with whom Circuit Judges CUMMINGS and EASTERBROOK join, dissenting:
In dissent, I would rely primarily on my opinion for the panel that originally heard this appeal, Torres v. Wisconsin Department, 838 F.2d 944 (7th Cir.1988). After parsing the district court’s opinion to find errors I think more fanciful than compelling, the majority has sent this matter back. The district court may take additional evidence and then apply a new standard, the outlines of which are quite indistinct.
As a primary point, I share Judge East-erbrook’s view, contained in his persuasive dissenting opinion, that the employer must establish the necessity of a BFOQ in order to avoid Title VII’s strictures against employment discrimination. For this case that rule means that the burden of bridging the yawning gaps in our understanding of rehabilitation falls on the state, not the plaintiffs.
To meet its burden here, the state has relied on what could be characterized as a “feminist” theory of rehabilitation. It is argued that men are at the root of many of these female inmates’ difficulties in social adjustment. The presence of men allegedly creates tension and stress in the living areas and men must therefore give up their jobs in those areas. This theory, propounded by the prison warden and supported by the testimony of two other witnesses, is fine in the abstract. It is logically coherent and its recognition of the problems men can cause women in our society is certainly praiseworthy. But the district judge found that the state’s support for this theory, as applied to the rehabilitation of female prison inmates, was woefully inadequate to meet the demanding BFOQ standard. Under Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977), to justify discriminatory employment practices employers, public no less than private, must show that these practices are “necessary to safe and efficient job performance.” Id. at 332 n. 14, 97 S.Ct. at 2728 n. 14. The record here contained, in opposition to the state’s position, testimony from plaintiffs’ expert Dr. Frederick Fosdale, who opined that the presence of males in the living quarters was valuable to inmates’ eventual ability to adjust to the outside world. Tr. 6 at 14-15. Dr. David Kalinich, another expert for the plaintiffs, testified that TCI’s BFOQ plan had not enhanced any rehabilitative efforts. Tr. 7 at 6. The plaintiffs also called Robert E. Ellsworth, a former Director of Wisconsin’s Bureau of Adult Institutions, who testified that employing staff of both genders can have a beneficial effect on the treatment of inmates, provided privacy concerns are met. Tr. 6 at 64. The record also contained one meager bit of quantitative evidence: Superintendent Switala’s concession that TCI’s recidivism rate had not declined since the gender-based restrictions had been imposed. Tr. 10 at 177-82; see 639 F.Supp. 271, 280 (E.D.Wis.1986). Concededly, the period over which potential improvements were measured was short and trends in recidivism rates may not reflect all of an institution’s improvements in “enhancing the individual’s potential for personal achievement.” Maj. op. at 1532 n. 4. But the burden here, a heavy one, was on the state to show that gender discrimination was necessary, not on the guards to show that gender neutrality was harmless.
Also significant, in my view, is what the record did not contain. If the presence of men in the living units was as stressful and disruptive as the state claims, one would think that specific, concrete evidence of negative inmate reactions would have been available. There was none of this. The only evidence presented in this vein was Superintendent Switala’s general description of complaints received from inmates. *1534Tr. 9 at 83-85. This gaping void in the evidence is perhaps what Chief Judge Warren had in mind when he commented upon the state’s lack of “objective evidence.”
The majority finds conclusive evidence of legal error in the district court’s statement that the defendants “offered no objective evidence, either from empirical studies or otherwise” in support of their rehabilitation theory. 639 F.Supp. at 280. According to the majority, this statement demonstrates that Judge Warren’s legal standard was “unrealistic” and “unfair.” Maj. op at 1532. I think the majority makes far too much of an isolated comment in a detailed and thoughtful opinion. Perhaps the district judge should have stressed his recognition that expert testimony is competent in this kind of case. I do not believe that Judge Warren’s language, which is juxtaposed with a summary of the contradictory expert testimony of the defendants’ and plaintiffs’ experts, even remotely establishes that the district judge carried the skepticism required in assessments of BFOQ claims to an extreme.
