When the first federal census was conducted in 1790, the total population of the United States was less than 4,000,000.1 New York City, the country's largest municipality, had only 33,000 inhabitants.2 Despite these relatively small numbers, the census, which was conducted by United States Marshals, did not go smoothly. Transportation and communication were inadequate.3 Boundaries of towns and cities were often undefined.4 Cooperation by the citizenry was less than complete. Some people distrusted the newly-formed federal government; others were influenced by a fear of increased taxation; many opposed the basic concept of a census.5

Because art. 1, § 2, cl. 3 of the then newly-enacted Constitution provided that members of the House of Representatives and direct taxes should be apportioned among the States according to population, it had been anticipated that the census would provide an accurate count. A State's temptation to exaggerate its population count for purposes of congressional representation would be offset by its desire to reduce its apportioned share of direct taxes.6 However, when the census was finally completed in 1791, there was a fairly wide-spread belief that it had resulted in a substantial undercount.7

Although the mechanics of the counting process have been improved in each of the nineteen ensuing censuses, there has never been a perfect count.8 This is concededly true of the 1980 census. In keeping with the spirit of the times, States and municipalities in various parts of the country are seeking to remedy this most recent imperfection through litigation.9 Approximately fifty lawsuits have been brought by or on behalf of subordinate government bodies, in each of which the claim of a substantial local or regional undercount is made.10

This is an appeal from a judgment of the United States District Court for the Southern District of New York, 508 F. Supp. 420, in an action alleging undercounts in the State and City of New York. The judgment orders the Census Bureau to adjust population figures for the State and City "in a reasonable and scientific manner" and to report to the court within thirty days the Bureau's plan to effectuate the court's ruling.11 For reasons hereafter assigned, we reverse and remand for a new trial.

Although the census is the delight of statisticians and sociologists and serves as a convenient measuring stick for the dispensing of federal funds, it was not created for those purposes. Its purpose under the Constitution was to determine the apportionment of Representatives among the States. "Representatives ... shall be apportioned among the several States ... according to their respective Numbers ...."12 In reviewing the judgment on appeal, it is important that we keep this basic constitutional purpose in mind.

The House of Representatives has 435 members, and this number must be apportioned among the fifty States. If one State gains a member, another must lose one. Following the 1960 census, seven States each gained one seat, one State gained four, and one gained eight. As a result, twelve States each lost one seat, three States each lost two seats, and one State lost three seats.¹³ Following the 1970 census, three States each gained one seat, one gained three, and one gained five. Seven States each lost one seat, and two States each lost two seats.¹⁴ In effect, House membership is a fund in which fifty States have an interest. No State's share can be increased without adversely affecting at least one other State. The question presented by litigation such as the one now before us is whether one State can be granted such an increase without full consideration having been given to its effect on other States.

Persons "who not only have an interest in the controversy but an interest of such a nature that a final decree cannot be made without either affecting that interest, or leaving the controversy in such a condition that its final termination may be wholly inconsistent with equity and good conscience" are traditionally considered to be indispensable parties.15 Equity suggests that a person be brought into the litigation if the case cannot be decided on its merits without prejudicing his rights.16 This not only prevents possible injury to the absent person; it avoids multiplicity of suits and the danger of inconsistent decisions.17

Although the present action has already been before this Court on an appeal from a preliminary injunction order, we did not consider any of the factors just set forth. The panel hearing that appeal held only that plaintiffs have standing to sue and that the issue of census mismanagement is justiciable rather than political.18 The panel did not address appellants' contention that the issuance of the preliminary injunction order was an abuse of discretion because it "overlook(ed) the impact on other cities and states."19 The panel concluded that "(t)he argument that a court decision may provide special treatment for the parties involved is one for the ultimate trial on the merits and decision on appeal."20 We have now reached that point in the litigation.

We think it clear beyond cavil that a statistically formulated increase in the population of only one State, such as New York, will have an adverse effect on other States which are entitled to, but do not receive, the benefit of a similar adjustment. Even if the increase is insufficient to change House membership, it will nonetheless increase New York's share in the numerous revenue sharing plans that are tied into the census. The adversely affected States therefore fall within the category of parties who should be joined in the instant litigation if feasible.21 Because compulsory joinder of all fifty States was not feasible in the district court, pragmatic equitable alternatives should have been considered.22

The first alternative, which the prior panel of this Court has already rejected, would have been to substitute Bureau and congressional review for that of the court.23 This is the recommendation of the Association of the Bar's Special Committee on Empirical Data in Legal Decision Making which concludes its report with the following observation:

The path to a better census is more likely to be found in scrutiny of Bureau procedures by the Bureau itself, Congress, other federal agencies, and interested professional groups than in litigation to compel adjustment.24

A second alternative would have been to require that notice of suit be given to all of the States, with permission to intervene given any State which felt that its interests were imperiled. Apparently, no notice was given anyone; only the County of Suffolk sought intervention, and its application was denied.

