This is аn interlocutory appeal and cross-appeal from a pretrial order of a district court granting in part and denying in part plaintiffs’ petition for temporary injunctive relief in a case alleging racial discrimination in the process of hiring defendant City’s firemen and policemen.
We affirm the district court’s order and remand the case for further proceedings.
I.
On March 21, 1986, plaintiffs-appellants N.A.A.C.P. and Walter Harris (“plaintiffs”) filed a class action against defendant-ap-pellee/cross-appellant City of Mansfield, Ohio, and various city officials (“defendants)”.
Both parties have appealed the court’s decision. The central issues on appeal are (1) whether the district court’s subsidiary factual determinations are clearly erroneous, and (2) whether the court abused its discretion by issuing the partial preliminary injunction.
We hold that the district court’s findings of fact are not clearly erroneous and that the court did not abuse its discretion by not enjoining the City of Mansfield from filling vacancies in its police department. However, with respect to the injunction relating to firefighter positions, we conclude that a remand to the trial court is necessary because subsequent events have cast doubt upon the magnitude of the potential “harm to others” and the “countervailing public interest” that originally militated against a broader injunction.
II.
Plaintiffs claim the testing procedures employed by the City of Mansfield in hiring policemen and firemen are a vehicle for discrimination. In particular, they challenge the written tests that were administered by the City’s Civil Service Commission (CSC) and given to police and fire department applicants on March 22, 1986, the agility tests subsequently given to firefighter applicants who successfully completed the written test, the scoring of all three tests, and the lists of policemen and firefighter applicants who were certified by the CSC as being eligible for available positions.
The City of Mansfield is located in Rich-land County in north central Ohio. According to the 1980 census, Mansfield has a total population of 53,927, and a black population of 8,653 (sixteen percеnt). Richland County has a total population of 131,205, and a black population of 9,373 (seven percent). Thus, ninety-two percent of the blacks in Richland County live in the City of Mansfield. Statewide, blacks comprise 6.1% of the civilian labor force, and 6.1% of the aggregate police and fire positions.
A.
The police department eligibility test is a written, multiple-choice examination consisting of 200 questions. It was composed by the Center for Criminal Justice (CSC) at Case Western Reserve University. Applicants who answer seventy percent of the questions correctly pass and may there
B.
The fire department’s test is somewhat different. While it includes a written, multiple-choice examination comprised of 200 questions composed, administered, and scored by the CSC, it also includes an agility test administered by fire department officials. Only those applicants who earn a score of seventy percent on the written test are eligible to take the agility test. Sixty percent of an applicant’s overall score is based on his score on the written test, and forty percent is based on his agility test score.
The agility test consists of nine events that simulate the kinds of physical exertions a firefighter might encounter in the course of his duties.
The department forwards the agility test results to the CSC for totaling. An applicant can earn an additional number of points equal to twenty percent of his combined score upon proof of military service. If there is an opening in the fire department, the CSC certifies the names of the top three candidates from the eligibility list for further review, including a personal interview.
C.
The CSC administered both the police and firefighter exams on March 22, 1986. A total of one hundred fifteen persons, fourteen blacks and one hundred one whites, took the police examination. Seven blacks and seventy-six whites passed the test. A total of ninety-six persons took the firefighter exam, ten blacks and eighty-six whites. One black and thirty-eight whites passed the test. Plaintiff Harris, the only black to pass the firefighter written exam, subsequently received the lowest score of all the firefighter applicants who performed the agility test. The CSC thereafter drew up the respective departmental eligibility lists.
