AMERICAN FEDERATION OF STATE, COUNTY AND MUNICIPAL EMPLOYEES COUNCIL 79, Richard Flamm, Plaintiffs-Appellees, v. Rick SCOTT, in his official capacity as Governor of the State of Florida, Defendant-Appellant, 717 F.3d 851

MARCUS, Circuit Judge: This appeal presents two closely related issues: first, the extent to which an executive order that mandates suspicionless drug testing of 85,000 state employees violates the Fourth Amendment; and, second, the propriety of the district court’s decision to enjoin the Governor of Florida from testing all 85,000 covered employees.

The district court, confronted with a suspicionless drug testing policy that almost certainly sweeps far too broadly and hence runs afoul of the Fourth Amendment in many of its applications, granted relief that also swept too broadly and captured both the policy’s constitutional applications and its unconstitutional ones. We therefore vacate the district court’s order and remand for further proceedings. Confusion regarding the scope of the relief that the plaintiffs requested has plagued this lawsuit from its inception in 2011.

In that year, Appellant Rick Scott, the Governor of Florida, issued Executive Order 11-58 (“EO”), which mandated two types of suspicionless drug testing: random testing of all employees at state agencies within his control, and pre-employment testing of all applicants to those agencies. Appellee American Federation of State, County, and Municipal Employees Council 79 (“Union”), which represents many employees covered by the EO, sued in the United States District Court for the Southern District of Florida to invalidate the EO, and to enjoin its implementation, as unconstitutional under the Fourth Amendment.

Initially, as the Union itself has conceded, its challenge was exclusively facial in nature and sought to strike down the entire EO rather than to limit its applicability. By the

summary-judgment stage, however, the Union urged the district court to construe its complaint as making both a facial and an as-applied challenge. The Union’s as-applied challenge contended only that the EO was unconstitutional when applied to employees not occupying safety-sensitive positions—a group that the Union estimated to be roughly 60 percent of the covered employees. The district court granted

summary judgment to the Union and denied

summary judgment to the State. In its order, the district court concluded that the State’s justifications for testing all of its employees, including those in non-safety-sensitive positions, were insufficient. The court then turned to the question of what relief it would grant. The district court granted relief that it described as “as-applied”- but that remained essentially facial in .nature: the court invalidated the EO, and enjoined its implementation, as to all 85,000 current state employees.

This relief covered every single employee and disregarded any distinction between safety-sensitive and non-safety-sensitive positions. Yet, as the Supreme Court has established, a party is entitled to facial invalidation of a law on Fourth Amendment grounds only if the party can demonstrate that there are no constitutional applications of that law. In this case, the district court declared the EO unconstitutional as to all current state employees.

This relief swept too broadly, enjoined both constitutional and unconstitutional applications of the EO, and did so without examining the specific job categories to be tested. What the Supreme Court’s case law requires, in contrast, is that the trial court balance the governmental interests in a suspicionless search against each particular job category’s expectation of privacy.

Among the covered state employees, for example, are law enforcement personnel who carry firearms as well as employees tasked with operating heavy machinery or large vehicles—groups that the Supreme Court has held, in a line of precedent beginning with Skinner v. Ry. Labor Execs.’ Ass’n, 489 U.S. 602, 109 S.Ct. 1402, 103 L.Ed.2d 639 (1989), may be drug tested without individualized suspicion. As to those safety-sensitive employees, the EO’s application would most likely be constitutional, and, therefore, the district court’s order cannot stand as written.

The State, however, asks us to do more than vacate and remand. It argues that the Governor is entitled to

summary judgment, and that we should reverse the district court, because the EO is constitutional as applied to all 85,000 state employees. At bottom, the State wants us to approve of a drug testing policy of far greater scope than any ever sanctioned by the Supreme Court or by any of the courts of appeals. In order to meet its burden of justifying the EO, the State offers several reasons, stated only at the highest order of abstraction, for why it can drug test all of its employees without any individualized suspicion.

However, the Supreme Court has approved of suspicionless drug testing only when the government has demonstrated heightened interests, such as a serious threat to public safety, that apply narrowly to specific job categories of employees. Yet during the

summary judgment proceedings, the State refused to provide reasons that apply narrowly to specific job categories, which undoubtedly hindered the district court from conducting its balancing calculus at the proper level of specificity. On remand, the State must meet its burden of demonstrating important special needs on a job-category-by-category basis. Its current arguments have failed to convince us to direct

summary judgment in its favor. I. A. On March 22, 2011, Governor Scott issued Executive Order 11-58. The EO directed all state agencies “within the purview of the Governor ... to provide for pre-employment drug testing for all prospective new hires and for random drug testing of all employees within each agency.” The EO further instructed the agencies to “provide for the potential for any employee ... to be tested at least quarterly.” Approximately 85,000 people, or 77 percent of the State’s workforce, are covered by the EO.

Although the Executive Order does not specify a method of drug testing, the State indicated in the district court that urinalysis would be the method used to implement the testing program. The testing process would afford the person providing the sample “individual privacy” unless there is reason to believe that a particular individual intends to alter or substitute the sample. In addition, the results of the drug tests cannot be used as evidence, obtained in discovery, or otherwise disclosed in any public or private proceeding.

The EO represented a significant expansion of the State’s employee drug testing regime. Prior to the EO’s issuance, Florida’s Drug-Free Workplace Act (“DFWA”), Fla. Stat. § 112.0455, permitted drug testing in more limited instances.

State agencies were authorized to test: job applicants to “safety-sensitive position[s],” meaning “any position, including a supervisory or management position, in which a drug impairment would constitute an immediate and direct threat to public health or safety,” § 112.0455(5)(f) & (m); current employees, if the employer had reasonable suspicion; current employees, if the test was “conducted as part of a- routinely scheduled employee fitness-for-duty medical examination”; and current employees who entered “an employee assistance program for drug-related problems.” See § 112.0455(7)(a)-(d).

This version of the statute notably did not provide for random suspicionless testing of any current employees, even those employed in safety-sensitive positions. Other statutes or administrative regulations provided for suspicionless testing of current employees in specific departments. The Department of Corrections (“DOC”), for instance, provided for random suspicionless testing of its employees. See Fla. Stat. § 944.474. The Department of Juvenile Justice (“DJJ”) also required random suspicionless drug testing of its employees.

The Department of Transportation (“DOT”) and the Department of Environmental Protection (“DEP”), meanwhile, required random suspicionless testing of their safety-sensitive employees, particularly those who held commercial driver’s licenses. In 2012, the Florida Legislature amended the Drug-Free Workplace Act and substantially broadened it. The current version of Fla. Stat. § 112.0455 permits random testing of all employees at three-month intervals, see § 112.0455(7)(c) (2012), and expands the definition of “job applicant” to cover all job applicants, see § 112.0455(5)(f) (2012).

In essence, the current version of the DFWA authorizes what the EO mandates.

The text of the Executive Order offers several justifications for this sweeping policy, including, among others, that: (1) “the State, as an employer, has an obligation to maintain discipline, health, and safety in the workplace”; (2) “illegal drug use has an adverse [ejffect on job performance,” including the risk of absenteeism, greater burden on state health benefit programs, and a decline in productivity; and (3) drug use poses a risk to the public, which “interacts daily with state employees.” Prior to the issuance of the EO, the State had collected data from random drug testing of job applicants and employees at three departments—the Department of Transportation, the Department of Juvenile Justice, and the Department of Corrections.

Random testing at DOT and DJJ yielded positive results in less than one percent of cases between 2008 and 2011; random testing at DOC produced positive results in less than one percent of cases in 2008 and 2009, then increased to 2.4 and 2.5 percent in 2010 and 2011. The State presented this data as evidence that there was a preexisting drug problem among the state employee population. B. On May 31, 2011, before any agency implemented the EO, the Union filed suit, alleging that the EO violated the Fourth Amendment.

Using the terminology of a facial challenge, the Union described its suit as “an action ... for a preliminary injunction and a permanent injunction against the Governor of the State of Florida, ordering him to cease, or not implement, all employee drug-testing mandated by his Executive Order Number 11-58,” and also for “declaratory judgment declaring that the drug-testing regime mandated by Executive Order 11-58 violates the Fourth Amendment of the Constitution.” Compl. ¶ 1.