A third alternative would have been to seek multidistrict coordinated or consolidated pretrial proceedings pursuant to 28 U.S.C. § 1407. It appears that thirty-one of the above listed actions have been transferred to the District of Maryland for coordinated or consolidated proceedings.25 The wisdom vel non of adopting a uniform nationwide method of making statistical adjustments might perhaps be fully explored if complete multidistrict joinder were agreed upon.

As a fourth alternative, the district court could have stayed the operation of its judgment pending a definitive appellate ruling as to whether the Census Bureau's conduct of the 1980 census could properly be challenged in fifty separate and unrelated actions and whether the requirements imposed on the Bureau by the district court in this action were proper.26

Without passing on the wisdom or necessity of any of the foregoing alternatives, we conclude that, at the very least, the district court was required to conduct the trial in such a manner that the interests of other States and subordinate government bodies were not prejudiced by rulings that prevented a full and fair development of the facts. Because error and prejudice resulted from certain of the district court's discovery orders and the sanctions imposed in connection therewith, for this reason, if for no other, there must be a new trial.

As part of their pretrial discovery, plaintiffs requested disclosure of the Master Address Registers for New York City and every other municipality in the State, together with a listing of all the vacant housing units in every New York State municipality. The Master Address Registers contain information based upon interviews and questionnaires and include data as to householders, house members, street names, apartments or unit numbers, descriptions of location, numbers of units, numbers of persons, etc. The Census Bureau estimated that there are more than 550,000 pages of address registers for the New York region alone. The vacant housing forms, known as D-160 Unit Status Review forms, contain information as to the vacant or occupied status of a unit, as obtained from the owner, custodian, manager, a neighbor, or another individual with knowledge. Between 500,000 and 600,000 D-160 forms are estimated to exist for the New York region. The Bureau took the position that the requested information was confidential, that its disclosure would violate 13 U.S.C. § 9 and would undermine the public confidence needed for the effective conduct of the census.27

The district court directed the Bureau to produce the documents in question, the court's sole concession to the Bureau's claim of confidentiality being the grant of permission to redact "only" the names of persons residing in a household. The court ordered that any person granted access to the documents should be sworn to confidentiality pursuant to 13 U.S.C. § 23(c).28 Upon the Bureau's refusal to comply with the production order, the district court precluded the defendants from offering any evidence to prove "any fact, matter or circumstance that would be reflected in or could be derived from the (Master Address Registers or D-160 forms) or that could be contradicted or impeached by such documents." Defendants were also precluded from "denying or offering evidence or testimony to disprove" the following allegations in plaintiffs' complaint:

(a) that there was a disproportionate undercounting of Blacks, Hispanics, other racial and ethnic minorities, legal and illegal aliens, persons whose principal language is other than English and persons living in poverty or high-crime urban areas;

(b) that the persons above described live in New York State or New York City in substantially disproportionate numbers as compared to the nation as a whole; that therefore the undercount for New York will be substantially higher in percentage terms than it will be for the rest of the United States and in absolute numbers it will be larger than the undercount for any other State;

(c) that the 1980 census operations in New York, and particularly in New York City, have been carried out in a manner that will greatly exacerbate the undercount in New York as compared to other parts of the nation;

(d) that the particular difficulties encountered by the Census Bureau and the poor management of census operations in New York State, and especially New York City, have made judicial relief essential;

(e) that the Census Bureau with knowledge of the difficulties it would encounter in New York State and New York City and with knowledge that undercounting would be disproportionately severe there, especially among the class described in paragraph (a) above, knowingly and intentionally failed to devote adequate resources to develop special plans and procedures, to assign the most qualified personnel and to carry out their own planned procedures in New York State, and particularly New York City;

(f) that the first follow-up operations in New York State and New York City were also seriously mismanaged (Six paragraphs, covering four printed pages in the Appendix, set forth alleged examples of mismanagement under this general allegation of wrongdoing);

(g) that the second follow-up operations were not being adequately staffed or managed in New York; they were not designed to count additional households, correct inaccuracies in the size of households, or to recheck the casual count inaccuracies;

(h) that the local review program was being improperly conducted by an inadequate staff and the City had been given neither adequate time nor adequate information to review and challenge the results.