At the time of the hearing on plaintiffs’ motion for a preliminary injunction, the Mansfield police department had eighty-four authorized positions, eighty-two of which were funded. Of these eighty-two positions, seventy-seven were manned, five by blacks and seventy-two by whites. Thus, there were five available police positions. The fire department had ninety-two authorized positions, ninety of which were funded. A total of eighty-seven of these positions were manned by eighty-six whites and one black. Thus, there were three available firefighter positions. Apparently, if all the vacancies in both departments
The City indicated a desire to fill all eight positions as soon as possible by certifying applicants from the respective eligibility lists, pursuant to the prevailing CSC procedures. Plaintiffs request for a preliminary injunction sought to enjoin the City, the police department, and the fire department from certifying any names from the current lists and from appointing any individuals to the police or fire departments.
In an extensive written opinion, the district court considered plaintiffs’ claims for injunctive relief against the police and fire departments separately. The court reasoned that the rationale for the need each department had for filling its respective positions were separate and unrelated, and that the departments’ testing procedures were distinguishable. After separately applying this court’s four-factor analysis for determining whether to grant a preliminary injunction, see, e.g., Mason County Medical Ass’n. v. Knebel,
III.
Plaintiffs argue on appeal that the district court erred by refusing to enjoin police department hiring from the March 1986 eligibility lists and by incompletely enjoining similar hiring practices by the fire department. Defendants argue that the court abused its discretion by limiting the fire department to filling just three firefighter positions.
It is well-settled that district courts evaluating requests for preliminary injunctive relief must consider four interrelated criteria:
1) Whether the plaintiffs have shown a strong or substantial likelihood or probability of success on the merits;
2) Whether the plaintiffs have shown irreparable injury;
3) Whether the issuance of a preliminary injunction would cause substantial harm to others;
4) Whether the public interest would be served by issuing a preliminary injunction.
Mason County,
Application of the abuse of discretion standard can be summarized as follows:
“In reviewing the decision of a district court to grant or deny a preliminary injunction, this court has continued to invoke the phrase ‘abuse of discretion’ in articulating the applicable standard.” In deciding whether to afford injunctive relief, the district court must first make findings of fact and conclusions of law, and then exercise its discretion to grant or deny the injunction. This court reviews the findings of fact under the clearly erroneous standard; legal conclusions are given de novo rеview. “[A] factual or legal error may alone be sufficient to establish that the court ‘abused its discretion’ in making its final determination.... However, in the absence of such an error the district judge’s weighing and balancing of the equities should be disturbed on appeal only in the rarest of cases.”
Baja Contractors, Inc. v. City of Chicago,
Under the Mason County criteria, plaintiffs’ initial burden in demonstrating entitlement to preliminary injunctive relief in this case was a showing of “a strong or substantial likelihood or probability of success on the merits” of their class action under 42 U.S.C. § 1981. Mason County,
Once plaintiffs produced substantive evidence of intentional discrimination on their § 1981 claims, the court was required to balance the apparent strength of that evidence against the irreparable harm plaintiffs would suffer in the absence of injunc-tive relief, the harm which would be caused to others by the issuance of a preliminary injunction, and whether the public interest would be served by an injunction. Mason County,
<
>
Plaintiffs motion to enjoin appointments to the Mansfield police department relied heavily upon the testimony of Dr. Douglas Schultz, an expert witness in “tests and measurements.” Dr. Schultz testified about several statistical indicators which, in his opinion, show that the department’s testing procedures have a “statistically significant impact.”
Dr. Schultz employed a “chi square” analysis to determine whether the differing pass ratios on the written examination for white and black applicants demonstrated a statistically significant disparate impact on black applicants. A chi-square value is a test of association which measures deviations from expected behavior. Certain deviations are expected to oсcur as a pattern of chance. However, at some point a discrepancy becomes so large that it is no longer expected to occur as a result of chance alone. The chi-square value for any two series is determined from a standard statistical table. Where a discrepancy becomes larger than that number, it means that the differences have not occurred as a result of chance alone. However, if the chi-square value becomes smaller than that number, the change is probably a result of chance variations. Based on the police department examination pass rate for white applicants of seventy-five percent (seventy-six out of one hundred one applicants) and the pass rate for black applicants of fifty percent (seven out of fourteen applicants), Dr. Schultz determined the applicable chi-square value to be 3.90. He testified that because that number exceeded 3.84, the chi-square value suggested statistical significance warranting the conclusion that the differing examination pass ratios for whites and blacks on the police test were not attributable to chance. The court found that Dr. Schultz’s chi-square analysis
Dr. Schultz also testified that the results of the рolice department’s test revealed an adverse impact on blacks under the “four-fifths rule” used by the Equal Employment Opportunity Commission (EEOC). See 29 C.P.R. § 1607.4(D). Under the EEOC’s rule:
A selection rate for any race ... which is less than four-fifths (Vs) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact....
