dissenting.
I would affirm the judgment of the district court. The case is now reduced to a damage suit against two individuals, Anderson and Keve, resting on two separate incidents, Anderson’s rejection of Winsett’s application for work release in 1974, and Keve’s veto in 1976. Although the liability theory underlying each is the same, it is desirable to treat the claims against the defendants separately. I believe the record requires a conclusion that Keve is entitled to official immunity and that the plaintiff has not established a claim against Anderson.
A court should not decide a constitutional question if the record supports some other ground that would obviate the need to make such a determination. Ashwander v. Tennessee Valley Authority, 297 U.S. 288, 347, 56 S.Ct. 466, 483, 80 L.Ed. 688 (1936) (Brandeis, J., concurring). In my view, it is consistent with this policy to decide a case, when possible, on the basis of immunity, and I therefore apply that rationale to Keve’s case. As to Anderson and the 1974 incident, however, the findings of the district court are not sufficient to support a decision on immunity at this point. Thus, it is necessary to address the constitutional issue in his case. Although this prevents complete avoidance of the constitutional question, deciding Keve’s immunity confines the liberty interest inquiry to Anderson’s case.
I.
I would not remand the case to the district court for specific findings on the immunity of defendant Keve. I agree that the guidelines discussed by the court, particularly the two-step analysis set out in Procunier v. Navarette, 434 U.S. 555, 98 S.Ct. 855, 55 L.Ed.2d 24 (1978), are the considerations that apply. In my view, however, the record is adequate to make the determination in this court. For example, in Reese v. Nelson, 598 F.2d 822 (3d Cir. 1979), we analyzed the record in a case submitted to a jury and found that the evidence required a finding of immunity as a matter of law. Here, by contrast, the task of evaluating the immunity issue is less complex because we have the benefit of the district judge’s findings and opinion.
The first step is to decide whether a clearly established constitutional right should have been apparent to Keve in 1976. That assumes, of course, that such a constitutional right exists, a premise I accept only arguendo. The majority opinion does not cite any cases handed down before 1976 *1011even arguably close to the circumstances present here, nor has my research uncovered any such authority. No Supreme Court case to that date had discussed the subject of conditional release eligibility, nor were there any opinions of this court holding that such an interest was protectible. Accordingly, there is nothing to demonstrate that the interest that Keve arguably violated was “clearly established at the time of [the] challenged conduct.” Procunier v. Navarette, supra at 562, 98 S.Ct. at 860. Because he could not reasonably have been expected to be aware of a constitutional right that had not yet been judicially declared, Keve did not act in disregard of established law.
The second inquiry of the immunity analysis is subjective — whether the defendant acted with malicious intent to cause a deprivation of constitutional rights or other injury to the plaintiff. Stated inversely, “The official himself must be acting sincerely and with a belief that he is doing right . . .” Wood v. Strickland, 420 U.S. 308, 321, 95 S.Ct. 992, 1000, 43 L.Ed.2d 214 (1975).
• The trial court credited Keve’s testimony that because of the likely public and legislative reaction to the plaintiff’s work release status, Winsett faced far more hazards than other inmates. Under these circumstances, it was more feasible to secure parole or work release after a transfer to another state, and Keve was negotiating such a transfer at the time he disapproved Win-sett’s application. The court said, “In the case of the second denial, I am convinced Acting Commissioner Keve believed that he had an alternate approach to work release which would better serve Winsett’s interests.” Winsett v. McGinnes, 443 F.Supp. 1369, 1371 (D. Del.1978). This finding belies any inference of malice on the part of defendant Keve and satisfies the subjective test of sincerity and a “belief that he is doing right.” Having met both phases of the immunity test, Keve is entitled to the judgment entered in his favor by the district court.
II.
A.
Since Anderson’s conduct occurred two years before Keve’s, it is even more apparent that the former meets the requirements under the first phase of the immunity inquiry. But I cannot complete the second step in the analysis because the district court did not make any findings on Anderson’s subjective intent, and the record does not support appellate resolution of this question.
Anderson did testify that he declined to approve the work release application because he feared for Winsett’s safety. This apprehension was the result of at least one telephone call to the prison threatening harm to Winsett if he were put on work release. Anderson attributed the threats to public reaction fueled by press reports, letters to the editor, and the opposition of the state police. The defendant bore Winsett no animosity and said that some years earlier he had offered use of his home to Win-sett as a place to visit his son on Christmas Day. Indeed, in 1976, Anderson appeared in Winsett’s behalf before the Parole Board to urge an out-of-state parole. The defendant testified that his job was protected by the merit system and denied that he succumbed to pressure from the state senator in Winsett’s case.
