dissenting.
On October 8, 2001, Aaron Reynolds (“Reynolds”) lost control of the car he was driving in an illegal drag race. The vehicle careened into a crowd of spectators, striking and killing Denise Jones (“Jones”). Several Lincoln Park police officers arrived at the scene before the race began and did not stop it, but this is emphatically not what this case is about. What really matters is Jones’s evidence that the race was about to be called off for fear of police intervention but ultimately went forward after the officers actively encouraged the race to proceed, reassuring the drag racers and the crowd that nobody would be arrested for conducting the race. Because I believe that this evidence, if credited by a jury, establishes both an affirmative act and a special danger under the state-ereated-danger theory, I would reverse and remand the case to allow the district court to address the issue of state culpability in the first instance. Thus, I respectfully dissent.
I.
I present the following facts drawn from the record in order to give a fuller picture of the crucial events that transpired on October 8, 2001. Naturally, the defendants dispute much of this account, but given the procedural posture, we must view the evidence in the light most favorable to Jones.
There is evidence in the record to suggest that the race would have been can-celled because of the officers’ arrival at the scene. According to an eyewitness, “[everybody was fittin’ to leave because they [saw] the police.” Joint Appendix (“J.A.”) at 281 (Young Dep. at 22). Indeed, some vehicles drove away soon after the officers’ arrival. J.A. at 1219 (Muncey Statement at 6). Moreover, Reynolds, the driver of the car that struck Jones, “was afraid to *700race when the Lincoln Park Police first drove up” for fear of being arrested, so he “was in the process of putting [his] car back onto the trailer, to call off the race.” J.A. at 1041 (Reynolds Aff. at 2); see also J.A. at 281 (Young Dep. at 22) (“[T]hey [were] about to pack up and leave.”).
Of course, the race was not called off. At least one officer — apparently Officer Mohamed Nasser (“Officer Nasser”)— briefly spoke to Reynolds and the other drag racer, Mustapha Atat (“Atat”). J.A. at 281-82 (Young Dep. at 22-23), 1041 (Reynolds Aff. at 2), 1046 (Ricks Aff. at 2), 1053 (Moore Aff. at 1), 1239 (Request for Warrant/Investigator’s Report at 1). Officer Nasser told Reynolds that the officers “want[ed] to see the race.” J.A. at 1041 (Reynolds Aff. at 2). On the way back to his vehicle, Officer Nasser was seen placing a wager on the drag race. J.A. at 1053 (Moore Aff. at 1). Officer Nasser returned to his cruiser, “got on his loud speaker[,] and said, ‘We’re not here to arrest anyone, go ahead.” ’ J.A. at 282 (Young Dep. at 23); see also J.A. at 1041 (Reynolds Aff. at 2) (“[W]e were told by the Lincoln Park Police to go ahead and race.”), 1046 (Ricks Aff. at 2), 1049 (W. Turner Aff. at 2), 1054 (Moore Aff. at 2), 1109 (T. Turner Aff. at 3). Officer Douglas Muncey played rap music over a loudspeaker before the race began. J.A. at 284-85 (Young Dep. at 25-26), 662 (Muncey Dep. at 69), 1042 (Reynolds Aff. at 3), 1047 (Ricks Aff. at 3), 1050 (Turner Aff. at 3), 1053 (Moore Aff. at 1), 1109 (T. Turner Aff. at 3), 1234 (Nasser Statement at 22). In response to the music, the crowd “got hyper” and “everybody start[ed] hollering and ... dancing.” J.A. at 287-88 (Young Dep. at 28-29); see also J.A. at 1050 (Turner Aff. at 3). According to an eyewitness, “[i]t’s like the police initiated the race to go on.” J.A. at 288 (Young Dep. at 29). Indeed, Reynolds attested, “I would not have raced if the police officers had not told me that we could go ahead and race.” J.A. at 1042 (Reynolds Aff. at 3).
II.
In Kallstrom v. City of Columbus, 136 F.3d 1055 (6th Cir.1998), we recognized the state-created-danger theory of due process liability and laid out three requirements: an affirmative act that creates or increases the risk, a special danger to the victim as distinguished from the public at large, and the requisite degree of state culpability. Id. at 1066-67. Jones has offered sufficient evidence to establish a genuine issue of material fact with respect to at least the first two prongs of the Kallstrom test. Because the district court did not rule on the third prong,1 I would reverse and remand the case to permit the court to address the issue in the first instance.