The ratio of the black pass rate of fifty percent to the white pass rate of seventy-five percent is sixty-seven percent, less than the eighty percent cutoff used by the EEOC.
Plaintiffs also offered evidence of the statistical relationship of past police department hiring prаctices to race.
Referring to the concept of the “standard deviation” as a measure of the significance of statistical disparities in a Title VII Civil Rights Act minority school teacher hiring case, the Supreme Court said:
“[I]f the difference between the expected value and the observed number is grеater than two or three standard deviations,” then the hypothesis that [the] teachers were hired without regard to race would be suspect.
Hazelwood School Dist. v. United States,
The plaintiffs argue that the district court “abused its discretion” in concluding that the plaintiffs have shown no more than a “marginally statistically significant disproportionate impact on blacks in the police department.” But “abuse of discretion” is not the standard of review for the trial court’s factual findings. We are not free to disturb the district court’s factual finding of “marginally” disproportionate impact, unless we are able to say that the finding is clearly erroneous. See Baja Contractors, Inc. v. City of Chicago,
Thus, if an employer selects 60% of the blacks and 80% of the whites, this very likely indicates a real difference in selection procedures if he is choosing 600blacks out of 1,000 and 800 whites out of 1,000. On the other hand, if he chooses 3 blacks out of 5 applicants and 4 whites out of 5 applicants, both common sense and rigorous statistical analysis tell us that it is much more likely that mere chance is the controlling factor. Thus, the small size of the sample might be a reason to rule that there was no prima facie case even if the % rule were violated....
Only fourteen of the one hundred fifteen March 1986 police test applicants were black. The EEOC’s rule itself states in part:
Greater differences in selection rates may not constitute adverse impact where the differences are based on small numbers and are not statistically significant. ...
29 C.F.R. § 1607.4(D) (emphasis added).
In sum, the district court’s factual finding cannot be said to be clearly erroneous. “Where there are two permissible views of the evidence, the factfinder’s choice between them cannot be clearly erroneous.” Anderson v. City of Bessemer City, North Carolina,
Plaintiffs also sought to prove discriminatory intent through circumstantial evidеnce. Most significantly, plaintiffs showed that seven more names appeared on the police eligibility list than the number of applicants who passed the written test, an apparent “[departure from the normal procedural sequence.” Village of Arlington Heights,
Plaintiffs also showed that Anna Coe, a black member of Mansfield’s Human Relations Commission, took both the police and fire department examination. Ms. Coe received a score of eighty-seven out of a possible one hundred on the police examination, but received no extra credit for her education experience. However, as the district court noted, “it was not clear whether Coe’s decision to take the two tests was based on her desire to monitor the testing process as a member of the Human Relations Commission, or on a genuine desire to become a member of either the police or fire department.” Further, Coe made no attempt to rectify the scoring error after learning that she had not received her extra credit. The district court held that “Coe’s testimony is inconclusive at best with respect to establishing circumstantial evidence of discrimination in the hiring of police officers.”
