United States v. Samuel Mullet, Sr.

*589OPINION

A string of assaults in several Amish communities in Ohio gave rise to this prosecution under Section 2 of The Matthew Shepard and James Byrd, Jr. Hate Crimes Prevention Act of 2009. The assaults were not everyday occurrences, whether one looks at the setting (several normally peaceful Amish communities), the method of attack (cutting the hair and shaving the beards of the victims), the mode of transportation to them (hired drivers), the relationship between the assailants and their victims (two of them involved children attacking their parents), or the alleged motive (religious-based hatred between members of the same faith). A jury found that four of the five attacks amounted to hate crimes under the Act and convicted sixteen members of the Bergholz Amish community for their roles in them.

At stake in this appeal is whether their hate-crime convictions may stand. No one questions that the assaults occurred, and only a few defendants question their participation in them. The central issue at trial was whether the defendants committed the assaults “because of’ the religion of the victims. 18 U.S.C. § 249(a)(2)(A). In instructing the jury on this point, the district court rejected the defendants’ proposed instruction (that the faith of the victims must be a “but for” cause of the assaults) and adopted the government’s proposed instruction (that the faith of the victims must be a “significant factor” in motivating the assaults). Regrettably for all concerned, a case decided after this trial confirms that the court should have given a but-for instruction on causation in the context of this criminal trial. Burrage v. United States, — U.S.-, 134 S.Ct. 881, 887-89, 187 L.Ed.2d 715 (2014). Because this error was not harmless, and indeed went to the central factual debate at trial, we must reverse these convictions.

I.

In 1995, Samuel Mullet bought land in Jefferson County, Ohio. That land became the Bergholz Amish community in 2001, when a sufficient number of ordained ministers qualified it as a separate Amish church district. The new community appointed Samuel as its bishop. As bishop, Samuel controlled all aspects of life in the Bergholz compound and had the ability to order the “shunning”—excommunication— of community members who failed to follow the tenets of their Amish faith. R. 540 at 292.

In 2006, Samuel excommunicated several church members who questioned Ber-gholz community practices and his leadership. Included in the group were Lavern and Mattie Troyer, whose son Aden was married to Samuel’s daughter Wilma, as well as Melvin and Anna Shrock, whose son Emanuel was married to Samuel’s daughter Linda. The excommunications were not good for relationships between and within the affected families. In one case, they led to a divorce. Aden left Wilma to join his parents in a Pennsylvania Amish community after unsuccessfully trying to convince her to join him. In another case, they led to parent-child animosity. Emanuel refused to leave Ber-gholz with his parents despite their repeated efforts to persuade him to do so.

The Bergholz excommunications also tested church doctrine. Amish communities as a general rule practice strict shunning, meaning that if one Old Order Amish community excommunicates a community member, all other Old Order communities must excommunicate him until he obtains forgiveness from the community that first shunned him. The Bergholz excommunications proved to be an exception. After *590fleeing Bergholz for another community in Pennsylvania, the Troyers asked not to be subject to the strict-shunning rule. Citing unusual practices in Bergholz, the Troyers asked their new bishop to admit them to the Pennsylvania church without requiring them to seek forgiveness from Samuel— the Elmer Gantry of the Amish community to their mind. Given the number of former Bergholz residents in similar situations, Amish bishops from all over the country met in September 2006 to address the issue. Three hundred bishops convened, and they voted unanimously to reverse the Bergholz excommunications.

At the same time that the ruling allowed the Troyers to settle into their new Pennsylvania community, it also exacerbated a custody battle between Wilma and Aden over their two children. The dispute began when a SWAT team took the children under an emergency temporary custody order issued to Aden. It ended two years, and one trial, later when Aden’s temporary custody of the children became permanent in an order declaring that “[a]ll parenting time shall be in Pennsylvania. Under no circumstances shall parenting time take place in Bergholz, Ohio.” Id. at 152.

Losing Wilma’s children brought the Bergholz community to its knees and sparked a change in their faith-based traditions. Typically, Amish men do not trim their beards and Amish women do not cut their hair as a way of symbolizing their piety, demonstrating righteousness and conveying an Amish identity. Believing that the loss of Wilma’s children resulted from their lack of faith, several Bergholz residents cut their own hair and trimmed their own beards as a way to atone for their sins. The Bergholz community saw these acts as penance and as a symbol of rededication to their faith.

The Bergholz community did not confine this ritual to their own ranks. They also used it to punish or harm others who were not members of the church district. From September 6 to November 9, 2011, several Bergholz community members committed five separate attacks on nine different individuals, slicing off the men’s beards and cutting the women’s hair. Religious and personal ties connected the nine victims of these attacks to the Bergholz community. Some were parents of Bergholz residents, some were friends, and some were associated with family members who had left Bergholz for other Amish districts. Also linking the victims was that they participated in overturning the Bergholz excommunications and that, in the eyes of the assailants, they were “Amish hypocrites.” R. 539 at 35; R. 540 at 11-12.

A federal grand jury indicted sixteen members of the Bergholz community for violating, and conspiring to violate, the Hate Crimes Prevention Act: Samuel Mullet, Johnny Mullet, Daniel Mullet, Lester Mullet, Levi Miller, Eli Miller, Emanuel Shrock, Lester Miller, Raymond Miller, Freeman Burkholder, Anna Miller, Linda Shrock, Lovina Miller, Kathryn Miller, Emma Miller and Elizabeth Miller. It also indicted Samuel Mullet, Levi Miller, Eli Miller and Lester Mullet for concealing evidence and indicted Samuel Mullet for making false statements to the FBI.