The gravamen of the complaint was that “[t]he Supreme Court of the United States has held that suspicion-less drug-testing by the government is an unreasonable search violative of the Fourth Amendment, except under certain special circumstances,” none of which applied to the EO. Compl. ¶ 11. More precisely, the EO “violate[d] the Fourth Amendment ... because it command[ed] state agencies to conduct random, suspicionless searches of all employees, without limiting the searches in any way to employees in safety-sensitive positions where there is a concrete danger of real harm.” Compl. ¶ 13.

Regarding its standing, the Union averred that it represented more than 50,-000 employees at the agencies covered by the EO. Its members were subject both to the random testing requirement for current employees as well as the pre-employment testing requirement for new hires because “employees represented by [the Union] who seek a promotion to another job are considered new employees.” Compl. ¶ 15. Thus, the Union “sue[d] on its own behalf’ as well as “in its organizational capacity on behalf of those state employees it represented].” Compl. ¶ 16. In the final

section of the complaint, the Union reiterated its request for facial relief. The Union first asked the district court to declare “that Defendant’s Executive Order 11-58 is quashed because it violates the right of the people to be free from unreasonable searches, under the Fourth Amendment.” The Union further urged the district court to issue a permanent injunction ordering “the Defendant [to] immediately direct all agencies and persons affected by Defendant’s Executive Order 11-58 to cease all drug-testing implemented in compliance with the order.” Compl. at 6-7. C. The parties filed cross motions for

summary judgment. The Union argued that the Executive Order was unconstitutional because it failed to separate safety-sensitive from non-safety-sensitive positions and thus moved the district court to issue both a declaratory judgment declaring that the EO violated the Fourth Amendment and a permanent injunction barring the EO’s implementation. Notably, at this stage, the Union began recasting its complaint in the terminology of an as-applied challenge.

The Union stressed that it “ehallenge[d] only the new drug-testing regime that tests the rest of the State’s workers [not covered by the then-current version of Fla. Stat. § 112.0455]—those not suspected of drug abuse and those who don’t hold safety-sensitive jobs.” And, in its opposition to the State’s cross-motion for

summary judgment and its reply brief, the Union expressly insisted it had made an as-applied challenge.

The Union argued that “the Complaint, fairly read, clearly put the Governor on notice that [the Union] was bringing both a facial and as-applied challenge,” and that its as-applied challenge contended that the statute was “unconstitutional as applied to [Union] bargaining unit members who are not reasonably suspected of drug abuse and who are not in safety-sensitive positions.” The Union further clarified that, for purposes of its as-applied challenge, it was “not challenging drug-testing of those in safety-sensitive positions.” In support of its motion, the State argued: (1) that the Union lacked standing; (2) that the Union could not succeed on what the State maintained was a facial challenge to the Executive Order; (3) that, on the merits, the EO was constitutional because individuals consented to the test; or, alternatively, (4) that the EO was constitutional because the State had a- special need justifying suspicionless drug testing.

In its special-needs analysis, the State offered its interest in a safe, productive, and efficient workplace as the primary need justifying the EO.

The State expressly declined to specify which groups of employees presented heightened safety concerns, instead arguing generally that “even if safety concerns were the only permissible justification, the notion that only intoxicated employees with certain duties present a danger to others ... is untenable.” Thus, according to the State, the proffered safety need applied across the board and to all employees: An employee need not drive a train, carry a gun, or interdict drugs to present a safety risk.

Even a desk-bound clerk ... may become violent with other employees or the public, may present a danger when driving a car in the workplace parking lot, or may exercise impaired judgment when encountering any of the myriad hazards that exist in the workplace environment .... The State also asserted that the privacy interests of state employees were diminished for several reasons. First, drug testing among private employers had become common. Second, Florida had a tradition of open government. Finally, the policy was clearly announced, so employees could not have any expectation of privacy.

As for the Union’s as-applied challenge, the State declined to meet it head-on. Instead, it argued only that the district court should reject the Union’s attempt to recast its pleadings because “pri- or to the about-face in its [opposition [to defendant’s motion for

summary judgment], Plaintiff repeatedly relied on the solely facial nature of its claim.” According to the State, therefore, the district court should consider and reject only the Union’s facial challenge. D. On April 25, 2012, the district court granted

summary, judgment to the Union and, in turn, denied the State’s motion. After finding that the Union had standing to challenge the Executive Order, the district court conducted the special-needs balancing test established in Skinner and weighed the State’s asserted public interests against the employees’ privacy interests.

The district court first determined that the public interests asserted were “notably broad and general compared to the interests that the Supreme Court ... held justify suspicionless drug testing.” The court then rejected the State’s assertion that state employees possessed a diminished privacy interest. The district court therefore concluded that the EO was unconstitutional. The district court turned to crafting the remedy.

Although the State argued that the Union had mounted exclusively a facial challenge, the court pointed out that the Union had conceded that the Fourth Amendment permitted drug testing of state employees in safety-sensitive positions. However, the district court then characterized the Union’s challenge “as consistent with an ‘as-applied’ challenge .... [that] asserts at most that the EO cannot be constitutionally applied to any current employee at a covered agency.” (Emphasis added.) Accordingly, the district court granted far more sweeping relief than was consistent with the Union’s concession.

The Court granted a declaratory judgment holding the EO unconstitutional and issued an injunction coextensive with that declaration, which barred drug testing of “both Union and non-Union employees .... currently employed at covered agencies” as of the date of the district court’s order. In short, the district court struck down the EO insofar as it covered all 85,000 current state employees.

The only thing that the judgment and injunction did not address was the application of the EO to “pre-employment testing of non-current employees,” a group the district court labeled “prospective new hires,” and “the random testing of those hired after the issuance of the EO.” The State timely appealed. II.

Summary judgment is appropriate where “the movant shows that there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law.” Fed.R.Civ.P. 56(a). We review a district court’s grant of

summary judgment de novo, viewing the facts and drawing all reasonable inferences in the light most favorable to the non-moving party. Moore ex rel. Moore v. Reese, 637 F.3d 1220, 1231 (11th Cir.2011). We review the decision to grant a permanent injunction for abuse of discretion but review the district court’s underlying legal conclusions de novo. See Alabama v. Ctrs. for Medicare & Medicaid Servs., 674 F.3d 1241, 1244 n. 2 (11th Cir.2012). A. The parties first dispute whether the relief the district court granted in this case was facial or as-applied in nature.

Although the boundary between these two forms of relief is not always clearly or easily demarcated, the district court’s decision to strike down the EO and enjoin its implementation as to all 85,000 current employees has the essential characteristics of facial relief. From the outset, the Union mounted a facial challenge to the Executive Order. That much is apparent from the face of the complaint. We look to the scope of the relief requested to determine whether a challenge is facial or as-applied in nature. See Doe v. Reed, — U.S. -, 130 S.Ct. 2811, 2817, 177 L.Ed.2d 493 (2010).

The heart of the Union’s requested remedy was two-fold: first, that the district court broadly declare “that Defendant’s Executive Order 11-58 is quashed because it violates the right of the people to be free from unreasonable searches, under the Fourth Amendment”; and, second, that the district court issue an injunction ordering “the Defendant [to] immediately direct all agencies and persons affected by Defendant’s Executive Order 11-58 to cease all drug-testing implemented in compliance with the order.” Compl. at 6-7 (emphasis added). There can be no doubt that this relief would be facial in nature.

And, indeed, the Union expressly maintained that its challenge was facial prior to filing a motion for

summary judgment. However, the Union began requesting both facial and as-applied relief at the

summary-judgment stage. In requesting as-applied relief, the Union explained that it “challenged only the new drug-testing regime that tests ... those not suspected of drug abuse and those who don’t hold safety-sensitive jobs,” and that it was “not challenging drug-testing of those in safety-sensitive positions.” The Union identified the non-safety-sensitive category of employees to be roughly 60 percent of all employees covered by the EO.

Insofar as the Union mounted a facial challenge to the Executive Order— and it surely did that—it had to meet an especially demanding standard. “A facial challenge, as distinguished from an as-applied challenge, seeks to invalidate a statute or regulation itself.” United States v. Frandsen, 212 F.3d 1231, 1235 (11th Cir.2000). “[W]hen a plaintiff mounts a facial challenge to a statute or regulation, the plaintiff bears the burden of proving that the law could never be applied in a constitutional manner.” DA Mortg., Inc. v. City of Miami Beach, 486 F.3d 1254, 1262 (11th Cir.2007).