Looking first to the discovery order, we conclude that it was improvidently granted. In McNichols v. Klutznick, 644 F.2d 844 (10th Cir. 1981), one of the actions pending in other jurisdictions, the district court had granted the identical relief ordered by the court below. The Court of Appeals reversed, stating that "(b)oth the history of the Census Act and the broad language of the confidentiality provisions of § 9 make abundantly clear that Congress intended both a rigid immunity from publication or discovery and a liberal construction of that immunity that would assure confidentiality."29 Although the Third Circuit appears to have taken a contrary position,30 we agree with the holding of the Tenth Circuit.

Assuming for the argument, however, that the court below did not err in granting discovery, it clearly abused its discretion in the imposition of sanctions. As already pointed out, New York is seeking to gain an advantage over other States in the competition for congressional seats and federal funds. The extremely broad sanction order punished these unrepresented States by preventing full and fair disclosure of the facts and an adjudication on the merits of plaintiffs' claims.31 The severity of sanctions should be tempered by a consideration of the equities involved.32 The equities in the instant case dictated that the unrepresented States not be penalized because the Census Bureau was attempting to recognize the public's interest in a confidential census.

Reversed and remanded for further proceedings consistent with this opinion.

STEWART, District Judge (concurring in the judgment):

Although I concur in the judgment, I write to express my disagreement with the court's analysis in overturning the order of the district court that the follow-up address registers (FARs) and D-160 Unit Status Review forms be disclosed to the plaintiffs with limitations. I also consider the scope and the effect of the holding overturning the preclusion order.

The statutory privilege raised by appellants as a barrier to disclosure is substantial and undoubtedly necessary. However, I do not believe that the statute under which the government claims an absolute privilege from disclosure is so broad as to protect most of the information sought by plaintiffs from the limited disclosure ordered below.1

The starting point in analyzing the scope of a statutory privilege must be the text of the statute, see United States v. Missouri Pacific R. R. Co., 278 U.S. 269, 277-78, 49 S. Ct. 133, 136-37, 73 L. Ed. 322 (1929), and the statute must be strictly construed. St. Regis Paper Co. v. United States, 368 U.S. 208, 218, 82 S. Ct. 289, 295, 7 L. Ed. 2d 401 (1961). The defendants bear the burden of proving that a privilege exists. Heathman v. United States District Court, 503 F.2d 1032, 1033 (9th Cir. 1974) (citing 8 Wright & Miller, Federal Practice & Procedure: Civil, § 2016 at 126).

§ 9. Information as confidential; exception

(a) Neither the Secretary, nor any other officer or employee of the Department of Commerce or bureau or agency thereof, may, except as provided in section 8 of this title

(1) use the information furnished under the provisions of this title for any purpose other than the statistical purposes for which it is supplied; or

(2) make any publication whereby the data furnished by any particular establishment or individual under this title can be identified; or

(3) permit anyone other than the sworn officers and employees of the Department or bureau or agency thereof to examine the individual reports.

No department, bureau, agency, officer, or employee of the Government, except the Secretary in carrying out the purposes of this title, shall require, for any reason, copies of census reports which have been retained by any such establishment or individual. Copies of census reports which have been so retained shall be immune from legal process, and shall not, without the consent of the individual or establishment concerned, be admitted as evidence, or used for any purpose in any action, suit, or other judicial or administrative proceeding.

13 U.S.C. § 9(a) (1976). Appellants contend that the statute extends confidentiality to "all information or data provided to the Census Bureau under the Act." Appellants' Br. at 53-54.² The opinion of the court merely endorses the holding of the Court of Appeals for the Tenth Circuit that "the broad language of the confidentiality provisions of § 9 make abundantly clear that Congress intended both a rigid immunity from publication or discovery and a liberal construction of that immunity that would assure confidentiality." Supra, at 739, quoting McNichols v. Klutznick, 644 F.2d 844 at 845 (10th Cir. 1981), petition for cert. filed sub nom., McNichols v. Baldridge, 49 U.S.L.W. 3826 (U.S. April 24, 1981) (No. 80-1781). The express terms of the statute are not so broad. Subsection 9(a)(1) limits the use of "information furnished under the provisions of this title" to statistical purposes. This provision recognizes the purpose of the census as merely gathering sociological data rather than enabling law enforcement or dictating substantive policy. Subsection 9(a)(2) protects the confidentiality of respondents in the enumerated population by prohibiting dissemination of information identified with "any particular establishment or individual" beyond officers or agents of the Department of Commerce or the Bureau. This is also consistent with the benign, data gathering function of the census. Subsection 9(a)(3) is more restrictive than § 9(a)(2), limiting access to individual reports to "officers and employees of the Department or bureau." Thus, the express protection of the statute is circumscribed: (1) census information is to be used only for statistical purposes; (2) no one may publish information which identifies respondents; and (3) no one without statutory authority may peruse individual reports.