In responding to the hate-crime charges at trial, none of the Bergholz defendants disputed that the assaults happened, and few disputed that they participated in them. To prove that the defendants’ actions amounted to a federal hate crime, though, the prosecution had to show that the defendants assaulted the victims and that they assaulted the victims “because of’ their religious beliefs. See 18 U.S.C. § 249(a)(2)(A). That extra burden gave rise to a central issue at trial: Why did the defendants assault these individuals? The defendants argued that a mix of interper*591sonal issues—parental mistreatment, personality conflicts, harassment, power struggles, and interference with family relationships—motivated the assaults. The prosecution argued that faith—a desire to punish those whom Samuel saw as Amish hypocrites and who did not properly practice their Amish faith—motivated the assaults.

The jury sided with the prosecution on four assaults and with the defense on one of them. All told, it convicted all sixteen defendants of at least one violation of the hate-crime statute. It also convicted three of the defendants (Samuel Mullet, Eli Miller and Lester Mullet) on the concealing-evidence charge and convicted Samuel Mullet on the false-statements charge. All sixteen defendants appeal their hate-crime convictions. None of the defendants challenges their convictions for concealing evidence and lying to the FBI.

II.

The Bergholz defendants raise a number of issues on appeal. Because our resolution of one issue—whether the district court properly instructed the jury on the motive element of the crime—makes it unnecessary to address most of the others, we start with that claim of error.

The federal hate-crime statute prohibits “willfully causing] bodily injury to any person ... because of the actual or perceived ... religion ... of [that] person.” 18 U.S.C. § 249(a)(2)(A). Of note here, the crime contains a motive element, requiring the government to show that the defendant attacked the victim “because of’ the victim’s “actual or perceived” religion. Id. The district court instructed the jury that the motive element could be satisfied by showing that “a person’s actual or perceived religion was a significant motivating factor for a [defendant's action” “even if he or she had other reasons for doing what he or she did as well.” R. 542 at 28-29. In taking issue with this instruction, the defendants argue that the phrase “because of’ requires but-for causation—a showing that they would not have acted but for the victim’s actual or perceived religious beliefs. The defendants have the better of the argument.

In everyday usage, the phrase “because of’ indicates a but-for causal link between the action that comes before it and the circumstance that comes afterwards. John carried an umbrella because of the rain. Jane stayed home from school because of her fever. Dictionary definitions of the phrase reflect this common-sense understanding: “Because of’ means “by reason of’ or “on account of’ the explanation that follows. Webster’s Second New International Dictionary 242 (1950); see also Oxford English Dictionary, “because ” (2012). Put in the context of this statute, a defendant “causes bodily injury to a[ ] person ... because of [that person’s] actual or perceived ... religion,” 18 U.S.C. § 249(a)(2)(A), when the person’s actual or perceived religion was a but-for reason the defendant decided to act.

Consistent with these definitions, the Supreme Court has “insiste[d]” that “statutes using the term ‘because of ” require a showing of “but-for causality.” Burrage, 134 S.Ct. at 889. It has applied this requirement in criminal and civil cases alike. See id. at 888-89 (criminal); Univ. of Tex. Sw. Med. Ctr. v. Nassar, — U.S. -, 133 S.Ct. 2517, 2528, 186 L.Ed.2d 503 (2013) (civil); Gross v. FBL Fin. Servs., Inc., 557 U.S. 167, 176-77, 129 S.Ct. 2343, 174 L.Ed.2d 119 (2009) (civil); Safeco Ins. Co. of Am. v. Burr, 551 U.S. 47, 63-64 & n. 14, 127 S.Ct. 2201, 167 L.Ed.2d 1045 (2007) (civil). And it maintains this requirement regardless of whether “because of’ refers to an easier-to-show prohibited act or a harder-to-prove prohibited motive. See *592Burrage, 134 S.Ct. at 887 (the prohibited act, “distribution of heroin,” must be a but-for cause of death); Nassar, 133 S.Ct. at 2528 (the prohibited motive, “the desire to retaliate,” must be a but-for cause of the adverse employment action); Gross, 557 U.S. at 177, 129 S.Ct. 2343 (the prohibited motive, the employee’s “age,” must be a but-for cause of the adverse employment action). A defendant thus “causes bodily injury to a[ ] person ... because of [that person’s] actual or perceived ... religion,” 18 U.S.C. § 249(a)(2)(A), when the person’s actual or perceived religion was “the ‘reason’ ” the defendant decided to act, Gross, 557 U.S. at 176, 129 S.Ct. 2343— that is, “the straw that broke the camel’s back,” Burrage, 134 S.Ct. at 888.

Our court has said the same thing. We have held that the Court’s interpretation of “because of’ in prior cases “points the way” to the correct interpretation of that same phrase in later cases dealing with different statutes. Lewis v. Humboldt Acquisition Corp., 681 F.3d 312, 321 (6th Cir.2012) (en banc). “[B]ecause of’ in brief means what it says: The prohibited act or motive must be an actual cause of the specified outcome.