Put another way, “the challenger must establish that no set of circumstances exists under which the Act would be valid.” United States v. Salerno, 481 U.S. 739, 745, 107 S.Ct. 2095, 95 L.Ed.2d 697 (1987). The Supreme Court reaffirmed Salerno’s validity as recently as 2010, see United States v. Stevens, 559 U.S. 460, 130 S.Ct. 1577, 1587, 176 L.Ed.2d 435 (2010) (citing Salerno, 481 U.S. at 745, 107 S.Ct. 2095), and, just last year, a panel of this Court reiterated that the strict “no set of circumstances” test is the proper standard for evaluating a facial challenge. GeorgiaCarry.Org, Inc. v.

Georgia, 687 F.3d 1244, 1255 & n. 19 (11th Cir.2012). Salerno also applies when a court grants relief that is quasi-facial in nature—that is, relief that reaches beyond the plaintiffs in a case. In Doe v. Reed, for instance, the Supreme Court considered a challenge that a state law violated the First Amendment when applied to referendum petitions. 130 S.Ct. at 2817.

The Court noted that characterizing the challenge.as either facial or as-applied was problematic because the challenge “obviously ha[d] characteristics of both: The claim [wa]s ‘as applied’ in the sense that it d[id] not seek to strike the PRA in all its applications, but only to the extent it covers referendum petitions. The claim [wa]s ‘facial’ in that it [wa]s not limited to plaintiffs’ particular case .... ” Id. When a plaintiff brings this sort of quasi-facial challenge, “[t]he label is not what matters.” Id.

Where “an injunction ... reach[es] beyond the particular circumstances of these plaintiffs,” it “must therefore satisfy [the Supreme Court’s] standards for a facial challenge to the extent of that reach.” Id. Prior to considering the propriety of the Union’s facial challenge, the district court correctly attempted to construe the Union’s complaint as making a more limited, as-applied challenge to the EO.

The State objects that the district court could not have construed the Union’s suit as an as-applied challenge at all because the Union’s complaint requested only facial relief and the Union insisted during discovery that it was mounting a facial challenge. This objection is unconvincing. Ordinarily, it is true that, “[a]t the

summary judgment stage, the proper procedure for plaintiffs to assert a new claim is to amend the complaint in accordance with Fed.R.Civ.P. 15(a). A plaintiff may not amend her complaint through argument in a brief opposing

summary judgment” or one advocating

summary judgment. Gilmour v. Gates, McDonald & Co., 382 F.3d 1312, 1315 (11th Cir.2004). In this case, however, the Union was not stating a new claim, only clarifying the scope of its desired remedy. As the Supreme Court has explained, “the distinction between facial and as-applied challenges is not so well defined that it has some automatic effect or that it must always control the pleadings and disposition in every case involving a constitutional challenge. The distinction ... goes to the breadth of the remedy employed by the Court, not what must be pleaded in a complaint.” Citizens United v. Fed.

Election Comm’n, 558 U.S. 310, 331, 130 S.Ct. 876, 175 L.Ed.2d 753 (2010); see Jacobs v. Fla. Bar, 50 F.3d 901, 905 n. 17 (11th Cir.1995) (we are not bound by a party’s characterization of the complaint as facial, but rather look to whether “the complaint sets forth a cause ' of action for an as-applied challenge”). As a general matter, courts strongly disfavor facial challenges, and for good reason: Claims of facial invalidity often rest on speculation. As a consequence, they raise the risk of premature

interpretation of statutes on the basis of factually barebones records. Facial challenges also run contrary to the fundamental principle of judicial restraint that courts should neither anticipate a question of constitutional law in advance of the necessity of deciding it nor formulate a rule of constitutional law broader than is required by the precise facts to which it is to be applied. Finally, facial challenges threaten to short circuit the democratic process by preventing laws embodying the will of the people from being implemented in a manner consistent with the Constitution. Wash. State Grange v. Wash.

State Republican Party, 552 U.S. 442, 450-51, 128 S.Ct. 1184, 170 L.Ed.2d 151 (2008) (citations and internal quotation marks omitted). Thus, courts construe a plaintiffs challenge, if possible, to be as applied. See Jacobs, 50 F.3d at 905 n. 17; see also Stupak-Thrall v. United States, 89 F.3d 1269, 1288 (6th Cir.1996) (Boggs, J., dissenting) (“[U]nless a plaintiff expressly disavows an ‘as:applied’ challenge, the complaint that a regulation is invalid should be construed, if possible, as an as-applied challenge.”).

However, despite explicitly saying that it was granting only as-applied relief, the district court in this case granted what effectively amounted to facial relief by declaring the Executive Order unconstitutional and enjoining its application to all 85,000 current employees. As the district court itself acknowledged, the concession that transformed the lawsuit into an as-applied challenge was the Union’s admission that the Fourth Amendment permitted drug tests of state employees in safety-sensitive positions.

Yet the district court did not follow that reasoning to its necessary conclusion, which was that the proper scope of the as-applied challenge— and the scope of the relief that it could have granted based on the Union’s motion for

summary judgment—was limited to those employees not occupying safety-sensitive positions. Instead, the district court characterized the Union’s concession “as consistent with an ‘as-applied’ challenge .... [that] asserts at most that the EO cannot be constitutionally applied to any current employee at a covered agency.” (Emphasis added.) In doing so, the district court attached an as-applied label to what essentially amounted to a facial challenge concerning all 85,000 current state employees. This led the district court to grant both a declaratory judgment and a corresponding injunction that were too broad.

In determining the scope of its relief, the court began by dividing the individuals subject to the EO into three groups: (1) employees at the covered agencies prior to the issuance of district court’s order; (2) “prospective new hires,” which meant “individuals who are not currently employed at covered agencies”; and (3) employees at the covered agencies hired after the district court’s order. The district court then granted the Union declaratory judgment declaring the EO unconstitutional, and an injunction that mirrored the scope of that declaration, as to the first group.

The court stated that its order left “unresolved” the question of the EO’s constitutionality with regard to the latter two groups, since “[t]he Union ma[de] no claims as to the constitutionality of the EO as it relates to pre-employment testing of non-current employees, or the random testing of those hired after the issuance of the EO.” This limitation, however, did not transform the district court’s relief from facial to as-applied.

As we’ve said, the line between facial and as-applied relief is a fluid one, and many constitutional challenges may occupy an intermediate position on the spectrum between purely as-applied relief and complete facial invalidation. The Supreme Court itself has weighed challenges with both facial and as-applied characteristics, see, e.g., Doe, 130 S.Ct. at 2817, and perhaps the best understanding of constitutional challenges is that “[t]here is no single distinctive category of facial, as opposed to as-applied, litigation.” Richard H. Fallon, Jr., As-Applied and Facial Challenges and Third-Party Standing, 113 Harv.

L.Rev. 1321, 1321 (2000). As both parties acknowledged at oral argument, the district court’s order has characteristics of both facial and as-applied relief. On the one hand, it reaches far beyond the scope of the Union’s as-applied challenge and encompasses all current state employees. On the other hand, the district court did not invalidate the EO in its entirety. Nonetheless, we conclude that the district court granted what effectively amounted to facial relief—or, at the very least, relief that had enough characteristics of facial relief to demand satisfaction of Salerno’s rigorous standard.

The essential point is that the district court invalidated the EO across the board covering all 85,000 state employees, the overwhelming majority of those subject to the EO. The scope of the district court’s judgment is extremely broad and, notably, its relief was not limited in any way by the concession the Union itself made: “[0]n March 22, 2011 (the date of promulgation) there was at least one employee ... who held a high-risk, safety-sensitive job, and was subject to EO 11-58.

And we admit that the Fourth Amendment does not bar the random drug testing of government employees in high-risk, safety-sensitive jobs.” Notwithstanding that concession, the district court’s judgment and injunction bar the State from testing that employee and, indeed, any other current employee who, for example, occupies a law enforcement position that requires carrying a firearm. Nor does the district court’s cutoff of the scope of its judgment and the accompanying injunction transform that relief into as-applied relief.

The district court invalidated the Executive Order and enjoined its implementation as to the vast majority of individuals covered by the EO. To be sure, the district court did not declare unconstitutional or enjoin the implementation of the entirety of the EO. But the district court’s decision not to cover pre-employment testing of prospective new hires does not alter our view that the relief it did grant was facial as to all -85,000 current employees.

If a statute has two distinct provisions, and a court strikes down one as unconstitutional (and indeed, one that covers so many employees)* we would not say that the relief was as-applied simply because a part of the statute remains. Rather, we would say that, as to the provision. the court struck down, the plaintiff obtained facial relief. See Doe, 130 S.Ct. at 2817. Here, that is precisely what the Union received.