The majority claims that the plaintiffs’ evidence does not prove to its satisfaction that Superintendent Switala’s decision to establish the BFOQ was not “the product of a reasoned decision-making process, based on available information and experience.” Maj. op. at 1532. This approach to formulating a standard cannot, of course, be reconciled with Dothard. Further, the majority’s assumption that there are decisive differences in the rehabilitation of men and women convicts runs totally counter to Title VIPs mandate to view such assumptions with suspicion. Although it eschews any bold statement, the majority opinion amounts to a prison exception to the generally applicable BFOQ standard. We are told that prison officials, particularly prison officials who answer to legislatures that believe, at least in theory, in attempting to rehabilitate inmates, face “intractable problems” which they must address under conditions of radical uncertainty. To have any hope of accomplishing their daunting task these prison administrators must have the latitude to innovate in ways that violate otherwise applicable strictures against employment discrimination. Courts must therefore recognize this experimental license as a necessary element of prison administration in BFOQ cases.
To confine this special treatment to prisons, however, we would have to rely more on edict than on principle. Society asks many public institutions to attack seemingly intractable problems in areas beset by uncertainty. The jobs of maintaining law and order and providing adequate health care and effective education, particularly in our poorest neighborhoods, come immediately to mind. In constitutional cases, one source of the majority’s special deference to prison administrators, the Supreme Court has recognized that the special demands placed on public school administrators, like the demands placed on prison administrators, necessitate special leeway. See, e.g., Hazelwood School Dist. v. Kuhlmeier, — U.S. -, 108 S.Ct. 562, 98 L.Ed.2d 592 (1988); New Jersey v. T.L.O., 469 U.S. 325, 105 S.Ct. 733, 83 L.Ed.2d 720 (1985). I doubt that the majority would be prepared to permit school administrators to adopt BFOQs involving race, gender, religion or age based solely on a showing that the discriminatory measures were “innovative” and adopted in good faith in the face of intractable educational problems and pedagogic uncertainty. I do not doubt, however, that reversal in this case has far more to do with the majority’s disagreement with Judge Warren’s weighing of the evidence than with any particularly compelling need for prison administrators to innovate in ways that violate strictures against employment discrimination.
Even with its newly crafted experimental license doctrine, the majority’s determination to reverse necessitates some energetic massaging of the plaintiffs’ evidence. Apart from the recidivism numbers, which the majority feels safe in disregarding due to the short time period they cover and the presumed shortcomings of recidivism as a measure of rehabilitation, there is the testimony of Fosdale and Kalinich. The majority finds this evidence wanting because Ka-linich’s work had primarily involved male prisons — though he also had done consult*1535ing and administrative work involving female prisons, see tr. 6 at 140-41 — and because both expressed skepticism about the prospects for rehabilitation in general. Ordinarily, we leave assessments of the persuasiveness of expert testimony to district courts. Anderson v. City of Bessemer City, 470 U.S. 564, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985), requires us to apply the deferential clearly erroneous standard of review to district court factual determinations — even to factual determinations based on physical or documentary evidence, which would appear to present a weaker case for deference than the assessment of testimony at issue here. Id. at 573-76, 105 S.Ct. at 1511-13. Moreover, this circuit has long held that assessments of expert testimony are no different from other factual determinations in this respect. See, e.g., George R. Churchill Co. v. American Buff Co., 365 F.2d 129, 130 (7th Cir.1966). I would have thought that such matters as determining the boundaries and transferability of Kalinich’s expertise (a point on which the district judge expressly ruled, see tr. 6 at 143) and deciding how Fosdale’s and Kalinich’s general doubts about the possibilities of rehabilitation affected their credibility on the question at hand were part of the fact-finder’s function.
Finally, the majority suggests that a “common-sense understanding of penal conditions,” as well as deference to prison administrators, supports its rejection of the trial judge’s evaluation of the evidence. Maj. op. at 1531. The conventional view is that common sense in a particular field, though perhaps partly innate, develops through exposure to events and experiences. Ordinarily, we assume that the comparative advantage on matters of common sense rests with district judges. In this particular instance, the rule of thumb seems especially apt. Judge Warren has served as district attorney of one of Wisconsin’s larger urban counties and as Attorney General of the state for a number of years. He has certainly had considerable exposure to pe-nological practice and to penological theory. His experience has endowed him with an uncommon amount of common sense. I do not believe it is necessary to seek new evidence and pursue a new analysis of the matters he has ably addressed.
I therefore respectfully dissent.