That conclusion makes good sense in the context of a criminal case implicating the motives of the defendants. The alternative proposed definition of the phrase (“significant motivating factor”) does not sufficiently define the prohibited conduct. How should a jury measure whether a specific motive was significant in inspiring a defendant to act? Is a motive significant if it is one of three reasons he acted? One of ten? “Uncertainty of [this] kind cannot be squared with the beyond-a-reasonable-doubt standard applicable in criminal trials or with the need to express criminal laws in terms ordinary persons can comprehend.” Burrage, 134 S.Ct. at 892 (rejecting the “substantial” or “contributing” factor test). Even if there were some doubt over which of these definitions Congress had in mind, which we do not think there is, the rule of lenity would require us to adopt the more lenient of the two in a criminal case. See id. at 891; see also id. at 892 (Ginsburg, J., concurring in the judgment). In point of fact, the members of the Court who do not think “because of’ means but-for causation in the setting of a civil statute, Nassar, 133 S.Ct. at 2546 (Ginsburg, J., dissenting), agree that it requires but-for causation in the setting of a criminal statute in view of the rule of lenity, see Burrage, 134 S.Ct. at 892 (Ginsburg, J., concurring in the judgment).

Any standard that requires less than but-for causality, moreover, treads uncomfortably close to the line separating constitutional regulation of conduct and unconstitutional regulation of beliefs. The government may punish “bias-inspired conduct ” without offending the First Amendment because bigoted conduct “inflict[s] greater individual and societal harm.” Wisconsin v. Mitchell, 508 U.S. 476, 487-88, 113 S.Ct. 2194, 124 L.Ed.2d 436 (1993) (emphasis added). But punishment of a defendant’s “abstract beliefs,” no matter how “morally reprehensible” they may be, violates the First Amendment. See Dawson v. Delaware, 503 U.S. 159, 167, 112 S.Ct. 1093, 117 L.Ed.2d 309 (1992). Requiring a causal connection between a defendant’s biased attitudes and his impermissible actions ensures that the criminal law targets conduct, not bigoted beliefs that have little connection to the crime.

What seems clear to us today, we must acknowledge, might not have looked as clear at the time of trial. It was not until 2014, when the Court decided Burrage, that several contours of this debate came into focus. In deciding that the phrase “results from” in a criminal statute re*593quires but-for causation, Burrage made several points directly applicable here. It noted that, under dictionary definitions, “results from” ordinarily “imposes ... a requirement of actual causality.” 134 S.Ct. at 887. It noted that, under its decisions in Nassar, Gross and Burr, “results from” and “because of’ customarily mean the same thing and that both phrases require but-for causation. Id. at 888-89. It cited decisions from other courts that reached the same conclusion, including one similar to this case—an Iowa Supreme Court decision interpreting “because of’ in the motive element of Iowa’s hate-crime statute to require a showing of but-for causation. Id. at 889 (citing State v. Hennings, 791 N.W.2d 828, 833-35 (Iowa 2010)). And it relied on the criminal setting of the statute to require but-for causation rather than some lesser causal connection between the defendant’s actions and the victim’s death. Id. at 891-92. Just so here: The “because of’ element of a prosecution under the Hate Crimes Act requires the government to establish but-for causation.

III.

The government challenges this conclusion on several grounds, each unconvincing.

A.

The government maintains that the defendants forfeited this argument and that the strictures of plain-error review thus apply. No forfeiture occurred. The district court addressed jury instructions twice, at the beginning and the end of the trial. Before trial, the district court acknowledged that “[s]ome [defendants] suggested ... the ‘but for’ ” instruction. R. 314 at 26. It then rejected the request, reasoning that it would be “impossible” “to ask the jury to determine beyond a reasonable doubt the defendants] would not have done this but-for the victim’s religion or perceived religion.” Id. at 25-26. The defendants renewed this objection as the trial came to a close. The court’s response remained the same: It “underst[oo]d ... that the [defendants [we]re requesting a ‘but for’ instruction as a fall back from exclusive motivation, and ... declined] to give that” because it “d[id]n’t quite see how [the government] could prove [‘but for’ causation] beyond a reasonable doubt.” R. 541 at 247-48. Twice the defendants objected to the district court’s explanation of the motive requirement, twice they requested a “but for” instruction, and twice the district court rebuffed their request. The defendants sufficiently “inform[ed] the court of thefir] specific objection and the grounds for th[at]' objection,” Fed. R.Crim.P. 30(d), and thus preserved the error for appeal.

B.

Turning to the merits, the government contends that, before passage of the Hate Crimes Prevention Act in 2009, other courts had construed “because of’ to mean “significant motivating factor” in related statutes and that Congress meant to incorporate that definition here. Appellee Br. 105-08. But this argument overlooks the vintage of the cited cases—United States v. McGee, 173 F.3d 952 (6th Cir.1999), United States v. Ebens, 800 F.2d 1422 (6th Cir.1986), and United States v. Bledsoe, 728 F.2d 1094 (8th Cir.1984)—-which all predate Gross and Burrage. And it overlooks the reality that none of the three cases addressed the possibility that “because of’ required but-for causation, see McGee, 173 F.3d at 957; Ebens, 800 F.2d at 1429; Bledsoe, 728 F.2d at 1097-98, making it difficult to understand how they could have provided a model for the 2009 law on this issue.

*594The argument also overlooks the reality that two Supreme Court eases—Gross and Burr—predated the passage of the federal hate-crime law and were assuredly more conspicuous exemplars of what Congress meant when it used “because of’ language in § 249(a)(2)(A). See Gross, 557 U.S. at 176, 129 S.Ct. 2343; Burr, 551 U.S. at 63-64 & n. 14, 127 S.Ct. 2201. And it overlooks other cases dealing with the motivation requirement in hate-crime statutes. Before 2009, several state courts construed “because of’ language in a similar setting to this law—state hate-crime laws—and most of those state courts refused to adopt a significant-factor interpretation of “because of.” State v. Plowman, 314 Or. 157, 838 P.2d 558, 561 (1992) (reading a “because of’ requirement in a hatecrime enhancement to “expressly and unambiguously require[ ] the state to prove a causal connection between the infliction of injury and the assailants’ perception of the group to which the victim belongs”); State v. Hennings, 776 N.W.2d 112, 2009 WL 2960616 at *6-8 (Iowa Ct.App.2009) (table decision) (holding that the State must show that the discriminatory motive was the cause in fact of the offense to establish a hate crime under Iowa law), aff'd in relevant part, 791 N.W.2d 828 (Iowa 2010).