A. Affirmative Acts that Create or Increase the Risk
“Liability under the state-created-danger theory is predicated upon affirmative acts by the state which either create or increase the risk that an individual will be exposed to private acts of violence.” Id. at 1066. The majority’s conclusion that the officers committed no cognizable affirmative acts has two bases. First, the majority claims that the officers simply failed to stop or discourage the race from happening, such that all the state did was fail to act. See Cartwright v. City of Marine City, 336 F.3d 487, 493 (6th Cir.2003) (“[F]ailure to act is not an affirmative act under the state-created danger theory.”). Second, the majority opines that the offi*701cers did not increase the danger to Jones. According to the majority, even if the officers had not arrived at the scene, the race would have gone ahead as scheduled, placing Jones in the very danger she faced during the situation as it actually transpired.
Both strands of the majority’s reasoning depend on partitioning the officers’ conduct into two phases — (1) arrival at the scene and (2) everything else — and then considering only the first. I assume for present purposes that under our precedents, the police would not have committed an affirmative act if they had simply arrived at the race and done nothing, as that truly would have been a failure to act that did not create or increase the risk of the drag race. But those are not the facts of this case, because that is not all the officers did. Jones has offered evidence showing that the officers actually were quite busy upon their arrival: they reassured the racers, Reynolds and Atat, that the race could proceed; announced over a loudspeaker that they would not arrest anyone, so the race could “go ahead”; and played rap music over a loudspeaker, which had the effect (and perhaps the intent) of stirring up the crowd. The majority inexplicably refuses to acknowledge that Jones’s claim is based on all of this conduct rather than the mere fact that the officers did not stop the race.
The majority offers no persuasive reason for ignoring these actions. It tries to wish away the facts by repeatedly invoking the principle that a failure to act does not state a due process violation, but that gets the inquiry exactly backward: the only way to determine whether the officers merely failed to act is by considering all the facts. The decisions cited by the majority certainly do not compel its myopic view of the facts, as none involved state officials arriving at a scene and then actually encouraging private actors to engage in dangerous acts or telling them that the state would look the other way. Furthermore, it is difficult to square turning a blind eye to post-arrival conduct with the principle that the Due Process Clause “is phrased as a limitation on the State’s power to act.” DeShaney v. Winnebago County Dep’t of Soc. Servs., 489 U.S. 189, 195, 109 S.Ct. 998, 103 L.Ed.2d 249 (1989). This principle instructs us to look at the state’s actions, not ignore them. It is why encouraging and condoning private actors’ dangerous conduct is relevant to the due process inquiry and may form the basis of liability.2 E.g., Pena v. DePrisco, 432 F.3d 98, 111 (2d Cir.2005) (“[W]hen ... state officials communicate to a private person that he or she will not be arrested, punished, or otherwise interfered with while engaging in misconduct that is likely to endanger the life, liberty or property of others, those officials can be held liable under section 1983 for injury caused by the misconduct ....”).
I cannot help but wonder whether there is any post-arrival conduct that the majority wouldn’t ignore. Could an officer have waved a green flag to signal the start of the race? Could a second officer have served as a play-by-play announcer over the loudspeaker? Could yet another officer have gotten behind the wheel and actually driven in the race against Reynolds? In response to these questions, the majority concedes (albeit under the wrong rubric3) that officers would run afoul of *702the Due Process Clause by engaging in these three hypothetical acts. This means that the majority detects a constitutionally significant difference between an officer waving a flag to signal the start of a race and an officer announcing over a loudspeaker that the racers may “go ahead.” The line between these two acts is too fine for me to see. What rationale justifies treating these two acts differently? Is it the physical distance between the officers and the racers (i.e., although here the officers were just across the street from the race, they would have had to be even closer to wave the starting flag)?4 Is it the prop used by the officer (i.e., the racers might have been encouraged by the officers giving the “go ahead” over a loudspeaker, but waving the flag would have shown that the officers really meant it)? These reasons are obviously too flimsy to bear the weight of a constitutional distinction, yet the majority offers none of its own. All we are left with, then, is the fact that the relevant conduct occurred after the officers arrived at the scene. Perhaps the majority would remove its blinders for post-arrival conduct surpassing some higher level of egregiousness, but its reasoning certainly suggests no limits to its brand of willful ignorance.
