This appeal involves the validity of an eligibility requirement imposed by a union on candidates for officer positions. The union, which is a Springfield, Massachusetts local of the American Postal Workers Union, holds officer elections once every three years. Under its constitution, a candidate for office must have been a member of the union for at least one year (two, if running for president) and — this is the disputed provision — must have attended at least three of the local’s regular monthly meetings in the twelve month period before the meeting at which nominations are made.
The union holds meetings nine months of the year (all months but July, August and December), so to be eligible a candidate must attend one third of the nine scheduled meetings. A member is also credited with attending if he does not attend but has an excused absence justified by job requirements or “other compelling reasons”; excuses can be granted by vote of the membership at the missed meeting (if the excuse request is submitted in advance) or at the next meeting. Ultimately, the question for us is whether the three-meetings requirement conforms to the statutory requirement that
every [union] member in good standing shall be eligible to be a candidate and to hold office (subject to ... reasonable qualifications uniformly imposed).
Labor-Management Reporting and Disclosure Act of 1959 (“LMRDA”), 29 U.S.C. § 481(e).
In March 1997, union members nominated candidates for office, but three nominees were declared ineligible because they had failed to satisfy the three-meetings requirement. After the nomination meeting, a union member exhausted internal remedies and then, following the election in May 1997, filed a protest with the Secretary of Labor challenging the election on the ground, inter alia, that the three-meetings requirement was invalid. The Secretary thereafter brought suit against the union in the district court under the LMRDA, 29 U.S.C. § 482(b), which in substance empowers the court to provide equitable relief for violations of section 481, including its eligibility provision quoted above. Alleging such a violation, the Secretary sought to set aside the election.
On cross motions for summary judgment, the district court ruled in favor of the-union after concluding that the three-meetings requirement is a “reasonable qualification” under section 481(e). The Secretary has appealed, stressing that the requirement disqualified over 96 percent of the union membership. The union, supported by a very helpful amicus brief from another union (the United Steelworkers), supports the district court’s decision. The appeal turns on issues of law which we consider de novo. Philip Morris Inc. v. Harshbarger, 122 F.3d 58, 62 (1st Cir. 1997).
In enacting the LMRDA, Congress was concerned with abuses in the labor movement, including not only corruption but attempts by incumbent union leadership to entrench itself further. Local 3489, United Steelworkers v. Usery, 429 U.S. 305, 309-10, 97 S.Ct. 611, 50 L.Ed.2d 502 (1977) (“Steelworkers ”); Wirtz v. Hotel, Motel & Club Employees Union, Local 6, 391 U.S. 492, 497-98, 88 S.Ct. 1743, 20 L.Ed.2d 763 (1968). Section 481(e) aimed to improve union democracy by making it easier for rank and file members to participate in union governance. Hotel, Motel & Club Employees Union, 391 U.S. at 497-98, 88 S.Ct. 1743. For that reason, the Secretary and the courts have approached skeptically qualification requirements of the kind at issue in this case. Those that *3impose an appreciable burden on members — to attend many meetings or to decide to do so long in advance of election— usually fail.1
Still, the courts have purported to make “reasonableness” a case-specific issue, turning on all of the circumstances, e.g., Steelworkers, 429 U.S. at 313, 97 S.Ct. 611; Local 14.02, Int’l Longshoremen’s Ass’n, 617 F.2d at 98-99; and the Secretary says that attendance requirements can serve legitimate ends in ensuring that candidates are educated about union affairs. 29 C.F.R. § 452.38(a) (1999). Union election by-laws have sometimes been crudely exclusionary, e.g., Hotel, Motel & Club Employees Union, 391 U.S. at 500-01, 88 S.Ct. 1743, and it is easy to be cynical about motive; but there is no per se ban on all meeting-attendance requirements.
Instead, the courts have concerned themselves primarily with evaluating the extent of the burden imposed by the requirements at issue and their impact on candidacies. Burden and impact are not quite the same thing, since a burdensome requirement might still be met by many union members (e.g., if it were common practice for most members to attend most meetings) and a slight condition (e.g., attending the nomination meeting) might in practice be met by very few union members. Reversing the district court’s approach, let us begin with burden and return then to impact, which presents the more difficult issue in this case.
To make a refined judgment of burden, one would need a great deal of specific information about actual practice in each case. For example, one would need to know just how inconvenient is attendance for members of a particular union (which depends on the location(s) of meetings, the hour and length of the meetings, the distance from home or job of the member, and so on), and whether there are other incentives for members to attend meetings, as well as the nature and operation of any excuse regime. But it is rare that either side offers such detailed information, let alone reliable evidence, which might well require surveys and statistics.