The government separately points to a case involving Section 1 of the Hate Crimes Prevention Act and argues that it shows “because of’ means “substantial motivating factor” in Section 2 of the statute. See United States v. Maybee, 687 F.3d 1026, 1032 (8th Cir.2012). But that court, too, never addressed the possibility that the phrase “because of’ in the statute might require but-for causation, id. at 1031-32, offering no reason to listen to it here. All roads lead to the same conclusion: For an assault to be a federal hate crime, the victim’s protected characteristic must be a but-for cause behind the defendant’s decision to act.

C.

Even if the district court incorrectly instructed the jury on this score, the government adds, any error was harmless. But motive played a starring role at trial, and the defendants presented evidence of other, non-religious motives for the assaults. The error was not harmless.

To be harmless, it must “appearf ] beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.” Neder v. United States, 527 U.S. 1, 15, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999) (internal quotation marks omitted). On the one hand: If “a defendant did not, and apparently could not, bring forth facts contesting the omitted element,” that would establish the harmlessness of the error. Id. at 19, 119 S.Ct. 1827. On the other hand: If the court “cannot conclude beyond a reasonable doubt that the jury verdict would have been the same absent the error—for example, where the defendant contested the omitted element and raised evidence sufficient to support a contrary finding—[the court] should not find the error harmless.” Id. This case is more like the second hand. Motive was the key issue the defendants presented to the jury, and they presented enough evidence to support a finding in their favor on this score. Consider the evidence relating to each of the four assaults (out of five) that led to convictions.

1.

The defendants perpetrated the first hair-shearing and beard-cutting attack against Martin and Barbara Miller in September 2011. According to the defendants, this attack was “not about religion” but “about bad parenting.” R. 542 at 106.

Evidence backed up their position. The Millers were attacked by their children *595and their children’s spouses, and they had a strained relationship with their children before the assault. Several witnesses testified that the Miller children resented their parents because they ‘‘belittle[ed] [them and] ... [a]lways put[ ] them down, never [making them feel] good enough.” R. 529 at 104 (Nancy Burkholder); see also R. 529 at 225 (John Aske); R. 537 at 238 (Daniel Shrock); R. 538 at 247-53 (Melvin Shrock, Jr.); R. 539 at 84-85 (Johnny Mast); R. 540 at 17-18 (Barbara Yoder). In the years before the assault, these feelings of resentment and disrespect grew. In late 2007 or early 2008, the Millers insulted their son Raymond and their daughter-in-law Kathryn by refusing to cook for or even attend their wedding. R. 528 at 247-49 (Barbara Miller). They then rebuffed their son-in-law Freeman Burkholder’s request for help refinancing jointly owned property in 2010, even though they knew Freeman might lose his home if they refused. R. 528 at 263-64; R. 529 at 136-37. When Barbara Miller’s daughters-in-law came to see her in January 2011, she “lock[ed] the door and t[old] them that they were not welcome in [her] house.” R. 528 at 277-78. A few months later in June 2011, the Millers snubbed their disfavored children by not inviting them to their youngest son’s wedding. R. 529 at 41 (Barbara Miller). And a month or so before the attacks, Martin Miller publicly chastised his daughter and criticized one or more of his sons when he ran into them at an Amish horse auction. R. 529 at 39-40 (Barbara Miller); id. at 110-12, 149 (Nancy Burkholder); R. 537 at 231 (Daniel Shrock); R. 538 at 247 (Melvin Shrock, Jr.); R. 539 at 85 (Johnny Mast).

On top of this, only the Miller’s children and their spouses perpetrated the beard cutting, and the children did not attack the Millers until four years after the Millers left Bergholz for another Amish sect, the key faith-based dispute in the community. R. 528 at 210 (Barbara Miller). Several intervening personal disagreements, described in part above, arose after the Millers left Bergholz, providing alternative motives for the assaults. The children, indeed, yelled accusations about “being rotten parents” during the attack. Id. at 243-44 (Barbara Miller). And many witnesses testified that bad-parenting disputes inspired the assault. See, e.g., R. 529 at 103 (Nancy Burkholder); R. 537 at 238 (Daniel Shrock); R. 538 at 247-48 (Melvin Shrock, Jr.).

Did some of this strife stem from religious discord? No doubt. But untangling the role of religion, family, personality and other issues in the assaults was the point of the trial. Just because the Millers and their children disagreed about the tenets of their religion and how to practice it does not make the Millers’ religious beliefs a but-for cause of their children’s attack on them. A failure of a child to meet the expectations of a parent (or the reverse)— whether those expectations stem from faith, tradition, vanity, familial hierarchy or something else—is hardly an unusual source of discord between parents and children. A jury could reasonably have found that the Millers’ constant criticism and rejection of their children, not the children’s disagreement with their parents’ faith, spawned the attacks.