Moving beyond the language of § 9, its legislative history does not support appellants' all-embracing interpretation. Appellants identify as indicative of the "strong congressional intent against all disclosure" the swift reaction of Congress to the decision of the Supreme Court in St. Regis Paper Co., supra. Appellants' Br at 51 n.*. The Court had decided that, as written, § 9 did not protect copies of census questionnaires voluntarily retained by respondents. 368 U.S. at 218, 82 S.Ct. at 295. Within one year Congress adopted the flush language following § 9(a)(3)3 to overrule St. Regis Paper Co. The legislative history pertaining to the amendment shows its limited purpose.

The purpose of this legislation is to provide specifically that company-retained copies of census reports submitted to the Bureau of the Census shall have the same confidential status which is afforded to the original census reports ....

S.Rep.No.2218, 87th Cong., 2d Sess. 2 (1962), reprinted in (1962) U.S.Code Cong. & Ad.News 3188, 3189. The Department of Commerce request for enactment of this measure stated that, with the requested amendment, " ... a respondent will be free to give information to Census for statistical purposes only, without fear that a copy of the return can be demanded by a regulatory agency and used against the respondent." Letter from Secretary of Commerce Luther H. Hodges to President Lyndon B. Johnson, reprinted in (1962) U.S.Code Cong. & Ad.News 3190, 3191. The relationship between respondents and the Bureau is repeatedly emphasized in Secretary Hodges' letter. Id. The statutory response to St. Regis Paper Co. was thus confined to the narrow holding of the case.

Although the St. Regis Court was specifically concerned with the relative protection afforded census forms retained by respondents and those held by the Census Bureau, the Court stated that "the prohibitions against disclosure contained in § 9 ... do not purport to generally clothe census information with secrecy." 368 U.S. at 217-18, 82 S.Ct. at 294-95. The Court analyzed each of the three subsections of 13 U.S.C. § 9 in the same manner as I have done here. Id. at 215-16, 82 S.Ct. at 293-94. The strict construction given to § 9 by the Supreme Court did not occasion any fundamental rethinking or clarification of the statute. The statute could have been amended to include a blanket prohibition on dissemination of the information obtained by the Bureau from any source, for any purpose. It was not so amended. In fact, the Committee on Post Office and Civil Service of the House of Representatives expressly rejected a statutory revision that would prevent use of "all copies of (census) information, reports and other data" in judicial and administrative proceedings. H.R.Rep.No.2437, 87th Cong., 2d Sess. 1-2 (1962). See also H.R.Rep.No.10569, 87th Cong., 2d Sess. (1962), printed in Hearing Before the Comm. on Post Office and Civil Serv., Confidentiality of Census Reports, 87th Cong., 2d Sess. (1962).

The limitation of confidentiality to information furnished to the Bureau by census respondents was emphasized in an early predecessor to § 9 which authorized the Director of the Census to collect and publish statistics relating to cotton seed. See Act of Aug. 7, 1916, Pub.L.No.64-177, ch. 274, § 2, 39 Stat. 436 (1916) ("1916 Act"). The 1916 Act provided "(t)hat the information furnished by any individual establishment under the provisions of this Act shall be considered as strictly confidential and shall be used only for the statistical purpose for which it is supplied." Id. (emphasis added).

I also derive support for this interpretation of § 9 from the express language of 13 U.S.C. § 8, a statute empowering the Secretary of Commerce to make certain disclosures subject to the limitations of § 9. Subsection 8(b) permits disclosure of tabulations or statistical materials "which do not disclose the information reported by, or on behalf of, any particular respondent." See McNichols v. Klutznick, supra, at 845 (emphasis added), petition for cert. filed sub nom., McNichols v. Baldridge, 49 U.S.L.W. 3826. Subsection 8(b) is entirely consistent with § 9(a) in seeking to protect the identity of census respondents. Subsection 8(c) illustrates another means of protection for census respondents. The information furnished to any respondent or government unit under § 8 may not "be used to the detriment of any respondent or other person to whom such information relates." 13 U.S.C. § 8(c) (1976).

