The Attorney General and the Federal Bureau of Prisons (collectively the Bureau) appeal a permanent injunction of the district court against the Bureau's drug testing program. We reverse in part and remand.

By executive order of September 15, 1986, the President of the United States directed each agency in the Executive Branch to establish a program to test employees in sensitive positions for the use of illegal drugs. Exec. Order No. 12564, 51 Fed.Reg. 32,889 (1986). The Bureau accordingly drew up a program including the regular annual testing of management (positions rated G-13 or above), the testing on reasonable suspicion of all employees, and the random testing of 5-10 percent per year of all the Bureau's 21,000 employees.

In June 1988, the American Federation of Government Employees and certain individual employees (collectively the Union), sought to enjoin the program. On May 12, 1992, the district court issued the following permanent injunction:

1. Random Testing. The Bureau of Prisons may conduct random urinalysis testing only of:

(1) those employees in primary law enforcement positions who, in the regular course of their duties, are issued or given access to firearms for use on a daily or weekly basis;

(2) those licensed physicians and dentists in primary law enforcement positions who, in the regular course of their duties, diagnose, treat, or directly supervise the diagnosis or treatment of patients on a daily or weekly basis;

(3) those employees in primary law enforcement positions who (a) have direct contact with inmates (b) on a daily or weekly basis (c) for periods of one hour or more each day of contact.

2. Post-Accident Testing. The Bureau may conduct urinalysis testing of those employees in primary law enforcement positions who apparently cause accidents or engage in unsafe practices (1) involving personal injury that requires immediate medical treatment, or (2) resulting in more than $2,000 damage.

3. Reasonable Suspicion Testing. The Bureau may test those employees of whom the Court has approved random drug testing, upon reasonable suspicion of on-duty or off-duty drug use or impairment by these employees. All other Bureau employees may be tested only where there is reasonable suspicion of on-duty drug use or impairment. "Reasonable suspicion" to test in either circumstance must be supported by (1) evidence of specific, personal observations concerning job performance, appearance, behavior, speech, or bodily odors of the employee; or, if based on hearsay evidence, (2) corroborative evidence from a manager or supervisor with training and experience in the evaluation of drug-induced impairment.

American Fed. of Gov't Employees, Council 33 v. Barr, 794 F. Supp. 1466, 1479 (N.D.Cal.1992).

The Bureau appealed.

Issues no Longer in Litigation. The Union does not contest that the Bureau may test employees with access to the Witness Security program and the Witness/Victim program of the Bureau. The Bureau does not appeal the injunction so far as it applies to the approximately 7 percent of its employees who are not employed within correctional institutions. The issues on appeal are the random urinalysis of between 5-10 percent each year of all Bureau employees working within correctional institutions, as well as the reasonable-suspicion testing of all Bureau employees.

Established Law. Urinalysis, when required by the government, is a search of the body that is constitutionally prohibited unless it is reasonable. National Treasury Employees Union v. Von Raab, 489 U.S. 656, 665, 109 S. Ct. 1384, 1390, 103 L. Ed. 2d 685 (1989); Skinner v. Railway Labor Executives' Ass'n, 489 U.S. 602, 616-619, 109 S. Ct. 1402, 1412-14, 103 L. Ed. 2d 639 (1989). What is reasonable depends upon the special needs of the government in terms of the government's requirements of those being tested and the reasonable expectations of privacy of those subjected to the test. Von Raab, 489 U.S. at 665-66, 109 S.Ct. at 1390-91; Skinner, 489 U.S. at 619-20, 109 S.Ct. at 1414. A warrant is not necessarily a precondition. Von Raab, 489 U.S. at 665, 109 S.Ct. at 1390. Particular suspicion of an individual is not necessarily a precondition. Id. What the government must show is a reasonable accommodation of the tested employees' privacy expectations to the special needs the government serves by testing.

The Bureau's Needs. The Bureau starts from the fact that more than half of the inmates of federal prisons are there for drug-related crimes. U.S. Dep't of Justice, Bureau of Justice Statistics, Drugs, Crime, and the Justice System 195 (1992). Although not all distributors of drugs are drug users, the familiarity bred from handling drugs may reasonably be expected to have frequently led to the use of drugs.

Federal prisoners are tested on suspicion and randomly for drug use. The percentage of drug tests that proved positive in the year July 1, 1989-June 30, 1990, was not large--1.1 percent for marijuana; 0.4 percent for cocaine; 0.4 percent for heroin; 0.01 percent for methamphetamines. U.S. Dep't of Justice, Bureau of Justice Statistics, Drug Enforcement and Treatment in Prisons, 1990 6 (1992). The percentages are a rough guide, because one sample may reflect more than one drug, and the measures are not exact. Id. at 7. Roughly speaking, the percentages suggest about 1000 federal prisoners using drugs. The percentages are markedly lower than those reported from state prisons where roughly 10 percent have been detected using drugs in a single year. Id. at 6. Interestingly, more drug use has been detected in maximum federal security prisons than in medium or minimum federal security prisons, id. at 8, suggesting that the nature of the prison population is more important than the level of prison security.

Drugs can reach prisoners only by smuggling. The smugglers have to be the prisoners themselves on arrival or return to the prison; visitors; correctional employees; or deliverymen. The prisoners themselves can be searched, as can their visitors. Id. at 3. The employees have substantially greater opportunity to smuggle drugs than do the visitors.