It is clear, then, that the proper point of comparison is not the risk that Jones would have faced if the officers had never arrived at all, but the risk that would have existed had the officers arrived but not reassured the racers, announced that the race could go ahead, and played rap music (i.e., if they had not committed the acts that the majority ignores). After the officers arrived but before they engaged in this additional conduct, people in the crowd were preparing to leave, some vehicles actually did leave, and one of the racers — Reynolds, the driver who struck Jones — was in the process of putting his car on a trailer and calling off the race. This evidence suggests that the race would not have gone forward if the officers had not committed these additional acts upon their arrival at the scene. By reassuring the racers and the people in the crowd that they could “go ahead” because they would not be arrested and by riling up the crowd with rap music played over a loudspeaker, the officers instead actually encouraged the race to proceed. The officers’ actions, which revived a drag race that was about to be abandoned, thus created or increased the risk of injury to Jones.
The majority attempts to foreclose this analysis by likening the case to Bukowski v. City of Akron, 326 F.3d 702 (6th Cir.2003), and DeShaney. In those cases, the state took temporary custody of the eventual victim based on suspicions of abusive circumstances; however, a lack of evidence left the state without legal authority to maintain custody, requiring the victims to be returned. Bukowski, 326 F.3d at 705-06; DeShaney, 489 U.S. at 192, 109 S.Ct. 998. Upon their respective returns, Lisa Bukowski was raped and Joshua DeSha-ney was beaten into a coma. Bukowski, 326 F.3d at 706; DeShaney, 489 U.S. at 193, 109 S.Ct. 998. In each case, the court held that the state did not create the danger because it “was merely returning a *703person to a situation with a preexisting danger.” Bukowski, 326 F.3d at 709 (relying on DeShaney, 489 U.S. at 201, 109 S.Ct. 998).
There are critical differences between the instant case and Bukowski and DeShaney. Rather than merely suspecting a danger, here the police arrived and actually saw the private actors preparing for the drag race, i.e., the very danger that ultimately harmed Jones. And rather than merely returning the victim to the status quo ante due to a lack of evidence of the risk, here the police encouraged the private actors to engage in the dangerous conduct even though they otherwise would have abandoned it. For Bukowski to be truly analogous to the case at bar, the officers would have had to know the rapist was about to attack Bukowski and then tell the rapist to go ahead despite his willingness to stop. Similarly, for DeShaney to be genuinely analogous, the state social workers would have had to know DeSha-ney’s father was about to beat his son and then encourage him to proceed even as he was about to give up the attack. Surely Bukowski and DeShaney would have come out the other way in these hypothetical scenarios.
Of course (and as the majority agrees), it is unimaginable that state officials would act in the manner posed in these hypothetical variations on Bukowski and DeShaney. But that is precisely the point: Jones has presented evidence that the officers in the instant case did act unimaginably, by encouraging Atat and Reynolds to engage in a dangerous and illegal drag race that ultimately killed Jones (and by egging on the crowd to boot).5 If this evidence were credited by a jury, it would demonstrate that the officers committed affirmative acts that created or increased the risk of harm to Jones.
The conclusion that the officers’ encouragement of the drag race constituted an affirmative act increasing the risk to Jones is consistent with other cases holding that where officers enable or embolden a private actor to drive dangerously, they commit an affirmative act for state-created-danger purposes. In Pena v. DePrisco, the officers drank excessively with the private actor (an off-duty officer) and rode in the car with him while he drove drunk. 432 F.3d at 110-11. The court held that a reasonable juror could find that the officers “implicitly but affirmatively condoned [the drunk driver’s] behavior and indicated to [the drunk driver] that he would not be disciplined for his conduct.” Id. at 111. In Reed v. Gardner, 986 F.2d 1122 (7th Cir.), cert. denied, 510 U.S. 947, 114 S.Ct. 389, 126 L.Ed.2d 337 (1993), the court held that “[p]olice officers who remove sober drivers and leave behind drunk passengers with keys may be said to create a danger or at least render others on the road more vulnerable.” Id. at 1125. In Nishiyama v. Dickson County, 814 F.2d 277 (6th Cir.1987) (en banc), a dangerous convicted felon in police custody was permitted to drive a patrol car unsupervised; he then pulled over a vehicle and beat its driver to death. Id. at 279. We explained that the officers who authorized the felon to drive the cruiser “facilitate[d] the crime by providing the criminal with the necessary means and the specific opportunity to commit his crime.”6 Id. at 281.