How could this be, the government responds, given the religious nature of the attacks: cutting the hair or beards of the victims? But assaults involving religious symbolism do not invariably stem from religious motives. Imagine that an adult male assaults a child. The father of the child confronts the man. After yelling at the man over what he has done, the father violently grabs a cross pendant hanging from the man’s neck, yelling ‘You hypocrite,” and injures the man’s neck in the process. The religious nature of the cross *596is important, to be sure, but it does not necessarily show, as the dissent claims here, that the father assaulted the man “because of’ of his faith. Or, given that this is the Matthew Shepard Act, imagine that a child tells his parents he is gay. As a result of their faith, the parents ask the child to undergo reparative therapy. The child resists, the parents dig in, all three fight verbally about everything from faith to family obligations. At some point, the child snaps. He assaults the parents and does so in a faith-offensive way—by physically forcing them to eat non-kosher food, by tattooing 666 on their arms or by taking some other action that deeply offends their faith. No doubt faith entered the mix from both sides of the assault, but there is doubt about whether the parents’ faith broke the camel’s back in terms of why the child committed the assault. That the means of assault involved religious symbolism confirms only that he knew how best to hurt his parents. It does not seal the deal that his parents’ faith, as opposed to their lack of support for him, was a but-for motive of the assault.

Keep in mind, moreover, what happened in this case. Despite the presence of beard cutting in all five indicted assaults under the Hate Crimes Prevention Act, the jury convicted the defendants of only four hate-crime violations. We will never know why the jury did what it did. But the mixed verdict casts some doubt on the idea that a faith-inspired manner of assault necessarily equals a faith-inspired motive for assault.

Nor is it fair to say that, because faith permeates most, if not all, aspects of life in the Amish community, it necessarily permeates the motives for the assaults in this case, no matter how mundane the personal, power, or getting-one’s-way disputes that formed the backdrop to these assaults. Even people of the most theocratic faith may do things—including committing crimes—for non-faith-based reasons. And even ostensible faith leaders, whether Samuel Mullet or Henry VIII, may do things, including committing crimes or even creating a new religion, for irreligious reasons.

2.

Another attack, this one on Melvin and Anna Shrock, was of a piece with the first one. According to the defendants, it too stemmed from family discord. Emanuel resented his parents’ constant scorn and criticism, and on the night of the November attack he thought “his dad was lying to him,” he was “upset that [his parents] called the police” before coming to his house, and he disliked the way “he perceived his father reacting to him” when he pressed his father to accept his decision to stay in Bergholz. R. 542 at 164-66.

Like the attack on the Millers, the assault on the Shrocks involved only immediate family. The perpetrators were the Shrocks’ son Emanuel, his wife Linda, and two of the Shrocks’ grandchildren. R. 539 at 208-09. The defendants attested to their frustration with Melvin and Anna’s constant criticism of them and claimed that this is why they attacked them. Daniel Shrock, one of Emanuel’s sons and a participant in the assault, testified that Emanuel wanted to cut Melvin’s beard because he kept “writing us letters and trying to get us to move out [of Bergholz]” and because he implied that Bergholz was a cult by “sen[ding the family] a cult book at one time.” R. 537 at 208-09. Daniel added that Emanuel would not have cut Melvin’s beard if he had admitted he was “wrong about the things he ha[d] said about Bergholz and the way [the Bergholz community] practice^] [its] faith.” Id. at 212, 279. Melvin Shrock, Jr., the other son who participated in the assault, echoed *597this theme. According to him, Emanuel planned to attack his father “[b]ecause [Melvin] kept on sending him letters trying to get him to move out of the community, to change his faith.” R. 538 at 242. Near the time of the attack, Linda expressed the same concerns to the women in Bergholz. One government witness admitted that “it bothered [Linda] that [Melvin and Anna] wanted Emanuel and Linda to move out [of Bergholz].” R. 529 at 99. From the defendants’ perspective, the attack was about the Shrocks’ refusal to accept them and repeated efforts to force them to change.

Witnesses also testified that Melvin’s decision to bring law enforcement to Emanuel’s home may have been the final straw that drove Emanuel to act. Samuel Mullet told the police that Emanuel attacked Melvin “based upon the fact that the sheriff was there ... [and] out of anger[ ] that the sheriff was there.” R. 540 at 220. Immediately after the attack, Linda suggested much the same motive when recounting events to the women. As she explained it, “Melvin had taught Emanuel not to go to the law. And he does exactly the opposite and comes with the sheriff. So Emanuel decided to take his hair and beard.” R. 529 at 98.

Emanuel’s actions give these statements credence. Emanuel wrote his parents and told them he felt like “the black sheep in the family” and was “hav[ing] a hard time trusting” them, and he asked his parents to drop their efforts to convince him to leave Bergholz. R. 539 at 194, 196-97. When his parents came to visit and sent the sheriff into his house, he “got upset” and “mad.” R. 539 at 222. Despite this, the family had a “friendly” meal together until the conversation turned into an argument between Emanuel and his father about Bergholz. R. 539 at 174, 206-07. “Emanuel ... ask[ed] his father a couple of questions, and [he felt] his father was lying to him.” R. 539 at 44. When Emanuel tried to get Melvin to admit that he was wrong about Bergholz, Melvin “made fun of [Bergholz] in a way” and told Emanuel that the community would fall apart as he had predicted if Emanuel “just wait[ed] a little while.” R. 537 at 270-71. ' Only then did Emanuel “g[e]t the scissors” and “cut [his father’s] hair and beard off’ while telling him “maybe this will help you.” R. 539 at 44-45; see also R. 537 at 212, 271. As with the attack on the Millers, considerable evidence would permit a reasonable jury to find that intra-family conflict motivated Melvin and Anna’s assault.