Rather than conducting any kind of job-category-by-category inquiry, and narrowly tailoring its decision to the precise contours of the constitutional violation, the district court facially -invalidated the provision of the Executive Order that provides “for random drug testing of all employees within each agency.” B. Having established that the district court granted facial relief, the essential question becomes whether that relief could meet Salerno’s- demanding standard.

To uphold the scope of the relief, we would have to be convinced that the State could never constitutionally require any of the 85,000 current state employees protected by the injunction to submit to a suspicion-less drug test. But the answer, plainly, is that there are some (how many is unclear) current state employees as to whom suspicionless drug testing is constitutionally permissible.

This conclusion ineluctably follows from the line of Supreme Court precedent beginning with Skinner, which held that the Fourth Amendment permits suspicionless drug testing of certain safety-sensitive categories of employees—for instance, employees who operate or pilot large vehicles, or law enforcement officers who carry firearms in the course of duty. The Fourth Amendment protects “[t]he right of the people to be secure in their persons, houses, papers, and effects, against unreasonable searches and seizures,” U.S. Const, amend.

IV, and applies to the states through the Due Process Clause of the Fourteenth Amendment. See City of Ontario v. Quon, 560 U.S. 746, 130 S.Ct. 2619, 2624, 177 L.Ed.2d 216 (2010). Testing a urine sample, which “can reveal a host of private medical facts about an employee,” and which entails a process that “itself implicates privacy interests,” is a search. Skinner, 489 U.S. at 617, 109 S.Ct. 1402; see also Chandler v. Miller, 520 U.S. 305, 313, 117 S.Ct. 1295, 137 L.Ed.2d 513 (1997). The basic question we are required to answer when confronted with a drug-testing policy is whether this search is reasonable.

Chandler, 520 U.S. at 313, 117 S.Ct. 1295. While “[i]n the criminal context, reasonableness usually requires a showing of probable cause” to obtain a search warrant, that standard is “unsuited to determining the reasonableness of administrative searches where the ‘Government seeks to prevent the development of hazardous conditions.’” Bd. of Educ. v. Earls, 536 U.S. 822, 828, 122 S.Ct. 2559, 153 L.Ed.2d 735 (2002) (quoting Nat’l Treasury Emps. Union v. Von Raab, 489 U.S. 656, 667-68, 109 S.Ct. 1384, 103 L.Ed.2d 685 (1989)).

The default rule in this context, therefore, is that “[t]o be reasonable under the Fourth Amendment, a search ordinarily must be based on individualized suspicion of wrongdoing.” Chandler, 520 U.S. at 313, 117 S.Ct. 1295. While individualized suspicion is the normal requirement, “particularized exceptions to the main rule are sometimes warranted based on ‘special needs, beyond the normal need for law enforcement.’ ” Id. (quoting Skinner, 489 U.S. at 619, 109 S.Ct. 1402).

When the government alleges that special needs justify this Fourth Amendment intrusion, “courts must undertake a context-specific inquiry, examining closely the competing private and public interests advanced by the parties.” Id. at 314, 117 S.Ct. 1295. “In limited circumstances, where the privacy interests implicated by the search are minimal, and where an important governmental interest furthered by the intrusion would be placed in jeopardy by a requirement of individualized suspicion, a search may be reasonable despite the absence of such suspicion.” Skinner, 489 U.S. at 624, 109 S.Ct. 1402.

Therefore, the test we apply is a job-category-by-category balancing of “the individual’s privacy expectations against the Government’s interests,” Von Raab, 489 U.S. at 665, 109 S.Ct. 1384, with other relevant factors being “the character of the intrusion”—particularly whether the collection method affords a modicum of privacy, see Vernonia Sch. Dist. 47J v. Acton, 515 U.S. 646, 658, 115 S.Ct. 2386, 132 L.Ed.2d 564 (1995)—and the efficacy of the testing regime, see Chandler, 520 U.S. at 319-20, 117 S.Ct. 1295.

At times, the Supreme Court has described the interests justifying suspicionless drug testing as “compelling.” See Von Raab, 489 U.S. at 670, 109 S.Ct. 1384; Skinner, 489 U.S. at 628, 109 S.Ct. 1402. In Vemonia, the Court clarified that “[i]t is a mistake, however, to think that the phrase ‘compelling state interest,’ in the Fourth Amendment context, describes a fixed, minimum quantum of governmental concern,” and therefore we cannot “dispose of a case by answering in isolation the question: Is there a compelling state interest here?” 515 U.S. at 661, 115 S.Ct. 2386.

Rather, a compelling interest is one “important enough to justify the particular search at hand, in light of other factors that show the search to be relatively intrusive upon a genuine expectation of privacy.” Id. The Supreme Court has had five occasions to evaluate suspicionless drug testing policies in the last twenty-five years. We therefore know the kinds of interests that are important enough to subject certain limited categories of individuals to suspicionless drug tests, and, moreover, we know that some of the 85,000 current state employees fall within those categories.

In Skinner, the Supreme Court established that the government has a compelling need to test railroad employees. In that case, the Federal Railroad Administration - (“FRA”) required suspicionless drug testing of workers involved in railroad accidents. 489 U.S. at 606, 109 S.Ct. 1402. As for the first factor in the balancing test, the FRA’s interest, the Court’s inquiry focused intently on the special characteristics of the railroad industry, where on-the-job intoxication was “a significant problem” that had resulted in “21 significant train accidents” in a ten-year period. Id. at 607, 109 S.Ct. 1402.

On the other side of the ledger, the Court reasoned that “the expectations of privacy of covered employees [we]re diminished by reason of their participation in an industry that is regulated pervasively to ensure safety.” Id. at 627, 109 S.Ct. 1402. As the Court pointed out, railroad “employees ha[d] long been a principal focus of regulatory concern,” with various federal laws subjecting railroad employees’ physical fitness to testing and regulation. See id. at 627-28, 109 S.Ct. 1402. The two other factors were the character of the intrusion and the efficacy of the policy.

The FRA’s urine testing was not overly intrusive because it did not require direct observation, id. at 626, 109 S.Ct. 1402, and testing was effective because it “deterr[ed] employees engaged in safety-sensitive tasks from using controlled substances or alcohol in the first place.” Id. at 629, 109 S.Ct. 1402; accord id. at 631-32, 109 S.Ct. 1402. In light of these factors, most notably the serious risks to public safety implicated by this specific category of employees, the Court upheld the constitutionality of the FRA’s policy. See id. at 633, 109 S.Ct. 1402.

The principle we draw from Skinner is that government “employees ... engaged in safety-sensitive tasks,” id. at 620, 109 S.Ct. 1402, particularly those involved with the operation of heavy machinery or means of mass transit, may be subject to suspicionless drug testing. In Von Raab, the Supreme Court identified several other job categories that a suspicionless drug testing policy may cover.

At issue in that case was the United States Customs Service’s required urinalysis testing for three job categories: first, those directly involved in drug interdiction; second, those who carried firearms; and third, those who handled classified material. 489 U.S. at 660-61, 109 S.Ct. 1384. The Court began by identifying the government’s special needs with regard to the first two categories. Id. at 668, 109 S.Ct. 1384.

Customs employees responsible for drug interdiction were “exposed to th[e] criminal element and to the controlled substances it s[ought] to smuggle into the country”; the Customs Service was concerned not only about those employees’ “physical safety” but also the risk of bribery or corruption. See id. at 669, 109 S.Ct. 1384. Thus, the Supreme Court found that “the Government ha[d] a compelling interest in ensuring that front-line interdiction personnel [we]re physically fit, and ha[d] unimpeachable integrity and judgment.” Id. at 670, 109 S.Ct. 1384. Similar logic applied to those who carried firearms.

Employees “who may use deadly force plainly discharge duties fraught with such risks of injury to others that even a momentary lapse of attention can have disastrous consequences.” Id. (internal quotation marks omitted). As for the privacy interests implicated by the search, the Supreme Court began by noting that “certain forms of public employment may diminish privacy expectations even with respect to such personal searches.” Id. at 671, 109 S.Ct. 1384.

The Court explained that, “[ujnlike most private citizens or government employees in general, employees involved in drug interdiction reasonably should expect effective inquiry into their fitness and probity.

Much the same is true of employees who are required to carry firearms.” Id. at 672, 109 S.Ct. 1384. “Because successful performance of their duties depends uniquely on their judgment and dexterity, these employees cannot reasonably expect to keep from the Service personal information that bears directly on their fitness,” and thus their privacy could not “outweigh the Government’s compelling interests in safety and in the' integrity of our borders.” Id. As for employees who handled classified information, however, the Court remanded.