Although Anderson’s testimony would support a finding that he acted in subjective good faith, I do not have the benefit of the trial judge’s credibility determinations and, consequently, cannot pass upon the defendant’s immunity at this point.1 I therefore turn to the issue of the alleged constitutional violation.
B.
I start with the assumption that a discretionary work release program and discre*1012tionary parole system basically require the same analysis to determine if there is a protectible interest that invokes due process standards. The majority’s reference to Greenholtz v. Inmates of the Nebraska Penal Correctional Complex, 442 U.S. 1, 99 S.Ct. 2100, 60 L.Ed.2d 668 (1979), leads me to believe that there is no disagreement between us on that score. We part company on whether the circumstances warrant finding that Winsett had a state-created liberty interest in an expectancy of work release status. In my view, the majority’s conclusion is contrary to the Court’s holding in Greenholtz.
In that case the Supreme Court addressed the question whether due process applies to discretionary parole-release determinations. Initially the opinion noted that decisions of the executive branch, especially with respect to the “sensitive choices” to grant parole, “however serious their impact, do not automatically invoke due process protection.” 442 U.S. at 7, 99 S.Ct. at 2104. The “public interest purposes of rehabilitation and deterrence” may prompt the state to be either general or specific in enumerating the factors that the parole authorities should consider. “It is thus not surprising that there is no prescribed or defined combination of facts which, if shown, would mandate release on parole. ... In each case, the decision differs from the traditional mold of judicial decision making” in that the parole determination is a “predictive judgment as to what is best for the individual inmate and for the community.” Id. at 8, 99 S.Ct. at 2104. In that connection, the Court found that in passing on parole the decisionmaker must consider whether, in light of the nature of the crime, the inmate’s release would minimize the gravity of the offense, weaken deterrent impact on others, and “undermine respect for the administration of justice.” Id.
Two theories were proffered in support of the claim to due process protection of a prisoner’s expectancy of release on parole: first, that the word “liberty” in the due process clause included a prisoner’s interest in the possibility of parole whenever a state establishes a parole system; second, that the particular statutory language involved in the case gave rise to a state-created liberty interest in release on parole. The Court rejected the former but accepted the latter. It found a liberty interest created by a state statute directing that the Parole Board “shall order” a prisoner’s release “unless” it found certain enumerated conditions to justify a denial.
The Court emphasized the essentially discretionary nature of parole but held that the “unique structure and language” of the statute under review created an expectancy of release that satisfied the claim to constitutional protection. As the majority recognizes, however, the Court stressed that “whether any other state statute provides a protectible entitlement must be decided on a case-by-case basis.” Id. at 12, 99 S.Ct. at 2106. Thus, discretion in the granting of parole is the general rule and generates no protectible interest. Only when discretion is restricted by statutory language that mandates the parole authorities to grant release does a liberty interest come into being. It is the statute’s command, not discretion, that creates the interest.
This, however, is not the test adopted by the majority. Instead the formula used is “that a state-created liberty interest in work release arises when a prisoner meets all eligibility requirements under the state regulation and the exercise of the prison authorities’ discretion is consistent with work release policy." Majority Opinion at 1007 (emphasis added). As I read it, Greenholtz does not recognize a protectible interest if the “discretion” is merely “consistent with” a state statute. The dissent in that case calls attention to the majority’s use of the word “shall,” 442 U.S. at 29 n.9, 99 S.Ct. at 2115 n.9, and as the Court of Appeals for the Second Circuit remarked in Boothe v. Hammock, 605 F.2d 661 (2nd Cir. 1979), “The ‘shall/unless’ formula was decisive for the Court.” Id. at 664. The Second Circuit found that in New York, while guidelines were “used to structure the exercise of discretion,” the state system did not establish a scheme whereby parole “shall be ordered unless specified conditions *1013are found to exist.” Id. Accordingly, no protectible interest arose. Other courts of appeals and district courts addressing the identical issue have adopted this interpretation of Greenholtz. See, e.g., Wagner v. Gilligan, 609 F.2d 866 (6th Cir. 1979); Shirley v. Chestnut, 603 F.2d 805 (10th Cir. 1979); Robinson v. Mabry, 476 F.Supp. 1022 (E.D.Ark. 1979); Austin v. Armstrong, 473 F.Supp. 1114 (D.Nev. 1979); Campbell v. Montana Board of Pardons, 470 F.Supp. 1301 (D.Mont. 1979).
In 1974 there was no mandatory language in the Delaware statute authorizing the promulgation of regulations. It merely stated that the department “may” establish such rules.2 That is not enough. I would concede that rules using mandatory language would suffice, but I find no such wording here. The regulations merely list some general (not all-inclusive) standards. They speak in terms of “basic criteria” and “basic procedures.” Nowhere is there language requiring that if these criteria are met the authorities “shall” grant work release status. Indeed, “basic procedures” state that approval must be given at three different levels. By no fair construction can this language be interpreted to mean that those approvals “shall” be given. See Shirley v. Chestnut, supra at 806-07.