3.

The defendants likewise framed the attack on Raymond, Andy and Levi Hersh-berger as stemming from interpersonal conflicts. Johnny Mullet, Samuel’s son, led the attack on Raymond Hershberger and his family purportedly to avenge harms done to his sister Wilma. Johnny targeted Raymond Hershberger in particular because he reversed the Troyers’ excommunications, eliminating “the possibility that [Johnny’s - sister] Wilma and her husband [Aden] would reunite” and resulting in ‘Wilma’s children [being] taken from her in a custody dispute.” R. 542 at 178-79.

The evidence at trial made the custody-dispute motive for this assault plenty plausible. Several witnesses testified that the loss of Wilma’s children deeply upset the community—enough so that some members may have resorted to retaliatory violence. See, e.g., R. 529 at 34 (Barbara Miller); R. 539 at 115-16 (Johnny Mast); R. 540 at 58-59 (Barbara Yoder). As these witnesses described it, the “community was in strife” over the loss of Wilma’s two little girls, R. 529 at 34, and the community members “were upset because they *598believed that it was wrong that the children were taken away from their mother,” R. 540 at 58.

The defendants’ confessions tied the community’s grief over the loss of Wilma’s children to their decision to attack Raymond and his family. When Johnny confessed to the attack against Raymond, he briefly talked about Raymond’s involvement in reversing the Troyers’ excommunications and how it affected Wilma’s two daughters. R. 588 at 161. So too did other confessed participants. Danny Mullet expressed anger that the Hershbergers “sided with the guys that kidnapped [Wilma’s] little girls,” id. at 173, and Eli Miller mentioned the “[t]wo little girls” and “the custody issue” in his police interrogation, id. at 204. The officer who arrested the other defendants involved in the Hershber-ger assault (Levi Miller, Lester Mullet, and Lester Miller) admitted that “[a] few [of those defendants] had mentioned a custody case” when discussing the assaults too, but he did “not know which ones for sure.” R. 538 at 105; see also id. at 75-76. Most, if not all, of the defendants referenced the custody dispute as a motive for the Hershberger assaults. As for those who did not, a jury could reasonably infer that they shared the same motive as the other confessed participants given that they agreed to carry out the attack as a group. All of this evidence considered, a reasonable jury could find that the government did not show beyond a reasonable doubt that a but-for cause of the attack on the Hershbergers was their faith, as opposed to other considerations, including the loss of Wilma’s children.

4.

As with the Hershbergers, Johnny Mullet argued that he led the assault on Myron Miller to avenge harm done to another of his siblings, Bill Mullet. He targeted Myron Miller for the assault because he “help[ed]” Bill “move out” of Bergholz and then kept him from “communicating] with [his] family” and forced him to return a horse and buggy that their father Samuel had sent as gifts. R. 542 at 179.

Ample evidence supported this theory. Myron Miller admitted that the events Johnny complained about had transpired. According to him, he “helped [Bill] move out” of Bergholz, R. 537 at 11, he advised Bill to “cut ties” with his family in Ber-gholz, id. at 19, and just one week before the attack, he “warned [Bill] if he d[id]n’t reject [the horse and buggy], he [could] not take communion with the church,” id. at 22. Witnesses tied the assaults to these events. Melvin Shrock, Jr., said that the men attacked Myron Miller “[b]ecause of the way he was treating my Uncle Bill Mullet.” R. 538 at 237. Johnny Mast echoed the point: The men targeted Myron Miller because of “[t]he way he was treating my Uncle Bill.” R. 539 at 37. Myron Miller himself believed the attack occurred because “in the Bergholz community members’ eyes, [he was] treating Bill[] bad[ly] and the precipitating event was the week before ... [he] ordered Bill[] to turn the horse and buggy back over to the Bergholz Amish community.” R. 537 at 75.

Myron Miller, to be sure, saw the horse- and-buggy dispute as a “religious issue,” but only because he did not make the decision to order their return alone. As bishop, he sought the advice of the community, and everyone in his church district had agreed with him that Bill should not keep the horse and buggy. Myron Miller’s classification of the dispute, though, sheds little light, least of all controlling light, on how the defendants saw the matter or how a jury might see it. A jury could reasonably side with the defendants, classifying the horse-and-buggy rebuff as a personal *599affront of the most mundane rather than religious order.

5.

What of Samuel Mullet, the alleged ringleader of the conspiracies and the self-anointed leader of the Bergholz community? The government as an initial matter did not present a separate harmless-error argument as to him. Samuel at any rate had close personal relationships with the victims and the other assailants, and the same theories of defense applicable to the other defendants applied to Samuel Mullet as well. Martin and Barbara Miller were his brother-in-law and sister, and the Miller children were his nieces and nephews. Melvin and Anna Shrock harassed his son-in-law Emanuel and daughter Linda, encouraging them to leave Bergholz with his grandchildren. Raymond Hershberger voted to overturn excommunications that injured Samuel’s daughter Wilma and caused his granddaughters to be taken to a different State. And Myron Miller hurt Samuel’s son Bill, forcing Bill to return gifts that Samuel had sent him by threatening excommunication if he kept them. Samuel had just as much of a personal stake in the assaults as the other convicted defendants.