While noting that the protection of “truly sensitive information” is “compelling,” id. at 677, 109 S.Ct. 1384, the Court questioned the Customs 'Service’s designation of several classes of employees—for instance, baggage clerks and messengers—as belonging to this category. See id. at 678, 109 S.Ct. 1384. Since the Court could not determine “whether the Service ha[d] defined this category of employees more broadly than is necessary,” it remanded for the lower courts to determine more precisely which employees truly dealt with sensitive information. See id.

The Supreme Court next approved of suspicionless drug testing in a far different context than government employment: schools. The Court upheld the constitutionality of two schools’ policies of randomly drug testing student athletes, Vernonia, 515 U.S. at 648, 115 S.Ct. 2386, and students participating in competitive extracurricular activities, Earls, 536 U.S. at 825, 122 S.Ct. 2559. The Supreme Court found that there was a special need in the public school context, where teachers were responsible for their young charges.

See Vernonia, 515 U.S. at 661, 115 S.Ct. 2386 (“Deterring drug use by our Nation’s schoolchildren is at least as important as enhancing efficient enforcement of the Nation’s laws against the importation of drugs ... or deterring drug use by engineers and trainmen .... ”); Earls, 536 U.S. at 829, 122 S.Ct. 2559. As for the students’ privacy interests, the Court noted that the students by definition were “(1) children, who (2) have been committed to the temporary custody of .the State as schoolmaster.” Vernonia, 515 U.S. at 654, 115 S.Ct. 2386.

The State, acting in loco parentis, exercised “a degree of supervision and control that could not be exercised over free adults.” Id. at 655, 115 S.Ct. 2386; see Earls, 536 U.S. at 831, 122 S.Ct. 2559. Those diminished privacy interests could not overcome the government’s important interests in protecting children from drug use. See Vernonia, 515 U.S. at 665, 115 S.Ct. 2386; Earls, 536 U.S. at 838, 122 S.Ct. 2559.

In contrast to the preceding cases, the Supreme Court rejected a Georgia statute that required all candidates for certain state offices to submit to a drug test at a time of their choosing prior to the election. See Chandler, 520 U.S. at 309-10, 117 S.Ct. 1295. Georgia attempted to justify its policy based on “the incompatibility of unlawful drug use with holding high state office,” contending that illegal drug use “draws into question an official’s judgment and integrity” and “jeopardizes the discharge of public functions.” Id. at 318, 117 S.Ct. 1295.

The Court dismissed these broad and general rationales, finding “[n]otably lacking ... any indication of a concrete danger demanding departure from the Fourth Amendment’s main rule.” Id. at 318-19, 117 S.Ct. 1295. Unlike the railroad employees in Skinner or the law enforcement officers in Von Raab, “th[e Georgia] officials typically d[id] not perform high-risk, safety-sensitive tasks, and the required certification immediately aid[ed] no interdiction effort.” Id. at 321-22, 117 S.Ct. 1295.

Worse still, Georgia’s testing program was not even well-crafted to detect drug use, since the candidates themselves scheduled the drug test and could easily evade a positive result. Id. at 319-20, 117 S.Ct. 1295. The Supreme Court therefore had little trouble declaring this policy unconstitutional. Although this Court recently has addressed the constitutionality of suspicion-less drug testing in a different context, see Lebron v. Sec’y, Fla.

Dep’t of Children & Families, 710 F.3d 1202, 1218 (11th Cir.2013) (affirming a preliminary injunction barring suspicionless testing of welfare recipients), we have not considered the propriety of testing current or potential government employees since Chandler v. Miller, 73 F.3d 1543 (11th Cir.1996), rev’d, 520 U.S. 305, 117 S.Ct. 1295, 137 L.Ed.2d 513 (1997). Our sister circuits, however, have confronted a wide variety of drug testing policies and have identified several other safety-sensitive job categories.

In cases similar to Skinner, the courts of appeals have upheld suspicion-less drug testing of categories of employees whose work involves heavy machinery or the operation of large vehicles, such as planes, trains, buses, or boats. Thus, although Skinner itself addressed railroad employees, the courts of appeals have extended its logic to those involved in the operation of aircraft. See, e.g., Bluestein v. Skinner, 908 F.2d 451, 457 (9th Cir.1990); Nat’l Fed’n of Fed. Emps. v. Cheney, 884 F.2d 603, 610-11 (D.C.Cir.1989).

Another category—a natural extension of the Supreme Court’s holding in Von Raab—encompasses police officers, see Carroll v. City of Westminster, 233 F.3d 208, 213 (4th Cir.2000), correctional officers who interact with parolees or inmates in a prison, see Int’l Union v. Winters, 385 F.3d 1003, 1013 (6th Cir.2004), and firefighters, see Hatley v. Dep’t of the Navy, 164 F.3d 602, 604 (Fed.Cir.1998).

The crucial point is that, to affirm the district court’s declaration and injunction in this case, we would have to find that none of the 85,000 current employees covered by the district court’s relief belong to the special-needs categories identified by the Supreme Court. However, the Union’s own submissions belie this.

Indeed, the Union itself observed that, “[o]f the approximately 85,000 employees in 2010, 33,052 of them ... served in arguably safety-sensitive positions.” More precisely, during discovery, the Union asked the State to identify: • “How many employees affected by EO 11-58 regularly carry firearms on the job?” (Interrogatory 16) • “How many employees affected by EO 11-58 are sworn law enforcement officers?” (Interrogatory 17) • “How many employees affected by ...

EO 11-58 regularly interact on the job with detainees in the correctional system?” (Interrogatory 18) • “How many employees affected by EO 11-58 regularly interact on the job with primary or secondary school students?” (Interrogatory 19) • “How many employees affected by EO 11-58 regularly work as mass transit operators?” (Interrogatory 20) • “How many employees affected by EO 11-58 regularly work as transportation safety inspectors?” (Interrogatory 21) The State provided fairly detailed figures in its responses, including, for example, the following categories of employees who carry firearms: 157 employees in the Department of Business & Professional Regulation, 146 inspectors in the Department of Corrections (along with another 1,088 employees who were authorized but not required to carry firearms), 136 employees in the Department of Environmental Protection, and 23 in the Department of Military Affairs.

Based on the holding in Von Raab, it is apparent that, at least as to these employees, the EO is very likely constitutionally applicable. The State further identified several distinct categories of employees who operate heavy machinery or large vehicles, with almost a thousand working for the Department of Transportation alone. Skinner makes it likely that the State also may subject these, or at least some of these, employees to suspicionless drug testing.

Yet by extending the declaratory judgment and injunction to all current employees, the district court effectively disregarded these portions of the record and barred testing of the safety-sensitive employees included among the 85,000 current employees. Under Salerno, the EO could not possibly be unconstitutional as to all current employees, and the district court’s order therefore cannot “satisfy [the Supreme Court’s] standards for a facial challenge to the extent of [the order’s] reach.” Doe, 130 S.Ct. at 2817.

Since it is well-settled that a district court abuses its discretion when it grants relief that is improperly or even unnecessarily broad, see Alley v. U.S. Dep’t of Health & Human Servs., 590 F.3d 1195, 1205 (11th Cir.2009), we vacate and remand the judgment and the injunction for the district court to more precisely tailor its relief to the extent the Executive Order may be unconstitutional. Nonetheless, the Union maintains that the scope of the injunction was proper anyhow and fell well within the district court’s broad discretion.

In fact, the Union continues to assert that the court “was also within its discretion to award facial relief’ because the Union had demonstrated that no set of circumstances exists under which the EO would be valid. This places the Union’s arguments in palpable tension. On the one hand, it concedes that suspicionless drug testing of safety-sensitive employees would be constitutional. On the other hand, it maintains that the EO is facially unconstitutional. The way that the Union squares the circle is by misapplying Salerno’s “no set of circumstances” test.

According to the Union, the Executive Order requires suspicionless drug testing of all employees, and “there are no circumstances in which suspicionless drug testing of all employees and applicants would be constitutional.” Therefore, the EO fails across the board. Under the Union’s

interpretation of Salerno’s test, a single application of the EO means its application to all employees. But under Salerno and our precedents, see, e.g., Harris v. Mexican Specialty Foods, Inc., 564 F.3d 1301, 1313 (11th Cir. 2009) (“Th[e] mere possibility of a constitutional application is enough to defeat a facial challenge to [a] statute.”), a single “application” of the EO must mean the suspicionless drug test of a single employee. The EO is facially valid, in other words, if the Fourth Amendment permits at least one covered employee to be tested.