I am convinced that the majority’s “consistent with” test does not comply with the stringent requirements of Greenholtz, which I read as insisting upon “mandated by” or “directed by.” Having failed to meet that standard, therefore, Winsett failed to establish a protectible interest. The only other way that a liberty interest could be created would be through the alternate theory of a prisoner’s expectancy of conditional release expressly rejected in Greenholtz and by the majority here. That approach would build on Morrissey v. Brewer, 408 U.S. 471, 92 S.Ct. 2593, 33 L.Ed.2d 484 (1972), and derive a liberty interest in the possibility of release from the Constitution itself.
The Morrissey Court held that a parolee’s interest in not having his parole revoked is an interest protected by due process. Central to this holding was the parolee’s reliance on at least an implicit promise that his freedom would be taken away only if he failed to live up to the parole conditions— the “essence of parole.” Id. at 477, 482, 92 S.Ct. at 2600. For this reason the Court concluded that the termination of parole liberty inflicts a “grievous loss” on the parolee. The interest in continued release, therefore, rose to the level of “liberty” protected by the due process clause independent of the statutory or regulatory language establishing the parole system.
In Greenholtz the claimants argued that the interest at stake in both a parole-revocation decision and a parole determination is the same, and the two situations should be accorded the same constitutional protection. 442 U.S. at 9, 99 S.Ct. at 2104. In rejecting this argument the Court again emphasized the predominately discretionary nature of parole and stressed the crucial difference between the grant and revocation of parole — “being deprived of a liberty one has . . and being denied a conditional liberty that one desires.” Id. In addition, the Court noted that revocation of parole involved the comparatively simple task of resolving the factual question of whether a violation had occurred. The parole release decision by contrast may be made “ ‘for a variety of reasons and often involves no more than informed predictions’ ” as to what best would serve correctional purposes or the interest of the prisoners. Id. at 10, 99 S.Ct. at 2105, quoting Meachum v. Fano, 427 U.S. 215, 225, 96 S.Ct. 2532, 2538, 49 L.Ed.2d 451 (1977).
The Court made it clear that the possibility of parole provides no more than a hope *1014that it will be obtained. “To that extent the general interest asserted here is no more substantial than the inmate’s hope that he [would] not be transferred to another prison, a hope which is not protected by due process.” Id. at 11, 99 S.Ct. at 2105. To support this last point, the Court cited Meachum v. Fano, supra, where it was held that a prisoner could be transferred at the discretion of prison officials, “for whatever reason or for no reason at all.” 427 U.S. at 228, 96 S.Ct. at 2540.
The practical effect of the holding in the case at bench — given the lack of regulatory or statutory language limiting the discretion of the prison authorities — is to find an implied promise in any work release program that an application will be denied only for cause. This is precisely the line of analysis that was rejected by the Green-holtz Court when it distinguished Morrissey, and I do not agree that Greenholtz should be circumvented in this fashion.3
C.
Even if contrary to my view, the “consistent with” standard can be squeezed into the Greenholtz mold, however, I do not accept the majority’s narrow reading of “work release policy.” The majority finds this policy not in the regulations or in the enabling statute, but in § 6531 of title 11 of the Delaware Code, which governs in general the treatment of “persons committed to the institutional care of the Department [of Correction].” The statute provides that such persons are to be treated humanely and guided toward rehabilitation. I gather that in assessing work release applications, the majority would approve such factors as risk of flight, the inmate’s personal safety, and the prognosis for the inmate’s behavioral adjustment. See Majority Opinion at 1006-1007. Although it understandably does not attempt to delineate definitely the outer perimeters of matters it believes to be within the officials’ discretion, “concern for public reaction,” at least, is outside the pale.
I cannot accept this restriction because nothing in the enabling statute suggests that the Department of Correction was not to have very wide discretion in implementing a work release program, and the regulations themselves do not impose any stringent guidelines. In recognizing such matters as “what is best for the community,” weakening deterrent effect on others, and undermining respect for the administration of justice as appropriate considerations for parole release, Greenholtz cannot be reconciled with the majority’s view that it is improper for prison officials to allow “concern for public reaction” to enter into the determination of whether to grant work release privileges.4
Moreover, reliance upon § 6531, without any consideration of the standards for parole enunciated in § 4347, is an illogical limitation on the prison officials’ discretion. Analogizing to the standards governing the parole decision would be appropriate since the regulations on work release refer to eligibility for parole and, obviously, the requirements for the two are interrelated. As shown in footnote 8 of the majority opinion, eligibility for parole was one of the critical factors affecting the decision to grant work release. According to the prison officials who testified at trial, the denial of parole after work release had been granted was a devastating experience for a prisoner, and, consequently, if early parole *1015could not reasonably be anticipated, work release was contraindicated.