Samuel’s criminal liability, moreover, depended on a jury finding of conspiracy, as Samuel was not present at and did not participate directly in any of the hair- and beard-cutting assaults. The essence of conspiracy is an agreement to commit an unlawful act. Iannelli v. United States, 420 U.S. 770, 777, 95 S.Ct. 1284, 43 L.Ed.2d 616 (1975). A jury might reasonably conclude that the unlawful agreement in this case included a consensus about which victims to target and why, and it thus could readily find that all of the defendants, Samuel included, shared similar non-faith-based motives—or more to the point that there was reasonable doubt as to whether they did. Whether Samuel shared the motives of his co-conspirators or had his own reasons for encouraging the assaults are precisely the kinds of questions that we normally ask juries to sort out.

Even if Samuel did not share his co-conspirators’ motives, his individual motives did not necessarily turn on the victims’ religious beliefs. Some evidence suggested that Samuel agreed to the assaults because he felt that the victims had disrespected him, and he wanted revenge for that disrespect. He told the authorities that he thought some of the attacks occurred because the men in the community were “mad” about the way the victims “had treated Sam[uel] Mullet.” R. 540 at 123. And he told the media that “the goal of the haircutting was to send a message to the Amish in Holmes County that they should be ashamed of themselves for the way they were treating Mullet and his community.” Id. at 143. Buttressing the point, Samuel had a reputation as a hothead and was not above resorting to violence when he felt disrespected. See R. 528 at 198. If people disagreed with him or things did not go his way, he could become “violen[t],” “[a]ngry,” and “[h]ateful,” and he took the position that “[i]f [a person] does [something bad] to me, I’ll do that and worse to him.” Id. On this record, a properly instructed jury must resolve whether religion or something else motivated Samuel to participate in the attacks.

When all is said and done, considerable evidence supported the defendants’ theory that interpersonal and intra-family disagreements, not the victims’ religious beliefs, sparked the attacks. And all of this evidence could have given a reasonable juror grounds to doubt that religion was a but-for cause of the assaults. As the dis*600trict court judge remarked in considering whether to give a but-for jury instruction, it would be difficult—maybe “impossible”—“to determine beyond a reasonable doubt [that] the defendant^] would not have [attacked] but for the victim’s religion or perceived religion.” R. 314 at 25-26. The defendants’ lawyers, we suspect, could not put it any better. Any such difficulty confirms that a properly instructed jury, not this court, must decide the but-for causality question for itself.

6.

Even though the dissent agrees that the trial judge incorrectly instructed the jury on the causation element of the crime, it claims that any error was harmless. We disagree, most pointedly because the causation question turns on.the state of mind of the defendants—the subjective motivations for the attacks. Assessing whether a prohibited motive was a but-for cause of a defendant’s actions presents a much more difficult task than assessing whether a defendant’s actions were a but-for cause of a subsequent event. As Justice Breyer aptly explained, “In a case where we characterize [a person’s] actions as having been taken out of multiple motives, ... to apply ‘but-for’ causation is to engage in a hypothetical inquiry about what would have happened if the [person’s] thoughts and other circumstances had been different. The answer to this hypothetical inquiry will often be far from obvious.” Gross, 557 U.S. at 191, 129 S.Ct. 2343 (Breyer, J., dissenting). No one in the Gross majority disagreed with Justice Breyer’s point—and Gross was a civil case, not a criminal one where the beyond-a-reasonable-doubt standard applies.

An improper instruction on motive and but-for causation poses a thorny issue that frequently will not be harmless. Experience bears this out. If the motive in this case was as obvious as the dissent suggests, why didn’t the government obtain a conviction on one of the five charged hate crimes even under the less-than-but-for-cause standard that the district court gave and even when the alleged religious mode of assault (cutting the victim’s beard) was the same? What of the two other prosecutions under Section 2 of the Hate Crimes Act? Even though considerable evidence showed that the defendants acted based on improper motives in those cases, no convictions occurred. United States v. Jenkins (Jenkins I), No. 12-15-GFVT, 2013 WL 3158210, at *2 (E.D.Ky. June 20, 2013) (acquitted); United States v. Mason, No. 3:13-CR00298, App. R. 97-1 at 1 (jury unable to reach a verdict); see also United States v. Jenkins (Jenkins II), No. 12-15-GFVT, 2013 WL 3338650, *6-7 (E.D.Ky. July 2, 2013).

That’s not all. Consider also that a mere change in the burden of proof on but-for causation altered the outcome in Gross. When the employer had the burden of proof, Gross prevailed. Gross, 557 U.S. at 171, 129 S.Ct. 2343. When the burden of proof shifted to Gross, the employer carried the day. Gross v. FBL Fin. Grp., Inc., 489 Fed.Appx. 971, 972-73 (8th Cir.2012). Consider too that our sister circuit did not find a similar error harmless in Nassar. Nassar v. Univ. of Tex. Sw. Med. Ctr., 537 Fed.Appx. 525, 525 (5th Cir.2013) (remanding for a new trial on but-for causation). All told, in the aftermath of Gross, Nassar and Burrage, only one court has found a but-for instructional error harmless and then only because the jury completed a special verdict form in which it expressly indicated “it relied only on a ‘but for’ theory of causation when finding liability” and “did not reach or rely on the alternative ‘mixed motive’ theory presented to it.” See Barrett v. Salt Lake Cnty., 754 F.3d 864, 868 (10th Cir.2014). That option was not given to the jury here. All *601of this demonstrates that a but-for instructional error typically will be harmful, especially in a criminal case like this one.