Weighing plaintiffs’ circumstantial evidence together with plaintiffs’ “marginally statistically significant evidence” of a discriminatory impact, the district court concluded, under the analysis of Mason County, that plaintiffs had failed to demonstrate a likelihood of success on the merits with respect to the hiring practices employed by the police department. This part of the district court’s holding is a legal conclusion subject to de novo review. See Baja Contractors, Inc.,
Plaintiffs argue that the court misapplied the legal standard used to determine whether a claim is sufficiently strong to merit relief in the form of a preliminary injunction. In that regard, this court recently held:
[A] stay may be granted with either a high probability of success and some injury or vice versa. However, we reiterate that the demonstration of a mere “possibility” of success on the merits is not sufficient, and renders the test meaningless.
State of Ohio ex rel. Celebrezze v. NRC,
It is obvious that if plaintiffs are able, eventually, to prevail on the merits of their claim, they will have suffered some actual harm in the absence of injunctive relief. This inevitability alone, however, is insufficient to show that the court below abused its discretion. Since plaintiffs have failed to demonstrate a high probability of success, they must compensate by showing a correspondingly high degree of injury. However, the potential injury they have shown, while conceivably direct and probable, is not likely given their otherwise marginal case. Further, it is undisputed that plaintiffs did not file their motion for in-junctive relief until nearly eleven months after they had filed the underlying suit, thus mitigating against the immediacy and irreparability of the injury perceived by plaintiffs. The district court carefully weighed these considerations. We think it did not abuse its discretion in denying preliminary relief.
B.
Ten percent, one out of ten, of the black applicants who wrote the Mаrch 1986 firefighter examination passed as did forty-four percent, thirty-eight out of eighty-six, of the white applicants. With respect to the fire department test, Dr. Schultz testified that the chi-square value for the pass rates was 4.34. He concluded that the difference in the pass rates for the blacks and whites was not attributable to chance because a chi-square in excess of 3.84 suggests statistical significance. Under the EEOC’s “four-fifth’s rule,” the ratio between the black pass rate and the white pass rate equals 22.7%, considerably less than the eighty percent standard.
In addition to his assessment of the test results, Dr. Schultz analyzed the 1986 racial composition of the fire department as cоmpared to the racial composition of the City of Mansfield. He determined that 1.14%, one out of eighty-seven, of Mansfield firefighters were black, and that according to 1980 census figures, sixteen percent of Mansfield’s citizens are black. Therefore, if black employment in the fire department in 1986 had been equal to the percentage of blacks in the City of Mansfield, there would have been fourteen black firemen rather than one. Applying a standard deviation analysis to these figures shows a statistically significant relationship between the representation of blacks in the fire department and their race. An expected number of fourteen firefighters minus three standard deviations results in a minimum expected representation of 3.8 blacks in the fire department. Therefore, since there is only one black firefighter, the defendants’ claim that fire department applicants have been selected without regard to race is suspect. See Hazelwood School Dist.,
In their cross-appeal, defendants challenge the court’s findings as clearly erroneous. They contend that the department’s employment figures are statistically insignificant when they are compared to the state and county population. They argue here, as they did below, that the black population of Richland County and not the City of Mansfield is the relevant criterion for measuring the proportionality of black representation in the Mansfield fire department because Richland County is the “civilian labor force” from which Mansfield
But proof of a disparate racial impact in the composition of the Mansfield fire department alone is insufficient to support a § 1981 claim. Intentional discrimination must be proved. Black v. City of Akron, Ohio,
The district court concluded that the plaintiffs presented sufficient circumstantial evidence to establish a likelihood of success on their § 1981 claim and we are satisfied that conclusion is well-supported by the evidence, given the underrepresentation of blacks in the fire department and the statistically significant race/test score relationship.
Plaintiffs asked the district court to enjoin defendants from filling any of the open positions remaining in the fire department. Despite its conclusion that plaintiffs had established a likelihood of success on their underlying claim of intentional discrimination, the court chose not to completely enjoin the fire department from making new appointments. Instead, the court granted injunctive relief restraining defendants from hiring more than three firefighters from its 1986 eligibility list.
The court did not deny the possibility that plaintiffs might be irreparably harmed in the absence of greater injunctive relief.