In the absence of such evidence, courts tend to make what they think are “common sense” judgments based on how many meetings members must attend, over what period of time, whether excuses are allowed, and how long in advance a candidate would have to decide to run in order to fulfil the conditions. E.g., Steelworkers, 429 U.S. at 313, 97 S.Ct. 611; Local 1402, Inti Longshoremen’s Ass’n, 617 F.2d at 98-99. As the numbers of excluded members go up, so does the courts’ readiness to find that the burden outweighs the supposed (and inherently speculative) benefits of the requirement in educating the candidate or demonstrating commitment. See cases cited at note 1, above.
By the standards of prior decisions, the qualification requirements in this case are certainly at the less burdensome end of the spectrum. The candidate need attend only three of the nine meetings in the year prior to the nomination meeting; the decision to run could be made as late as four months before the nomination meeting; and an excuse regime exists that would allow credit for a meeting not actually attended on a showing of job requirements or other compelling reasons. Most of the leading cases striking down requirements have dealt with ones at least marginally more demanding.
Of course, if one believed that meeting attendance requirements serve no purpose *4but to ward off challenges to incumbent union leaders, any burden at all might suffice to condemn them. But the amicus offers some commentary to the contrary— principally a statement by former Labor Secretary Ray Marshall, an expert in labor matters — and, more important, the Secretary does not assert that all meeting attendance requirements are a useless sham or point us to any evidence to support such a position. If and when the Secretary does so, it will be time enough to revisit that basis for banning the requirements.
Instead, the Secretary’s attack on the district court decision rests almost entirely on the alleged impact of the attendance requirements in this case, on her reading of case law, and on the deference she says is due to her judgment. The main datum relied on by the Secretary is the fact, conceded by the union, that over 96 percent of the members of the local in this case did not in practice attend enough meetings to qualify to run for office. Put differently, when the nomination meeting arrived, less than 4 percent of the local’s membership could run for office.
There is nothing crazy about the view professed in the Secretary’s brief that any requirement disqualifying so many potential candidates should be automatically deemed unreasonable. Such a rule would be easily administered (once one chooses the fatal percentage); it would lessen the need for speculative judgments about actual burden; • and it would enlarge “union democracy” by providing a larger body of possible candidates (albeit while contracting the unions’ ability to manage their own affairs). But to make a percentage test conclusive is not the most natural reading of the statute or the governing precedents, and there is doubt about the Secretary’s own position' — outside of this litigation.
The statutory “reasonableness” test in section 481(e) does not have much inherent content but, as already noted, the main thrust of the decisions implementing it is to consider election-qualification requirements on a case by case basis, taking account of all relevant factors and not just one. True, the Secretary’s brief reads Steelworkers as making conclusive the percentage of members excluded; but while the dissenting opinion in that case so accused the majority, Steelworkers, 429 U.S. at 315, 97 S.Ct. 611 (Powell, J., dissenting), the majority avoided any such explicit holding, id. at 310-13, 97 S.Ct. 611,2 and the facts of the case involved a requirement — attending one half the monthly meetings in the prior three years — that was far more burdensome than the requirement at issue in this case.
Further, looking at “impact” solely from the nomination day perspective is misleading. Since most union members do not intend to run for office in a given election, most are not going to trouble to meet any precondition that requires them to do anything different from what they ordinarily do — no matter how slight the burden or how reasonable a condition for those who do wish to be candidates. Suppose that the union required all candidates to send in a postcard 10 days before the election or to collect three signatures prior to the election; probably, at least 96 percent of the membership would on the day of the election still be disqualified from running for office.
One circuit court decision does — to some extent — support the view taken in the Secretary’s brief. In Doyle v. Brock, 821 F.2d 778 (D.C.Cir.1987), a divided panel overturned the Secretary’s refusal to sue to enjoin a meeting attendance requirement obligating a candidate to attend at least one half of the monthly meetings in the *5prior year. That requirement (viewed ex post) also disqualified a high percentage of candidates. Pointing to Steelworkers ’ emphasis on such percentages, Judge Edwards (Joined by a visiting judge) held that the Secretary had not adequately explained his refusal to challenge the union. Doyle, 821 F.2d at 785-86.
