense. Many of them served to diminish it. The court did say that the images on Bistline's computer were "horrendous," and that the "production of child pornography and the distribution of it is an extremely serious offense, one which should be punished accordingly." But notably omitted from that recitation (and virtually unpunished in this case) was the crime of possession of child pornography. Indeed, the court said there are "significant differences ... in the degree of culpability in the chain of events that leads to the display of child pornography[,]" with the "most culpable" persons being "those who are involved in actually performing these acts and photographing them." We agree with that statement so far as it goes. That the producers of child pornography are more culpable, however, does not mean that its knowing and deliberate possessors are barely culpable at all.

Yet the court took some long strides towards that view in explaining its decision not to sentence Bistline to a single day in prison. (The courthouse lockup is not prison.) The court said that "[t]he technology that we are presented with today has significantly changed the landscape as it relates to pornography in general, and child pornography specifically"; that "even the worst kind of child pornography is available at the mere click of a mouse through the internet without any payment whatsoever"; and that, "[i]n fact, I am sure that in some instances that it is displayed without any request in the form of viruses that are distributed or so-called spam distributions." None of those instances happened here (and nor would they violate the statute). As to what did happen, the court said that Bistline "hooked up to a free, on-line program called LimeWire, which provided ready access to this kind of material without cost, and it is obvious that he got caught up in it and permitted it to be downloaded onto his computer."

It is little wonder, then, that Bistline was not sentenced to a day in prison. We can only read the court's comments—as to both possession of child pornography generally and Bistline's conduct specifically— to reflect a sense that the offense is not serious, and that the hundreds of images found on Bistline's computer—a great many of them showing young girls being raped—somehow sought out Bistline more than he sought out them. We reject this sense of Bistline's crime. Knowing possession of child pornography in violation of § 2252(a)(4)(B) is not a crime that happens to a defendant. It is not a crime of inadvertence, of pop-up screens and viruses that incriminate an innocent person. Possession of child pornography instead becomes a crime when a defendant knowingly acquires the images—in this case, affirmatively, deliberately, and repeatedly, hundreds of times over, in a period exceeding a year.

The district court separately reasoned that "the Indictment and the conviction of this defendant reflect the seriousness of this offense"; that "the potential penalties that he is facing reflect the seriousness of the offense"; and that "the consequences of the guilty plea and conviction, which include registration as a sexual offender with the State of Ohio and the publication of that registration to the community and to his friends and neighbors also reflect[] the seriousness of the offense." But § 3553(a)(2)(A) plainly states that "the sentence imposed" should "reflect the seriousness of the offense," (emphasis added); and none of these things are Bistline's sentence. Nor are they consequences of his sentence, as opposed to consequences of his prosecution and conviction. The district court's recitation of these collateral consequences therefore does nothing to show that Bistline's sentence reflects the seriousness of his offense. Were it otherwise, these sorts of consequences—particularly ones related to a defendant's humiliation before his community, neighbors, and friends—would tend to support shorter sentences in cases with defendants from privileged backgrounds, who might have more to lose along these lines. And "[w]e do not believe criminals with privileged backgrounds are more entitled to leniency than those who have nothing left to lose." United States v. Stall, 581 F.3d 276, 286 (6th Cir.2009). It is true that the district court also cited Bistline's term of supervised release as something that reflects the seriousness of his offense; and that term is part of the sentence itself. But a term of supervised release is simply not enough to reflect the seriousness of the offense here.

That point is underscored by a victim statement offered by the government in this case. Among the 305 images of hard-core child pornography on Bistline's computer were some that depicted a ten year-old girl being raped by her father. That same girl, now a young woman, specifically asked that her statement be read aloud at Bistline's sentencing hearing. It was. She stated in part:

Bistline's crime inflicted great harm upon its victims. The district court's explanation of his sentence did not include any recognition of that fact. Neither does the sentence itself. Bistline's sentence does not reflect the seriousness of his offense.

2.

We discuss some other factors more briefly. One of them is "the need for the sentence imposed ... to afford adequate deterrence to criminal conduct[.]" 18 U.S.C. § 3553(a)(2)(B). The district court thought that factor was satisfied here, at least as to Bistline personally, because of "the humiliation of his arrest and the prosecution in this court with his family and close friends and the necessity that he register as a sex offender in this community[.]" But none of those consequences arise from Bistline's sentence, as opposed to his prosecution and conviction; and thus none of them count for purposes of § 3553(a)(2)(B). See supra at 765-66. The district court otherwise said that "[g]eneral deterrence, I believe, will have little [if] anything to do with this particular case." That statement is inexplicable, and in any event conflicts with our statement that "[g]eneral deterrence is crucial in the child pornography context[.]" United States v. Camiscione, 591 F.3d 823, 834 (6th Cir.2010). Suffice it to say that we do not see how Bistline's sentence would meaningfully deter him or anyone else.