The Union’s position completely inverts Salerno and renders a facial attack, far from being the “most difficult” of challenges, 481 U.S. at 745, 107 S.Ct. 2095, the easiest to make. To prevail under the Union’s version of Salerno, the Union needs to show only one employee as to whom suspicionless drug testing is unconstitutional. Then, it would follow, the EO is unconstitutional as a whole because there is no way that testing of all employees is constitutional.

Under the correct understanding of Salerno, we are compelled to conclude that the EO is not facially invalid since safety-sensitive employees may be subjected to suspicionless drug testing. The Union offers another argument: that the district court was required to facially invalidate the EO because otherwise the court would have been “put in the untenable position of having to rewrite” it. The Union claims that the Supreme Court’s case law cautions against partial invalidation and cites Ayotte v. Planned Parenthood of N. New England, 546 U.S. 320, 126 S.Ct. 961, 163 L.Ed.2d 812 (2006).

Ayotte, however, hardly supports this proposition. As the Supreme Court stated in that case, “the ‘normal rule’ is that ‘partial, rather than facial, invalidation is the required course,’ such that a ‘statute may ... be declared invalid to the extent that it reaches too far, but otherwise left intact.’ ” Id. at 329, 126 S.Ct. 961 (quoting Brockett v. Spokane Arcades, Inc., 472 U.S. 491, 504, 105 S.Ct. 2794, 86 L.Ed.2d 394 (1985)). In Sabri v.

United States, the Court identified the “few settings” in which it had “recognized the validity of facial attacks alleging over-breadth (though not necessarily using that term)”: free speech, the right to travel, abortion rights (the category to which Ayotte itself belongs), and legislation under § 5 of the Fourteenth Amendment. 541 U.S. 600, 609-10, 124 S.Ct. 1941, 158 L.Ed.2d 891 (2004). As the Court put it, “[o]utside these limited settings, and absent a good reason, we do not extend an invitation to bring overbreadth claims.” Id. at 610, 124 S.Ct. 1941.

The Supreme Court has not sanctioned this type of facial invalidation in the Fourth Amendment context, and we can discern no basis to do so here. C. As a fallback position, the Union suggests that we could refashion the judgment and injunction simply by cutting them down to cover only those categories of employees as to whom the Executive Order’s application is unconstitutional. While an appellate court undoubtedly has the power to modify injunctions, see United, States v. Nat’l Treasury Emps.

Union, 513 U.S. 454, 480, 115 S.Ct. 1003, 130 L.Ed.2d 964 (1995), or to affirm a judgment as to some plaintiffs but not others, see Allen v. Bd. of Pub. Educ., 495 F.3d 1306, 1320 (11th Cir.2007), we decline to do so because the sort of fact-intensive line-drawing required is a task that properly belongs to the district court. Unlike the typical case where we may affirm a judgment as to some plaintiffs but not as to others, we are dealing here not with a manageable number of individual plaintiffs but with a current workforce of some 85,-000 state employees.

Nor is the district court’s order as amenable to modification as the injunction in Nat’l Treasury Emps. Union, which the Supreme Court altered solely to exclude non-plaintiffs. In sharp contrast, in order to modify the judgment and injunction before us, we would be required to “differentiate[ ] between job categories designated for testing,” scrutinize the State’s rationale for testing each job category, and “conductf ] the balancing test” laid out in Skinner and its progeny. See Nat’l Fed’n of Fed. Emps. v. Vilsack, 681 F.3d 483, 489 (D.C.Cir.2012).

As it currently stands, the district court’s order does not break down the covered employees on a job-category-by-category basis, which leaves us with little basis for determining which portions of the declaratory judgment and the injunction are proper. Thus, while we could simply enjoin the EO as to all employees except those in certain safety-sensitive job categories—those who carry firearms in the course of law-enforcement duties, for instance, or those who operate heavy machinery—and end up probably being right, we would be pronouncing the law without really knowing the facts. Cf. United States v.

Banks, 347 F.3d 1266, 1271-72 (11th Cir.2003). Although the Union did divide the covered employees at least into an “arguably” safety-sensitive group (encompassing roughly 40 percent of all covered employees) and a non-safety-sensitive group, we understand that the Union’s position is that some of the employees in the arguably safety-sensitive group actually are not subject to suspicionless testing, while the State’s position is that some employees in the non-safety-sensitive group are subject to suspicionless testing.

Thus, for instance, the State included all employees at the Department of Corrections within its answer to the Union’s interrogatories. The Union will undoubtedly contest whether some categories of DOC employees should be included within the safety-sensitive category. Meanwhile, the State may be able to identify job categories that the Union has labeled non-safety-sensitive but that actually present real, substantial, and immediate threats to public safety. The Union’s interrogatories, for instance, never asked about the number of doctors or medical personnel employed by the State.

Yet some courts of appeals have held that government-employed medical residents or emergency medical technicians are safety-sensitive employees. See Pierce v. Smith, 117 F.3d 866, 874 (5th Cir.1997); Piroglu v. Coleman, 25 F.3d 1098, 1102 (D.C.Cir.1994).

In light of the wholly undefined nature of the non-safety-sensitive group, and the fact that the current division was found only in the submission of one party in an answer to some interrogatories rather than in the district court’s own finding, we are convinced that determining the proper composition of those groups is a task best left to the district court in the first instance. In order for the district court to accomplish this task, the parties must provide the court with more extensive, job-category-specific facts than the record currently contains.

It is difficult to imagine how this category-specific balancing task can be accomplished without additional discovery. Thus, we vacate and remand both the declaratory judgment and the corresponding injunction in order for the district court to conduct further factfinding and to recraft its relief to cover only those groups as to which the Executive Order’s application is unconstitutional. III. The State does not ask us merely to vacate and remand; boldly, it urges us to reverse the denial of its

summary judgment motion and to direct the district court to grant judgment in its favor. The State argues that there is no need for the district court to conduct the very job-category-by-category balancing that the Supreme Court’s case law commands. Instead, the State offers several reasons that, it claims, can justify suspicionless drug testing of all 85,000 government employees regardless of the nature of their specific job functions. Based on these generic reasons, the State asks us to approve a testing policy of unprecedented scope. We are unpersuaded.

The State’s arguments, which are stated so abstractly, cannot. satisfy the special-needs balancing test laid out in Skinner and its progeny. Those cases conducted the special-needs balancing test not at a high order of generality but in a fact-intensive manner that paid due consideration to the characteristics of a particular job category (e.g., the degree of risk that mistakes on the job pose to public safety), the important privacy interests at stake, and other context-specific concerns (e.g., evidence of a preexisting drug problem).

The State’s arguments have not convinced us that Skinner and its progeny are inapplicable, nor can they obviate the need for job-category-by-category scrutiny. Just as we know that some subset of state employees almost certainly can be tested due to specific, important safety concerns, we know that there are some employees who almost certainly cannot be tested without individualized suspicion. Again, the problem is that the factual record is almost barren, and the balancing calculus required by Supreme Court case law cannot be exercised in a vacuum. A.

The State’s first justification is that employees have consented to testing by submitting to the testing requirement rather than quitting their jobs, and that this consent renders the Executive Order’s search reasonable and hence constitutional. In effect, the State is offering its employees this Hobson’s choice: either they relinquish their Fourth Amendment rights and produce a urine sample which carries the potential for termination, ■ or they accept termination immediately.

Moreover, rather than treating this exacted consent as part of the special-needs balancing test, the State instead argues that this consent, standing alone, justifies suspicionless drug testing. To begin with, we do not agree that employees’ submission to drug testing, on pain of termination, constitutes consent under governing Supreme Court case law. See Lebron, 710 F.3d at 1214-15. Although a “search conducted pursuant to a valid consent is constitutionally permissible,” Schneckloth v.

Bustamonte, 412 U.S. 218, 222, 93 S.Ct. 2041, 36 L.Ed.2d 854 (1973), consent must be “in fact voluntarily given, and not the result of duress or coercion, express or implied.” Id. at 248, 93 S.Ct. 2041; see also Bumper v. North Carolina, 391 U.S. 543, 548, 88 S.Ct. 1788, 20 L.Ed.2d 797 (1968); Johnson v. United States, 333 U.S. 10, 13, 68 S.Ct. 367, 92 L.Ed. 436 (1948) (consent invalid when “granted in submission to authority rather than as an understanding and intentional waiver of a constitutional right”).