There are obvious parallels between these cases and the instant case. Officers *704knew or should have known of a potential driver who would be unquestionably dangerous if allowed on the road: a drunk driver in Pena and Reed; a dangerous felon who could masquerade as a cop in Nishiyama; and drag racers here. Rather than simply failing to act, the officers actively did something to enable or encourage the dangerous driver to proceed: in Pena, they sent implicit signals condoning drunk driving; in Reed, they removed the car’s driver and left behind the vehicle’s keys with a drunk passenger; in Nishiyama, they handed over the keys for a patrol car to an unsupervised, dangerous felon; and here, they reassured the drag racers that they could race, announced over a loudspeaker that nobody would be arrested and that the race could “go ahead”, and played rap music to energize the crowd.7
The majority’s decision rests on its conclusion that there was no affirmative act as a matter of law. Because I believe that Jones offered enough evidence to survive summary judgment with respect to the affirmative-act issue, I proceed to the remaining Kallstrom prongs.
B. Special Danger
In addition to an affirmative act, “plaintiffs alleging a constitutional tort under § 1983 [must] show [a] ‘special danger.’ ” Kallstrom, 136 F.3d at 1066. A special danger exists “where the state’s actions place the victim specifically at risk, as distinguished from a risk that affects the public at large.” Id. The district court appeared to conclude that Jones failed to meet this prong because she was not “in any more danger than any other citizen in the area that evening.”8 J.A. at 61 (Dist. Ct. Order at 19); see also id. at 65 (Dist. Ct. Order at 23) (noting that Jones was not “in more danger on that evening than any *705other person on the street that night”). Apparently, the court believed that there was no special danger because Jones was in the same danger as the other members of the crowd, even though the crowd as a group was at greater risk than the public at large.
The district court’s view of the special-danger requirement is inconsistent with our precedents. When we have rejected the existence of a special danger, the plaintiff faced the same danger as the general public. E.g., Schroder v. City of Fort Thomas, 412 F.3d 724, 729 (6th Cir.2005) (holding that traffic laws and enforcement practices posed “a general traffic risk to pedestrians and other automobiles”); Jones v. City of Carlisle, 3 F.3d 945, 949-50 (6th Cir.1993) (holding that an epileptic driver was “no more a danger to [the plaintiff] than to any other citizen on the City streets”), cert. denied, 510 U.S. 1177, 114 S.Ct. 1218, 127 L.Ed.2d 564 (1994); Janan v. Trammell, 785 F.2d 557, 560 (6th Cir.1986) (holding that the plaintiff “was simply a member of the public at large”); see also Mitchell v. Duval County Sch. Bd., 107 F.3d 837, 839 (11th Cir.1997) (“[The plaintiff] must show that the state affirmatively placed him in a position of danger which was distinguishable from that of the general public.”), cited with approval in Kallstrom, 136 F.3d at 1066.9 In contrast, when the danger threatened a group including the plaintiff rather than simply the public at large, we have found a special danger to exist. McQueen v. Beecher Community Schools, 433 F.3d 460, 468-69 (6th Cir.2006) (finding the existence of a special danger where the victim was a member of a group of children left in a classroom with an armed classmate); Waller v. Trippett, 49 Fed.Appx. 45, 50 (6th Cir.2002) (unpublished opinion) (finding that there was a special danger where the victim, who worked with inmates in a prison kitchen, “was a member of a limited and specifically definable group”); see also Schroder, 412 F.3d at 729 (rejecting the existence of a special danger where the state’s actions “did not create a ‘special danger’ to a discrete class of individuals (of which the [plaintiff] was a member)” (emphasis added)).