The Bureau has a strong interest in preventing the use of drugs by federal prisoners. The use of drugs can lead to disruptive behavior. The use of drugs interferes with whatever rehabilitative effect a prison may have. The use of drugs within a prison lowers the public's regard of the Bureau and makes the Bureau appear as corrupt or inefficient. The Bureau would be disgraced if a regular federal prison reached the condition reported to exist in Lorton Central Prison administered by the District of Columbia, see Robert Blecker, Haven or Hell? Inside Lorton Central Prison: Experiences of Punishment Justified, 42 Stan.L.Rev. 1149, 1191 (1990) ("drugs are everywhere inside Lorton Central"), or the condition of the New York state prisons as viewed by Adam Walinsky, former chairman of the New York State Commission of Investigation. Personal Responsibility in Criminal Law: Discussion, 77 Cornell L.Rev. 1071, 1075 (1992) ("Drugs are as easily obtainable in prison, or more so, than they are on the street.").

The Bureau believes that there is a nexus between the use of drugs by correctional employees and their smuggling of drugs, because: (1) drug-using employees are subject to blackmail by inmates; (2) drug-using employees need a higher income to sustain their habit; and (3) drug-using employees are likely to be indifferent to the criminal character of drugs.

The Bureau states that all correctional employees have the opportunity at some time or times during their work within the prison to have contact with prisoners. The Bureau adds that, in addition to forestalling the smuggling of drugs, detection of the use of drugs will ensure a desirable level of alertness for all correctional officers; for all are charged with responsibility for security while at work within the prison.

We need not consider here the Bureau's further alleged need, the need for all correctional officers to be drug-free because they all allegedly have access to firearms.

The Union's Position. The Union points to the large number of employees classified as "Correctional" or "Primary Law Enforcement Employees." The two terms are equivalent in this opinion. They include, for example, dieticians, telephone operators, plumbers, and mattress makers. The Union invites the court to be skeptical of the Bureau's position that these miscellaneous positions are occupied by persons who are engaged in primary law enforcement.

The Union challenges the Bureau's need for a program focussed on correctional officers. The Union points to two facts: (1) between 1984 and 1988 only 72 employees out of 21,000 left the Bureau because of connection with the use of drugs; and (2) in testing on reasonable suspicion between 1988 and 1991, 33 tests were conducted but only 11 tests detected illegal drug use.

The Union also points to the features of the urinalysis program that are unpleasant--the sudden call to provide a sample; the insistence on staying until a sample is furnished; the monitoring; and the measures taken to prevent introduction of foreign material into the sample. The Union goes so far as to characterize the procedure as "degrading."

Balance. The Bureau's need for the program has not been established with overwhelming force, given the small number of drug-using employees detected in the nineteen-eighties. It could be supposed that the Bureau has more interest in a display of bureaucratic zeal in response to a presidential directive than in the actual prevention of a pervasive evil. The call here is close. Nonetheless, the Supreme Court has sustained against constitutional challenge the program of the Customs Department where there was even less evidence of an evil to be detected and deterred. Von Raab, 489 U.S. at 669-70, 109 S.Ct. at 1392-93 (citing incidents of bribery and other integrity violations, not drug use, involving Customs officers). In the light of precedent we cannot say the Bureau is constitutionally at fault when it sets its sights on drug-free prisons and a drug-free corps of correctional officers.

The Union points to no evidence in the record countering the Director of Prisons' declaration that all correctional officers are primary law enforcement officers and have the opportunity for contact with prisoners.

No one would want to live in an Orwellian world in which the government assured a drug-free America by randomly testing the urine of all its citizens. The Union intimates that by allowing the correctional officers to be tested we take steps towards this dystopia. The Union overstates the argument. The Bureau's employees are a small fragment of even federal employees. The correctional officers deal with persons in a volatile environment peculiarly susceptible to drugs. By the nature of their work the officers have reduced expectations of privacy. The asserted degrading effect is arguably mitigated by the fact that all of management is subjected to the same random testing.

A balance of the employees' privacy expectations against the special needs of the Bureau sustains the Bureau against a constitutional challenge to its program as it affects correctional officers working within prisons.

Testing on Suspicion. The injunction of the district court did permit nonrandom testing on the basis of reasonable suspicion, but then defined reasonable suspicion so that the suspicion must be supported either by evidence of specific personal observation of the employee's performance, appearance, behavior, speech or odor, or, "if based on hearsay evidence," supported further by "corroborative evidence from a manager or supervisor with training and experience in the evaluation of drug-induced impairment." Barr, 794 F. Supp. at 1479. The second requirement of the district court has no basis in the constitutional protection of the employees from unreasonable search.

The criteria justifying reasonable suspicion used by the Department of Labor have been approved as constitutional by this court. American Fed. of Gov't Employees, Local 2391 v. Martin, 969 F.2d 788, 792-93 (9th Cir.1992). These criteria include:

4. Information provided either by reliable and credible sources or independently corroborated; or

5. Newly discovered evidence that the employee tampered with a previous drug test.

Although reasonable suspicion testing does not require certainty, mere "hunches" are not sufficient to meet this standard.

Id. at 790, n. 1.

That decision, which came two months after the injunction issued by the district court herein, may now be followed by it in setting out the criteria for testing for reasonable suspicion. We have already held that all correctional officers are primary law enforcement employees, so that the Bureau has a special need to test them randomly. By the same reasoning, all are subject to testing for reasonable suspicion of the use of drugs, on duty or off. See id. at 792.

The injunction of the district court must be modified so that it enjoins only the random testing of employees outside correctional institutions who do not have access to information regarding the Witness Security program or the Witness/Victim program and so that it allows testing on suspicion of all employees subject to random testing in accordance with the criteria of Martin.

REVERSED IN PART and REMANDED.