In Delaware, parole may be granted when there is a reasonable probability that the person can be released “without detriment to the community or to himself . . . A parole shall be ordered only in the best interest of society . . . .” Del. Code Ann. tit. 11, § 4347 (Michie 1979). In the case of an offender convicted of first degree murder — as is true of Winsett — the immediate family of the victim may testify before the Parole Board. Id. § 4350(b). Others presenting “information or arguments to the Board” are to do so in writing. Id. § 4350(a). Thus, in Delaware, the decision to grant parole necessarily involves greater discretion than simply deciding whether “humane treatment and rehabilitation” are at stake. The Board may consider potential detriment to the community and the other factors listed in Greenholtz, including public opinion.5
If, as the majority holds, the improper exercise of discretion impinges upon a pro-tectible interest, the bounds of legitimate discretion must be expanded to include consideration of the public interest in deterrence and respect for the administration of justice. Public opinion is not an impermissible factor.
The trial judge stated that the denial of Winsett’s application was motivated, in substantial part, by “prison officials’ concern about a possible retaliation by the Delaware General Assembly.” Winsett v. McGinnes, supra at 1371. Whether the court referred to personal or institutional repercussions is not apparent from the opinion. In his testimony, Keve said that over the years the legislature had given the Department of Correction the privilege of implementing work release and furlough programs. But the legislature also had demonstrated that it would quickly revoke that power if the program proved disappointing. As Keve explained, he “had to be concerned with the inmate population as a whole and not exclusively with the advantage that should be given to any one case.” He faced the possibility that anything done for Winsett that would smack of extra privilege “might cause legislative reaction that would cut back the privileges for all other inmates, and this was a distinct, you know, kind of jeopardy . . . .”
The point at which “legislative oversight” exceeds its proper limits and unduly infringes upon executive or agency action as well as individual interest is not always clearly marked. The majority feels that the boundary was crossed in this case and faults the two corrections officials for not stoutly resisting what is viewed as an invasion. Because they and not the invaders are defendants, it is the corrections officials who will be exposed to a damage assessment should their immunity defense fail in the trial court.
I cannot help but wonder if the court has reached this stage because it, too, has wandered outside its sphere by groping to find a protectible interest where none exists. Arbitrariness is anathema to the judicial mind. But carving out a due process right to work release because of its otherwise illusory character confuses what process is due with the question whether a protectible interest exists. Wolff v. McDonnell, 418 U.S. 539, 556-58, 94 S.Ct. 2963, 2974-75, 41 L.Ed.2d 935 (1974). I fear the court has disregarded Greenholtz’s caution that “there simply is no constitutional guarantee that all executive decisionmaking must comply with standards that assure error-free determinations.” 442 U.S. at 7, 99 S.Ct. at 2104. In a similar vein, Meachum warned against “subject png] to judicial review a wide spectrum of discretionary actions that traditionally have been the business of prison administrators rather than of the federal courts.” 427 U.S. at 225, 96 S.Ct. at 2538.
I would affirm the judgment of the district court.
. Winsett was not the only person who was denied work release after public outcry. Norman Parson, who had been convicted of the attempted rape and brutal murder of a young girl, also alleged that because of intense public resentment he had been denied furloughs and work release. See Parson v. Keve, 413 F.Supp. 111 (D. Del.1976).
. The statute was amended in 1976 to substitute the word “shall” for the word “may.” 60 Del. Laws c.705, § 1 (1976), amending Del. Code Ann. tit. 11, § 6533. Even if this amendment antedated Winsett’s initial application for work release, however, my view of this case would remain unchanged. Under either version of the statute, the regulations actually adopted would have to be taken into account because the enabling statute does not delineate eligibility requirements.
. In meeting the plaintiff’s equal protection claim, the district court observed that there was no proof of class-based discrimination but only an assertion of an individual misapplication of state law. The court, relying on Beck v. Washington, 369 U.S. 541, 82 S.Ct. 955, 8 L.Ed.2d 98 (1962), found the claim to be without merit. Winsett v. McGinnes, supra at 1373-74. Under the circumstances I do not find this conclusion to be erroneous and, in view of the majority’s approach, see no need for elaboration.
. In late 1976, the Delaware legislature created the Institutional Release Classification Board to pass on all releases and consisting of seven members, three of whom were to be private citizens appointed by the Council on Administration of Justice. Del. Code Ann. tit. 11, § 6529A (Michie 1979). According to testimony, the three private citizens were to reflect public sentiment.
. In his letter to the state senator, Superintendent Anderson referred to the concern of Mrs. Paris, the widow of the state police officer whom Winsett shot and killed.