Another factor is "the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct[.]" 18 U.S.C. § 3553(a)(6). The sentencing guidelines already take account of Bistline's lack of a prior criminal record and specific conduct in violating § 2252(a)(4)(B). And yet his guidelines range was 63 to 78 months' imprisonment. We do not mean to imply that only a sentence in or around that range will avoid disparities with other similar defendants. But we do not see how the sentence imposed here avoids them.

The final factor we discuss here is "the history and characteristics of the defendant." 18 U.S.C. § 3553(a)(1). As to this factor, the district court said that Bistline's "age, his otherwise unblemished record as a productive citizen, his health, and his family circumstances, specifically, the need for his assistance to care for his wife, all call for mitigation of his sentence[.]" Although the district court was entitled to consider these circumstances, they cannot justify the sentence imposed here. On this point United States v. Christman, 607 F.3d 1110 (6th Cir.2010), is controlling. The facts there are remarkably similar to those here. Christman pled guilty to possession of child pornography; his guidelines range was 57 to 71 months' imprisonment; and the district court sentenced him to only five days in prison, citing among other circumstances that "Christman had severe back pain, he was his elderly, ailing mother's primary caregiver, [and] his family believed that he was remorseful[.]" Id. at 1112. We held that Christman's sentence was substantively unreasonable, noting that a defendant's age, physical condition, and family responsibilities are all discouraged factors under the guidelines. Id. at 1119 (citing U.S.S.G. ch. 5, part H). Although we did recognize that the guidelines are only advisory, we said that "section 3553(a)(5) requires that the district court consider applicable policy statements issued by the Sentencing Commission" and that "the Guidelines' disfavored view towards age and physical condition is, therefore, relevant to our reasonableness review." Id. (internal citations omitted).

The district court did not consider any of those policy statements here. In addition, the court overlooked that Bistline has four adult children who live not far from his home and who presumably could help care for his wife. In arguably more difficult circumstances than this, we rejected the same caregiver rationale in Christman. See id. at 1120. Moreover, the court here did not address the government's challenges to Bistline's assertions about his condition and that of his wife, instead accepting them at face value.

There is also an important distinction between Christman and this case. There, it appeared that "Christman was remorseful for his crimes[.]" Id. at 1116. We cannot say the same here. Although Bistline did express "regret" for his crime— not always the same thing as remorse—he never expressed (on the record at least) any recognition of the harm it inflicted on Kylie and others like her. To the contrary, Bistline said he could not figure out why it was illegal to possess child pornography; and he expressed genuine anger that the government had taken his downloaded music along with his computer, and that (as a convicted felon) he could no longer possess firearms. The sentence imposed in this case only validated these perceptions. Bistline's sentence is substantively unreasonable.

3.

It remains only to explain why our decision in this case is not controlled by our decision in Stall, where we affirmed a sentence very similar to this one. The basic explanation is the same one we recited in Christman: in Stall we reviewed the sentence only for plain error, whereas here we do not. We also note the following facts about Stall: "the government submitted a cursory, two-page memorandum" in support of its proposed sentence, 581 F.3d at 280, "the government did not cite a single case in either its Sentencing Memorandum or at the sentencing hearing[,]" id. at 283, and "the government at sentencing put forward almost no evidence for why a sentence within the Guidelines was warranted and did not raise the same cogent arguments it presents only on appeal," id. at 278. The contrast here could hardly be more stark: the government submitted a detailed memorandum in support of its proposed sentence before the first sentencing hearing, submitted yet another after that hearing, presented extensive argument and evidence at both hearings, and objected on the spot to the court's announcement of sentence with a recitation of reasons that were virtually an outline of its arguments on appeal.

In Stall, we suggested that the sentence there was attributable in significant part to "the failure of prosecutors to defend their sentencing recommendations vigorously." Id. at 286. Having carefully reviewed our opinion in that case, we concur in that suggestion. Stall is more a cautionary tale about prosecutorial neglect, than it is a precedent important to our decision here.

* * *

In fairness to the district court, its decision here came before our decisions in McNerney, Camiscione, and Christman. But the fact remains that the sentence imposed in this case does not remotely meet the criteria that Congress laid out in § 3553(a). We vacate Bistline's sentence and remand his case for prompt imposition of one that does.