Employees who must submit to a drug test or be fired are hardly acting voluntarily, free of either express or implied duress and coercion. See Bostic v. McClendon, 650 F.Supp. 245, 249 (N.D.Ga.1986); cf. Garrity v. New Jersey, 385 U.S. 493, 497-98, 87 S.Ct. 616, 17 L.Ed.2d 562 (1967) (holding that the government cannot require its employees to relinquish their Fifth Amendment rights on pain of termination because “[t]he option to lose their means of livelihood or to pay the penalty of self-incrimination” was “the antithesis of free choice”).

Moreover, consent has already been adequately incorporated into the special-needs balancing test, which obliges us to evaluate whether an employee’s choice of profession necessarily diminishes her expectation of privacy. In Skinner, the Court weighed the railroad employees’ “participation in an industry that is regulated pervasively to ensure safety,” 489 U.S. at 627, 109 S.Ct. 1402, as a factor militating in favor of drug testing.

In Von Raab, the Court explained that employees’ choice of “certain forms of public employment may diminish privacy expectations even with respect to ... personal searches.” 489 U.S. at 671, 109 S.Ct. 1384. For instance, “[employees of the United States Mint ... should expect to be subject to certain routine personal searches when they leave the workplace every day.” Id. Finally, the Court echoed this view of consent in Vernonia, in which the student athletes and their parents had signed explicit consent forms granting the school the right to test the athletes. See 515 U.S. at 650, 115 S.Ct. 2386.

Nonetheless, the Court did not treat this factor as dispositive. Instead, as the Court saw it, the athletes’ choice to participate was a choice to “voluntarily subject themselves to a degree of regulation even higher than that imposed on students generally,” and amounted to “an additional respect in which school athletes have a reduced expectation of privacy.” Id. at 657, 115 S.Ct. 2386 (emphasis added). Thus, there seems to be no way to square Skinner and its progeny with the argument that consent justifies the Executive Order’s drug testing requirement.

This Court’s recent decision in Lebrón rejected a similar argument that welfare recipients had consented to suspicionless drug testing when the State required testing as a precondition to the receipt of their benefits. As the panel in Lebrón put it, a welfare recipient’s “mandatory ‘consent’ ” was of no “constitutional significance” because it was a “ ‘submission to authority rather than ... an understanding and intentional waiver of a constitutional right.’ ” 710 F.3d at 1214-15 (quoting Johnson, 333 U.S. at 13, 68 S.Ct. 367).

The panel in Lebrón also canvassed the suspicionless drug testing cases and concluded that, to the extent consent was relevant, it had already been incorporated into the balancing calculus. While the context in Lebrón was different because the State sought to test a population of private citizens, which implicates somewhat different privacy concerns, the panel’s logic and reasoning are fairly applicable to these circumstances.

As the panel in Lebrón explained, every time the Supreme Court has addressed á suspicionless drug testing policy—whether those tested were private citizens or government employees—it has analyzed the issue through the prism of SkinnePs special-needs balancing test. See id. at 1215. Surrendering to drug testing in order to remain eligible for a government benefit such as employment or welfare, whatever else it is, is not the type of consent that automatically renders a search reasonable as a matter of law.

Indeed, at least one court of appeals has rejected a similar argument to the one that the State has made here. In McDonell v. Hunter, a case decided even before Skinner and its progeny lent further support to our position, the Eighth Circuit squarely rejected the idea that “employees who signed consent forms have no legitimate expectation of privacy.” See 809 F.2d 1302, 1310 (8th Cir.1987). “If a search is unreasonable, a government employer cannot require that its employees consent to that search as a condition of employment.” Id. (citing Pickering v.

Bd. of Educ., 391 U.S. 563, 568, 88 S.Ct. 1731, 20 L.Ed.2d 811 (1968)); see also United Teachers of New Orleans v. Orleans Parish Sch. Bd., 142 F.3d 853, 856-57 (5th Cir.1998). The courts of appeals have also applied the special-needs balancing test, rather than treating consent as the sole determinant of a policy’s constitutionality, in cases where the government attempted to compel consent to drug testing as a condition for obtaining some privilege. See, e.g., Joy v. Penn-Harris-Madison Sch.

Corp., 212 F.3d 1052, 1055 (7th Cir.2000); id. at 1067 (upholding a policy insofar as it provided for alcohol testing of student drivers but striking it down insofar as it provided for nicotine testing, despite the fact that student drivers signed consent forms authorizing both). In short, the State’s consent argument cannot, standing alone, render the EO constitutional. B.

Next, the State argues, again at a high order of abstraction, that the Executive Order is constitutional under Skinner's special-needs balancing test because the need for a safe and efficient workplace necessarily outweighs state employees’ expectations of privacy. This argument, however, does not entitle the State to

summary judgment. The State’s abstract reasons do not fit within the narrow scope that the Supreme Court has given to the special-needs exception and, therefore, cannot justify testing every category of employee covered by the EO. Indeed, if those reasons could suffice, then there would never be any need to balance anything or consider any job-category-specific rationales. We repeat that individualized suspicion is the normal requirement in this context, and the special-needs eases are only “particularized exceptions to the main rule.” See Chandler, 520 U.S. at 313, 117 S.Ct. 1295.

To the extent the State’s justifications hinge on drug-related productivity loss and other expenses, such as medical care, they are insufficient. Although at oral argument, counsel suggested that the State’s need to maintain an orderly and efficient workplace is enough of a special need to justify suspicionless testing, the authority cited—O’Connor v. Ortega, 480 U.S. 709, 107 S.Ct. 1492, 94 L.Ed.2d 714 (1987)—cannot sustain this proposition.

O’Connor held only that, in the workplace context, the “need for supervision, control, and the efficient operation of the workplace” meant that a workplace search was not subject to the warrant or probable-cause requirements. See id. at 720-26, 107 S.Ct. 1492. O’Connor neither held nor remotely suggested that the need for an efficient workplace could justify searches without individualized suspicion.

The only employment-related rationales that the Supreme Court has endorsed as being sufficient to justify suspicionless drug testing are a “substantial and real risk” to public safety or direct involvement in drug interdiction functions. Chandler, 520 U.S. at 323, 117 S.Ct. 1295; see also Von Raab, 489 U.S. at 670, 109 S.Ct. 1384. Indeed, if safety is the justification, then public safety must be “genuinely in jeopardy,” Chandler, 520 U.S. at 323, 117 S.Ct. 1295; see also Lanier v. City of Woodburn, 518 F.3d 1147, 1151-52 (9th Cir.2008).

Notably, in Chandler, the Court summed up the principle undergirding this line of precedent: [W]here the risk to public safety is substantial and real, blanket suspicionless searches calibrated to the risk may rank as ‘reasonable’—for example, searches now routine at airports and at entrances to courts and other official buildings. But where ... public safety is not genuinely in jeopardy, the Fourth Amendment precludes the suspicionless search, no matter how conveniently arranged. 520 U.S. at 323, 117 S.Ct. 1295 (citation omitted).

The State’s safety argument, at least in its current, global form, is insufficient. The State does not advance specific concerns relating to particular job categories and instead asserts only a broad concern for safety that applies to all employees. But we have little doubt that a clerk, for example, cannot be subject to suspicionless drug testing under the theory that she presents some vague and indefinite safety risk. In comparison, the safety risks that justified suspicionless drug testing regimes in Skinner and its progeny were far more pressing.

In Skinner, railroad accidents had led to 25 deaths, 61 non-fatal injuries, and extensive property loss. See 489 U.S. at 607, 109 S.Ct. 1402. In Von Raab, the concern was with law enforcement officers who carried firearms. See 489 U.S. at 671, 109 S.Ct. 1384.

Here, the State offers the hypothetical examples of an office employee “presenting] a danger when driving a car in the workplace parking lot” or falling prey “to the myriad hazards that exist in the workplace environment (from stacks of heavy boxes, to high stair cases, to files on high shelves, to wet floors, to elevators and escalators).” We reject the idea that a stack of heavy boxes or a wet floor falls within the same ballpark of risk as the operation of a ten-thousand-ton freight train or the danger posed by a person carrying a firearm.