However, Doyle was rendered over a dissent by Judge Silberman, id. at 787-89, and over strenuous arguments from the Secretary that Steelworkers did not create a per se rule, see id. at 784. Further, the facts in Doyle addressed a requirement somewhat more burdensome than the one before us, compelling attendance at six meetings (instead of three) and a head start of about six months (instead of four). Finally, if Steelworkers is taken to impose a rigid, one-factor test based on percentage, much of the opinion becomes surplus-age, particularly the section in which the Court approved the Secretary’s own flexible, multifactor test for reasonableness. 429 U.S. at 311-13, 97 S.Ct. 611.
This brings us to the Secretary’s more interesting argument, namely, that her present position on what is a reasonable qualification is entitled to substantial deference from the courts. Traditionally, the most deference is due to decisions by agencies made pursuant to delegated power, cf. Chevron U.S.A., Inc. v. Natural Resources Defense Council, 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), an authority that the Secretary does not claim here. But we have ourselves said that some weight is due even to interpretative rules that express merely the view on the matter at hand taken by the agency charged with enforcement.3 This deference does not depend on notions of delegation but on the expertise and comparative neutrality of the agency involved.
But there are major difficulties with according such deference here. So far as the Secretary’s position is based simply on a reading of Steelworkers, our own competence in parsing that decision is equal to that of the Secretary. See Valerio v. Putnam Assocs. Inc., 173 F.3d 35, 43 n. 5 (1st Cir.1999) (court not bound by agency’s interpretation of judicial precedent); accord Atchison, Topeka & Santa Fe Ry. Co. v. Pena, 44 F.3d 437, 443 (7th Cir.1994) (en banc), affd without discussion of this point sub. nom, Brotherhood of Locomotive Eng’rs v. Atchinson, Topeka & Santa Fe Ry. Co., 516 U.S. 152, 116 S.Ct. 595,133 L.Ed.2d 535 (1996). Indeed, as Doyle makes clear, the Secretary previously shared our view of the Steelworkers decision. See Doyle, 821 F.2d at 784.
Further, even in her modified interpretative regulation adopted after Doyle, the Secretary has taken an ambiguous position. The text of the regulation continues to assert that attendance requirements may be reasonable and that “reasonableness must be gauged in the light of the circumstances of the particular case,” including, among other considerations, burden, excuse provisions, and impact. 29 C.F.R. § 452.38(a). Only in the new “case law” footnote at the end does the Secretary, citing Doyle, say that the exclusion of a “large portion of members” alone “may” render an attendance requirement unreasonable. Id. at n. 25.
Thus, in her most recent formal statement of position, the Secretary has sought to have it both ways, supporting the union position in text4 but appending a footnote *6reference to Doyle that is more of a disclosure than an endorsement. Certainly the Secretary’s brief in this court is unqualified in supporting some percentage test as determinative — indeed, the brief scarcely mentions burden — but courts normally decline to defer to what may be the ad hoc position of counsel rather than the judgment of the administrator. See, e.g., O’Connell v. Shalala, 79 F.3d 170, 179 (1st Cir.1996); Massachusetts v. Blackstone Valley Elec. Co., 67 F.3d 981, 991 (1st Cir.1995).
We might well decide this case differently if the Secretary adopted an interpretative regulation setting forth a conclusive percentage test. This is not the most natural or traditional reading of the statute, but there are policy arguments — a few already mentioned — for this result, assuming (as we do) that the statute affords some latitude for the expert administrator to devise an enforceable regime. And while Steelworkers does not compel such a percentage test, arguably Steelworkers would permit it, if based on the Secretary’s independent judgment.
Such a flat percentage test would largely eliminate meeting attendance requirements and repudiate prior positions of the Secretary, and might therefore require more explanation than is normally required for a regulation, but agencies are entitled to change their views. Strickland v. Commissioner, Maine Dep’t of Human Servs., 48 F.3d 12, 20 (1st Cir.), cert. denied, 516 U.S. 850, 116 S.Ct. 145, 133 L.Ed.2d 91 (1995); Massachusetts v. Lyng, 893 F.2d 424, 431 (1st Cir.1990). Despite the able arguments of the amicus union, we do not attach much weight to the prevalence of such meeting attendance requirements in the past — a past not without abuses — or to Congress’s failure to ban them outright, which falls far short of implicit approval.
In sum, if the Secretary has the courage of her counsel’s convictions, she can revise her interpretative regulation, adopt a flat percentage test, and eventually bring an enforcement action outside the D.C. Circuit to test her position. There are apparently plenty of candidates, including the very unions involved in this case. Obviously, we do not commit ourselves as to the outcome but, with a clear-cut regulation adopted by the Secretary, in place of the current straddle, this would at least be a close contest.
Affirmed.