A drag race consists of driving at dangerous speeds for a limited time and distance. The confined scope of the race means that it posed a greater danger to the people lining the.street at that time than to the public at large, as most of the general population was not in the vicinity at the time of the race. In other words, the officers’ actions set an out-of-control car hurtling toward a specific crowd of people at a specific time. Thus, the spectator crowd constituted a “discrete class of individuals,” Schroder, 412 F.3d at 729, or “a limited and specifically definable group,” Waller, 49 Fed.Appx. at 50, that was at greater risk of injury than the general public. As a member of the spectator crowd, Jones faced a special danger.
The discreteness of the drag race distinguishes- it from the dangers in decisions like Jones v. City of Carlisle and Schroder v. City of Fort Thomas. In City of Carlisle, we rejected a claim premised on the state’s permitting a person known to suffer from uncontrollable epileptic seizures to maintain a driver’s license, because the driver “was no more a danger to [the *706plaintiff] than to any other citizen on the City streets.” 3 F.3d at 947, 950. The state’s issuance of a license did not let a danger loose at a specific time and place, because the epileptic driver could have suffered a seizure at any time and could have been anywhere when it happened. Thus, he posed no special threat to a “discrete class of individuals” or a “limited and specifically definable group.” In Schroder, we turned away a suit based on the state’s failure to lower or enforce the speed limit of a street where the decedent was struck by a speeding car, because the state “did not create a ‘special danger’ to a discrete class of individuals (of which the [decedent] was a member), as opposed to a general traffic risk to pedestrians and other automobiles.” 412 F.3d at 725, 729. The majority likens the instant case to Schroder because the asserted danger was spatially discrete, i.e., limited to a specific street, in both. Yet the state’s conduct in Schroder (failing to lower or enforce the speed limit) did not unleash a speeding car on that street at a specific time. Thus, once again, no discrete, limited, or specifically definable group was at risk there.
Finally, the majority claims that our cases actually require the government to be able to “specify whom it was putting at risk, nearly to the point of naming the possible victim or victims.” Majority Op. at 696. They say no such thing. We have never held that a group — short of the general public — is too large to be the target of a special danger. The folly of the majority’s revisionist rule demonstrates why we have not heretofore adopted it: the majority denies that Jones faced a special danger because she was part of a crowd, yet it would presumably conclude otherwise if Jones had been the only spectator. Am I to believe that the Due Process Clause, which is “a limitation on the State’s power to act,” DeShaney, 489 U.S. at 195, 109 S.Ct. 998, shields the state when it endangers a larger subset of its citizens? My copy of the Constitution does not demand such perverse results. Thus, Jones has presented sufficient evidence to survive summary judgment with respect to the second Kallstrom prong.
C. State Culpability
The final requirement of the state-created-danger test is that “[t]he state must have known or clearly should have known that its actions specifically endangered an individual.” Kallstrom, 136 F.3d at 1066. As I have explained above at footnote 8, this third prong was the stated basis for the district court’s grant of summary judgment to the defendants, but the substance of the district court’s analysis clearly demonstrates that it was actually analyzing the second prong. Because it did not rule on the third prong, I would remand to the district court to address the issue in the first instance.
III.
Whether a state official should intervene in private affairs is often a tough call, as potential land mines line the paths of both action and inaction. Indeed, this difficulty was a supporting rationale in DeShaney. The state was sued for leaving Joshua with his father, but “had [state social workers] moved too soon to take custody of the son away from the father, they would likely have been met with charges of improperly intruding into the parent-child relationship.” DeShaney, 489 U.S. at 203, 109 S.Ct. 998. We, too, have recognized this tension. “By failing to detain Bukowski, [state officials] face this lawsuit. If they had chosen to detain her, they may have faced another lawsuit based on charges of false imprisonment.” Bukowski, 326 F.3d at 712. That liability under the Due Process Clause is narrow reflects the urge to shield the state from liability when it faces such difficult choices. Here, however, the officers faced the following quandary when *707they arrived upon an illegal drag race and saw the racers and crowd preparing to forego the race before it started: (1) do nothing, leaving the race to be abandoned; or (2) reassure the racers that the race could proceed; announce over a loudspeaker that they would not arrest anyone, so the race could “go ahead”; and play rap music over a loudspeaker to rile up the crowd. The officers picked option two. Before the majority’s decision today, I would have thought it uncontroversial that this is not the type of state decisionmaking that DeShaney sought to protect. Due process means nothing if the officers’ conduct here is all the process that Denise Jones was due.