As the Supreme Court did in Chandler, the courts of appeals consistently have rejected testing policies that the government justified based only on generalized and indefinite safety concerns. Those cases underscore that, “where the government asserts ‘special needs’ for intruding on Fourth Amendment rights, ... the specific context matters.” Vilsack, 681 F.3d at 492. “[T]he governmental concern in the general ‘integrity of its workforce’ [i]s insufficiently important to warrant random drug testing .... ” •. Id. at 491-92. Thus, in Vilsack, the D.C.

Circuit rejected a random drug testing policy that covered all Forest Service Job Corps Center employees. I'd. at 499. Similarly, in Lanier, the-Ninth Circuit prohibited the application of a city’s drug-testing policy to a library page. See 518 F.3d at 1152. As the panel in Lanier explained, “the need for suspicionless testing must be far more specific and substantial than the generalized existence of a societal [drug] problem.” Id. at 1150.

Indeed, if the State’s rationale sufficed to justify suspicionless drug testing, then the. exception would swallow the rule and render meaningless Von Raab’s distinction between those employees for whom physical fitness, mental sharpness, and dexterity are paramount and “government employees in general.” 489 U.S. at 672, 109 S.Ct. 1384. Since the State’s generic justifications could apply to all government employees in any context, there would be nothing left of the individualized-suspicion requirement in any type of government employment, and no interests to balance.

Nor does the State shore up its case for across-the-board, suspicionless drug testing with evidence of a preexisting drug problem. Although the State does not need to present evidence of a drug problem in the group it seeks to test, see Von Raab, 489 U.S. at 674-75, 109 S.Ct. 1384, a- showing of an existing problem “would shore up an assertion of special need,” Chandler, 520 U.S. at 319, 117 S.Ct. 1295. The problem with the State’s evidence is that some of it is toó broad to be of any use, and the rest is too specific to justify the breadth of the testing regime the EO mandates.

The bulk of the evidence canvasses the prevalence and harms of drug use in the general population. But Supreme Court case law contemplates a more targeted showing of drug abuse in the group to be tested, not people as a whole. In Skinner, for instance, the Federal Railroad Administration identified a score of drug or alcohol-related train accidents, and industry participants admitted that there was a serious drug problem among railroad workers. 489 U.S. at 607-08, 109 S.Ct. 1402.

The State’s evidence is so general that, if accepted as evidence of a drug problem among state employees, it would have to be accepted in every other government employment context. On the other hand, the relevant data the State presents is too narrow to justify the EO. First of all, the evidence actually suggests that drug use is a relatively small problem in the three departments already subject to random testing prior to the EO’s issuance. The worst result the State obtained was when 2.5 percent of -DOC employees tested positive in 2011. This hardly demonstrates the existence of a serious drug problem.

In fact, as the State itself submitted, a 2010 national survey indicated that 8.4 percent of full-time employees nationwide were illicit-drug users. If anything, then, the results of the State’s random testing reveals that there is substantially less of a drug problem among state employees than among the general working population as a whole. Cf. Lebron, 710 F.3d at 1211 n. 6 (evidence showed that Florida Temporary Assistance for Needy Families recipients tested positive at a 5.1 percent rate, which “was lower than had been reported in other national studies of welfare recipients”).

There is still another problem with the State’s submissions. The data, even assuming it did indicate a drug problem among employees at DOC, DOT, and DJJ, does not demonstrate the prevalence of drug abuse in other state agencies. Thus, even if those results could bolster a case for testing employees at those three agencies—testing which in any event is independently authorized by state statutes not at issue in this case—it would not provide strong support for extending testing to all state employees. In short, the State has fallen far short of showing a preexisting drug problem that pervades its entire workforce.

On the other side of the balancing test, the State also claims that state employees’ expectations of privacy are diminished for two reasons other than consent. First, drug testing among private employers has become common, and this “customary social usage [has] a substantial bearing on Fourth Amendment reasonableness.” Georgia v. Randolph, 547 U.S. 103, 121, 126 S.Ct. 1515, 164 L.Ed.2d 208 (2006). Second, Florida has a tradition of open government that diminishes state employees’ expectations of privacy. We find neither argument persuasive.

The problem with the first one is that it confuses what the Supreme Court means by a diminished expectation of privacy— or, more precisely, what baseline courts should use to determine whether an employee’s expectation of privacy is diminished. The proper baseline is the ordinary government employee’s expectation of privacy.

In Von Raab, for example, the Supreme Court concluded that Customs Service employees involved in drug interdiction had a diminished expectation of privacy precisely because, “[u]nlike most private citizens or government employees in general, employees involved in drug interdiction reasonably should expect effective inquiry into their fitness and probity.” 489 U.S. at 672, 109 S.Ct. 1384 (emphasis added).

In other words, the appropriate inquiry is whether the employee being tested has a diminished expectation of privacy relative to the ordinary government employee because her position depends on physical fitness and judgment. The State’s broad-based argument that all of its employees have a reduced expectation of privacy contradicts binding case law. The second argument is similarly unpersuasive. Open government laws require state employees to disclose certain financial information and also their official work product.

The logical leap from disclosure of financial information and work product to a diminished expectation of privacy in an employee’s physical body is a substantial one. All of the Súpreme Court’s cases discuss the diminished expectation of privacy specifically with regard to physical or bodily privacy, not privacy more broadly conceived.

Thus, in Skinner, employees’ expectations of privacy were “diminished” because of regulations pertaining to their “health and fitness.” See 489 U.S. at 627, 109 S.Ct. 1402; see also Von Raab, 489 U.S. at 672, 109 S.Ct. 1384 (Customs Service employees should have expected inquiries into their “fitness” and “dexterity”). Vemonia is perhaps the clearest example of this focus on physical privacy.

When explaining why athletes have a lower expectation of privacy, the Court pointed out that “[sjchool sports are not for the bashful” and require “ ‘suiting up’ before each practice or event, and showering and changing afterwards” in “locker rooms ... not notable for the privacy they afford.” 515 U.S. at 657, 115 S.Ct. 2386. It is readily apparent, then, that when courts analyze employees’ expectations of privacy in this context, it is their physical privacy that is relevant.

None of the State’s arguments demonstrate that all state employees, including those who have no reasonable relation to safety-sensitive tasks, have a reduced expectation ' of privacy. Just as the State must demonstrate job-category-specific interests, so too must, it demonstrate why each particular job category it seeks to cover under the Executive Order has a diminished expectation of privacy compared to the ordinary government employee. In sum, we cannot find that the State’s proffered rationales warrant

summary judgment in the State’s favor concerning all job categories and all employees covered by the EO. In this case, the character of the intrusion is relatively noninvasive and, “if the ‘special needs’ showing had been made, the State could not be faulted for excessive intrusion.” Chandler, 520 U.S. at 318, 117 S.Ct. 1295. However, the State has failed to make that showing. As the district court concluded, the State’s case most closely resembles Georgia’s failed justification of the policy held unconstitutional in Chandler.

Unlike in Skinner or Von Raab, where the specific job categories subject to testing had a diminished expectation of privacy, the State has failed to demonstrate that all 85,000 state employees somehow have diminished privacy rights. Moreover, it has failed to provide a compelling or important reason for testing; indeed, it has offered only general and weak justifications regarding workplace efficiency and the possible—not “substantial and real,” see Chandler, 520 U.S. at 323, 117 S.Ct. 1295—risks to safety that any state employee may pose. IV.

One final issue has been raised by the parties: who bears the burden in a suspicionless drug testing case. In light of limited authority on this issue, and in order to provide the district court with guidance on remand, we clarify the precise burdens each party bears. There are several different burdens that arise in this case. For starters, on a motion for

summary judgment, “[t]he moving party bears the burden of showing that there are no ... genuine factual issues and that [it] is entitled to

summary judgment as a matter of law.” Gossett v. Du-Ra-Kel Corp., 569 F.2d 869, 872 (5th Cir.1978). Moreover, “in a [42 U.S.C.] § 1983 action, the plaintiff bears the burden of persuasion on every element.” Cuesta v. Sch. Bd. of Miami-Dade Cnty., Fla., 285 F.3d 962, 970 (11th Cir.2002). Thus, in Cuesta, when a § 1983 plaintiff alleged that she was subjected to a strip search without reasonable suspicion, it was “her burden to show that the County lacked reasonable suspicion to search her.” Id. There is no question, therefore, that the Union ultimately bears the burden of persuasion in this case.

In the drug testing context, a plaintiff may initially meet both the burden of going forward and the initial burden of persuasion by demonstrating that (1) there was a search; and (2) it was conducted without individualized suspicion, which ordinarily is

[…]