I can only conclude that the privilege conferred by § 9 is not as broad as the court's opinion states. Nonetheless, the express language of the statute amply protects the confidentiality interests of census respondents and promotes the purposes of accurate enumeration of the population. See In re FTC Corporate Patterns Report Litigation, 432 F. Supp. 291, 305 (D.D.C.1977), aff'd sub nom. Appeal of FTC Line of Business Report Litigation, 595 F.2d 685, 699 (D.C. Cir. 1978).

Broadening our search for the limits of the statutory privilege conferred by 13 U.S.C. § 9(a) to include judicial construction of the statute does not materially alter this picture of substantial, but finite, confidentiality. Defendants rely heavily on Seymour v. Barabba, 559 F.2d 806 (D.C.Cir. 1977) (per curiam), for the absolute confidentiality of matter gathered by the Bureau, whether or not the data is reported to the Bureau by census respondents. Seymour held that the Bureau did not have to disclose to third parties the names and addresses of companies falling within 21 of the Bureau's "standard industrial classifications." Id. at 809. The information was sought to notify potential class members of an antitrust suit. Plaintiff argued that the names and addresses were not furnished under the Census Act and were not gathered for statistical purposes. With respect to the latter point, the opinion reasoned that names and addresses are "sufficiently related to the statistical data which the firms eventually are called upon to report that it is protected under Section 9(a)(1)." Id. The statistical purpose of collecting names and addresses for standard industrial classifications is apparent: it permits cross-checking data to avoid double counting while ensuring completeness. To the extent that the Seymour Court found names and addresses could not be disclosed without violating the statutory proscription of use of "information furnished under this title" for other than statistical purposes, I agree with its holding. I cannot agree with Seymour, however, that information "gathered," "categorized and assembled" for the Bureau, but not reported by census respondents, acquires statutory protection. My reading of the statute and its purpose, as described above, of ensuring respondents' privacy, leads to the conclusion that "information furnished under the provisions of this title" means information obtained from the enumerated population.

This latter construction was adopted by Judge Weinfeld in United States v. Bethlehem Steel Corp., 21 F.R.D. 568 (S.D.N.Y.1958). Judge Weinfeld found that the "clear and unambiguous" purpose of the statute "was to protect (citizens) who complied with the command of the statute." Id. at 570. United States v. Little, 321 F. Supp. 388 (D.Del.1971), also focussed on the origin of the census data and its use, not merely on the Census Bureau as the repository of any information:

... (T)he information obtained by the census questionnaire is strictly confidential. 13 U.S.C. § 9. It may not be used other than for statistical reporting, and may never be disclosed in any manner so as to identify any individual who has answered the questions.4

Id. at 392. In Little, the court declined to dismiss an information charging the defendant with refusal to answer census questions. Id. The court relied on the statutory privilege protecting respondents' confidentiality in denying defendants' claim that the census questionnaire was an unconstitutional invasion of privacy.

It remains to assess the discovery order in this case in light of the specific and limited provisions of 13 U.S.C. § 9(a). The discovery order provided:

(T)hat (appellants) shall produce to (appellees) by September 10, 1980, the FARs for the City and State of New York for use in checking the accuracy of (appellants') conduct of the 1980 decennial census except that as to those Census Bureau District Offices in New York State which have been closed or from which FARs have been removed by the Census Bureau, the FARs for those offices shall be produced to the (appellees) by September 12, 1980; and it is further

Ordered that (appellants) shall produce to (appellees) by September 10, 1980, for use in checking the accuracy of (appellants') conduct of the 1980 decennial census:

(a) all lists compiled by the Census Bureau of vacant or nonexistent household units in the City and State of New York; and

(b) all Census Bureau D-160 forms for housing units in the City and State of New York other than those indicating that the household unit is occupied; and it is further

Ordered that (appellees) and their attorneys shall use the documents produced pursuant to this order solely for the purpose of helping to ensure an accurate count of the whole number of persons in New York in the 1980 decennial census, and such documents shall not be used or disclosed to any other person for any other purpose whatsoever; and it is further

Ordered that any person granted access to the documents produced pursuant to this Order shall be sworn to confidentiality pursuant to 13 U.S.C. § 23(c); and it is further

Ordered that in producing the documents required herein, (appellants) may redact such documents to remove from them only the names of persons residing in any household enumerated in the 1980 decennial census; and it is further

Ordered that (appellants') motion for a protective order is granted to the extent provided herein, and is in all other respects denied; ....