Because I believe that Jones’s evidence would satisfy the first two prongs of the state-created-danger test if credited by a jury, I would reverse the order granting summary judgment to the defendants and remand to the district court to address the third prong (and, if necessary, the issues of qualified immunity and municipal liability) in the first instance.
For the reasons set forth above, I respectfully dissent.
. Nor did the district court rale on the first prong. I address the issue here, however, because it is the basis of the majority's decision.
. This rule is not unique to direct state-created-danger claims. It is also, for example, a well-settled part of § 1983 supervisory-liability doctrine. See, e.g., Estate of Carter v. City of Detroit, 408 F.3d 305, 314 (6th Cir.2005) ("Supervisor liability attaches when a supervisor encourages or condones a constitutional violation.”).
. To be clear, I posit in each of these hypothetical questions that Reynolds's car ulti*702mately strikes Jones, so a private actor is the direct cause of the harm. Thus, the majority cannot brush the examples aside by simply claiming that the state-created-danger theory is not implicated at all.
. A fixation with physical proximity might explain why the majority repeatedly notes the fact that the race occurred in Detroit while the officers were in Lincoln Park. While this geographical circumstance might be relevant to whether state law permitted the officers to stop the race, it has no bearing on whether the Constitution permits them to actively encourage the race.
. One might argue that an impending rape or beating presents a more likely risk of harm than an impending drag race. This is beside the point, however, because the test is whether the affirmative act created or increased the risk of harm. Of course, the occurrence of a drag race presents a greater risk of harm than the absence of one.
. Quoting Stemler v. City of Florence, 126 F.3d 856 (6th Cir.1997), cert. denied, 523 U.S. 1118, 118 S.Ct. 1796, 140 L.Ed.2d 936 *704(1998), the majority notes that Nishiyama is “no longer an accurate statement of the law." Majority Op. at 694-95. Stemler’s repudiation of Nishiyama was only with respect to the requisite level of state culpability. Stemler, 126 F.3d at 869 (explaining that gross negligence — the standard endorsed by Nishi-yama — was no longer sufficient). Nishiyama's analysis of the officers' facilitation of private violence remains good law.
. Of course, the principle that officers commit affirmative acts when they encourage private misconduct that leads to violence is not limited to driving cases. See Dwares v. City of New York, 985 F.2d 94, 99 (2d Cir.1993) ("[T]he officers conspired with the ‘skinheads' to permit the latter to beat up flag burners with relative impunity, assuring the 'skinheads’ that unless they got totally out of control they would not be impeded or arrested. It requires no stretch to infer that such prior assurances would have increased the likelihood that the 'skinheads' would assault demonstrators. Thus, in the present case, the complaint asserted that the defendant officers indeed had made the demonstrators more vulnerable to assaults.”); Freeman v. Ferguson, 911 F.2d 52, 54-55 (8th Cir.1990) (remanding to permit amendment of the complaint to allege that the police chief, who was a close friend of the attacker, had prohibited officers from interfering with the attacker, because such an allegation "presents a claim that the violence the decedents were subjected to was not solely the result of private action, but that it was also the result of an affirmative act by a state actor to interfere with the protective services which would have otherwise been available in the community — with such interference increasing the vulnerability of decedents to the actions of [the attacker] and possibly ratifying or condoning such violent actions on his part. Without such affirmative actions on the part of the chief of police, the danger faced by the [victims] would have arguably been less.” (citation omitted)).
. The district court purported to resolve the summary-judgment motion on the basis of the third Kallstrom prong, i.e., state culpability. The substance of the analysis makes clear, however, that the court blended the second and third prongs and actually focused on the second prong, i.e., special danger.
. Jones v. Union County, 296 F.3d 417 (6th Cir.2002), which held that there was no special danger where the state did not timely serve an ex parte order of protection on the plaintiff's ex-husband, id. at 431, is not to the contrary. As we explained in McQueen v. Beecher Community Schools, 433 F.3d 460 (6th Cir.2006), the state's failure left the plaintiff in the same position as the general public, i.e., unprotected against the ex-husband’s potential attacks. Id. at 468 n. 8.