Chief Kopcial testified that the manning in the fire department currently is the lowest in the history of the fire department. He indicated that the decreased number of fire fighters affects the fire department’s ability to provide fire safety services to the city. He indicated that the understaffing decreases the team work and cohesion among individual engine units, because the decreased staffing requires the department to fill available positions by allowing firemen to work overtime and by transferring firemen from one station to another. Chief Kopcial also testified that the decreased staffing results in the, manning of positions within units by a person whomight not have the requisite experience required to perform adequately on a particular type of equipment. Chief Kopcial testified that these problems are lowering the quality of service by the fire department to the citizens of the City of Mansfield. Thus, the Cоurt finds public interest in not issuing the injunction,
and
The Court ... finds that the ability of the current City of Mansfield Fire Department to provide protection to the City of Mansfield against fires and other emergencies is severely curtailed by the current understaffing of the fire department. Thus, although the plaintiffs have demonstrated a likelihood of success and irreparable harm, the defendants have shown that ... harm to others and a countervailing of public interest weigh against the issuance of the injunction.
Although the district court acted within its discretion in weighing the competing factors underlying its decision to grant plaintiffs’ only limited injunctive relief, events subsequent to the trial court’s hearing indicate that the court may have given undue weight to factors militating against a broader injunction. Despite its vigorous assertion before the district court that issuance of the requested injunction would seriously jeopardize the public safety and the safety of its firefighters by leaving the fire department understaffed, the defendant City of Mansfield has not, in the eighteen months following entry of the district court’s order, filled any of the three vacant firefighter positions despite its ability to do so under the terms of the court’s order. This suggests that the defendants may not have fully and completely appraised the district court of the interests which underlie their opposition to plaintiffs’ requested injunctive relief. Accordingly, we remand the case to the district court to afford the court an opportunity to reconsider plaintiffs’ fire department-related motion in light of the subsequent developments.
V.
For the foregoing reasons, we AFFIRM the district court’s order denying plaintiffs’ motion to enjoin Mansfield’s police department. We AFFIRM the district court’s order to partially enjoin Mansfield’s fire department and REMAND the case for reconsideration in light of the factors previously discussed.
Notes
. In addition to the City, plaintiffs named as defendants Edward Meehan, Mayor; Pearl Adams, Personnel Director; William Friend, Service Safety Director; Matthew Benick, Chief
. The court also found, with respect to police department hiring, “that the plaintiffs will not suffer substantial irreparable harm by the issuance of the injunction” and "that there is a strong public interest in ensuring that the police department has adequate manpower to provide protection to the citizens of Mansfield.”
. The district court described the agility test as follows:
The agility test consists of two portions. The first portion consists of two timed events. Each applicant must climb a 50-foot ladder in five minutes, and each applicant must perform a fence and wall climb, during which the applicant runs 45 feet with a rope, tosses the rope over a 5-foot high fence, and then vault the fence. If the applicant successfully completes these two events, the applicant then performs the second portion of the test. The second portion consists of seven additional events. The applicant must perform a ladder lift, charged hose pull, balance beam walk, tower climb, hose lift, and hose stack.
. Even though the statute of limitations may have expired on claims arising out of the department's past hiring decisions, evidence of those decisions is admissible if relevant. See Black Law Enforcement Officers Ass’n. v. City of Akron, Ohio,
. The district court seemed unimpressed by the defendants insistence that 1) Dr. Schultz's chi-square analysis was incorrect becаuse it did not include a so-called “Yates correction” for chi-square numbers under 5, and 2) the relevant population group for police department applicants was Richland County and not the City of Mansfield.
. The court recognized that should plaintiffs eventually prevail, the positions within the department might no longer be available. Further, even if new positions were created in the intervening period, plaintiffs would not have the seniority they might have earned had the injunction been issued. See, e.g., Firefighters Inst. for Racial Equality v. City of St. Louis,