That this case involves the defendants’ subjective motivations rather than their actions differentiates it from the key case the dissent relies upon: Harrington v. California, 395 U.S. 250, 89 S.Ct. 1726, 23 L.Ed.2d 284 (1969). There, the harmlessness question focused on whether the defendant was present at the scene of the crime. Id. at 252-53, 89 S.Ct. 1726. The Supreme Court found a Confrontation Clause violation harmless because the erroneously admitted testimony “was of course cumulative” of the other evidence that “placed [the defendant] in the store with a gun at the time of the murder.” Id. at 253-54, 89 S.Ct. 1726. The defendant himself, moreover, “agreed he was” at the scene of the crime. Id. at 253, 89 S.Ct. 1726. Harrington thus tells us little about the harmlessness analysis in this case, where the contested element involves the defendants’ subjective motivations and where the defendants focused much of their defense on this point.

In an effort to support a contrary outcome, the dissent points to evidence that supports a religious-motive theory of the assaults. No doubt, such evidence exists. Had there not been probable cause to support the government’s theory of the case, there never would have been an indictment. But, as we have explained at length, this is not the only way to read the evidence; it is merely the best way to read the evidence—for the prosecution. That is not how harmless error works. One must consider the evidence in support of the government and the evidence in support of the defendants. That a criminal act in the abstract may have more than one but-for cause, moreover, does not take away from the defendants the right to have a jury find whether one motive, two motives, or several motives were beyond a reasonable doubt but-for causes of the beard-and-hair-cutting attacks and whether the victims’ religious beliefs were one such but for cause—a cause that “broke the camel’s back,” as Burrage puts the point. Nor does it make sense to pluck one defendant out of the mix (Samuel Mullett)—in a conspiracy case no less—and say that he does not get a trial under a proper instruction of causation while the other defendants do. There is not a lot of precedent for that approach to harmless error. Because a reasonable juror could reach more than one conclusion regarding the defendants’ motives, the only issue we must consider, this erroneous instruction on the key debate at trial—the state of mind of the defendants—was not harmless.

IV.

Although the absence of a but-for jury instruction requires a new trial, the Double Jeopardy Clause prevents the government from taking the case to a second jury unless it presented enough evidence before the first to get a conviction. See Patterson v. Haskins, 470 F.3d 645, 651 (6th Cir.2006) (“[T]he longstanding prudential rule in this circuit [is] that an appellate court faced with arguments both that the evidence was insufficient and that the trial was infected with other constitutional errors needs to address the sufficiency-of-the-evidenee issue, even if the court orders a remand on the basis of trial error.”); Hoffler v. Bezio, 726 F.3d 144, 162 (2d Cir.2013) (noting that this practice is uniform across the circuits). In this setting, we view the evidence in the prosecution’s favor and, only then, ask whether any rational trier of fact could have found the defendant guilty beyond a reasonable doubt. United States v. Djoumessi, 538 F.3d 547, 551 (6th Cir.2008). Our narrow inquiry here is limited to the sufficiency *602challenges that the defendants have raised in this appeal. See Patterson, 470 F.3d at 651. As concerns those challenges, the government met its burden.

A conspiracy exists where the defendants knowingly and intentionally agreed to violate the law and took a step towards effectuating that unlawful agreement. United States v. Blackwell, 459 F.3d 739, 760 (6th Cir.2006). A reasonable jury could find just such an agreement here given the evidence identified above, including: The crimes followed the same pattern and method of attack; they were relatively close in time to each other; the defendants met and discussed the attacks before and after they occurred; and many of the defendants participated directly in more than one assault. These facts, among many others, provide enough evidence of an agreement to take concerted action against the various victims to support a conspiracy conviction.

A few of the defendants also challenge their convictions under various requirements of the Hate Crimes Prevention Act, saying this or that element was not established by the government. Some defendants claim that the government failed to show that certain victims suffered “bodily injury.” 18 U.S.C. § 249(a)(2)(A). But by any reasonable measure, forced hair-shearing and beard-cutting amounts to a “bodily injury,” a phrase that in other criminal contexts extends to “disfigurement” or “any other injury to the body, no matter how temporary.” Id. § 1365(h)(4). Even if that were not the case, the government offered proof that at least one victim in each of the three challenged attacks suffered other physical harm: Barbara Miller had bruises on her arms, R. 528 at 224; Martin Miller had cuts and razor burns on his face, R. 529 at 52; Raymond Hershberger had cuts on his head, R. 528 at 64; and Melvin Shrock had a small gouge on his cheek, R. 556-3, R. 556-5.

Another defendant (Linda Shrock) says that the government presented insufficient evidence to prove she aided and abetted her husband’s commission of the crime. Yet a jury could reasonably find that Linda aided her husband in attacking the Shrocks. Linda, after all, grabbed Anna Shrock and covered her mouth when Anna tried to run outside for help during the attack against her husband Melvin. R. 539 at 209-10. That fact alone suffices to show that Linda intended to and did act in a way that made it possible for Emanuel to cut Melvin Shrock’s beard. See United States v. Lowery, 60 F.3d 1199, 1202 (6th Cir.1995). In sum, none of the defendants’ sufficiency challenges precludes a retrial under the Double Jeopardy Clause.

V.

The defendants raise other challenges to their convictions, including whether the district court properly admitted evidence of Samuel’s sexual misconduct under Evidence Rules 404(b) and 403, whether the crimes can be characterized as kidnapping for sentencing purposes, and whether the federal hate-crime statute exceeds Congress’s Commerce Clause powers as applied to the facts of this case. While these other arguments present not-inconsequential issues, we need not address them given our conclusion that the erroneous jury instructions require a new trial.

VI.

For these reasons, we reverse the defendants’ various hate-crime convictions and remand the case for further proceedings consistent with this opinion.