A.281-83.

I turn first to the FARs. The FARs at issue were completed in six steps. First, addresses were purchased from the commercial mailing firms "that supplied the best list for the area." A.437. That included 2,279,000 addresses. A.449. Second, because the commercial lists ("TARs") were known to be deficient, the TARs were compared to the 1970 census address list and 255,000 addresses were added. A.437, 449. Third, the addresses were printed on post cards and checked by the Post Office for address correction, duplication, business only, undeliverables and address missing. A.437. This added 458,000 addresses. A.449. Fourth, in a "precanvass" operation, enumerators checked the special address registers in the field to verify the accuracy of addresses, to identify missing units and to correct geographic errors. A.437. Precanvass added an additional 219,000 addresses to the master mailing lists. A.449. Fifth, another round of postal checks followed. A.437. Sixth, the FARs were up-dated after questionnaires were mailed out and either mailed back or answered by a personal visit or answered by the enumerators if no response was obtained. A.266-267.

It follows from the foregoing analysis of the statutory privilege that information obtained at the first, third, fourth and fifth steps is not privileged.5 Most importantly, the information sought at these steps was not supplied by census respondents. It is also significant that the information was to be used only for statistical purposes according to appellees' representations, A.270, and the express terms of the discovery order, A.282. There was no dissemination of information that identified any respondent. The names of persons residing in any enumerated household were to be redacted. A.282-83. There was no evidence that the respondent was otherwise identified on the address registers. Appellees did not seek and the district court did not grant access to individual reports. Also, under the protective aspects of the order, persons granted access to the documents were ordered not to disclose or use the FARs "for any other purpose whatsoever," and were sworn to confidentiality.

At the second and sixth steps, however, the statutory privilege may apply, as information furnished by respondents is sought. However, the statutory privilege is not absolute. The information furnished by respondents may be disclosed for statistical purposes as long as respondents are not identified and the individual reports are not disclosed. See text at app. 2-5, supra. So "(t)he potential harm from disclosure of any communication subject to a privilege must be weighed against the benefits of disclosure." Carr v. Monroe Mfg. Co., 431 F.2d 384, 390 (5th Cir. 1970), cert. denied, 400 U.S. 1000, 91 S. Ct. 456, 27 L. Ed. 2d 451 (1971); see United States v. IBM, 1975-2 Trade Cases, P 60,383 (S.D.N.Y.1975). It is appropriate in determining the availability of a privilege to consider protective orders tailored to the case at hand. Carr, 431 F.2d at 390, cert. denied, 400 U.S. 1000, 91 S. Ct. 456, 27 L. Ed. 2d 451. The district court heard argument and considered counsel's submissions before ordering disclosure. The disclosure order recited the competing considerations. The order was crafted to satisfy both the letter and intent of § 9(a). The information was sought only for statistical purposes and the disclosure order prohibited dissemination of information identifying census respondents. Access to the individual reports was never at issue. It was not an abuse of discretion to comply fully with the statutory privilege and order the FARs produced subject to a protective order.

In reviewing the disclosure order of the district court relative to the D-160s "other than those indicating that the household unit is occupied," we first determine what data is contained in D-160s. In addition to the address and vacant condition of the unit, D-160s specify the name of the person providing the information. A.320. This information is supplied by both census employees and respondents who are not residents of the vacant units. The protective order only permitted the Bureau to redact from the documents the names of residents of enumerated households. As disclosure of the D-160s would identify the individual respondents, who may not be census employees, such disclosure would violate § 9(a)(2) as construed above. As the disclosure order with respect to the D-160s inadequately protected the confidentiality of census respondents the order was erroneous.

I conclude that 13 U.S.C. § 9 is not as broad as claimed by appellants, but that it does reach some matter ordered disclosed by the district court.

Following the discovery order, the Census Bureau persisted in its claim of privilege. The district court certified its disclosure order to this court and stayed the production order. Following the October 1, 1980 order by this court denying review, appellees sought a preclusion order under Fed.R.Civ.P. 37(b)(2) for appellants' failure to comply with the disclosure order.

Our review is limited, of course, to determining whether the preclusion order was an abuse of discretion. National Hockey League v. Metropolitan Hockey Club, 427 U.S. 639, 642, 96 S. Ct. 2778, 2780, 49 L. Ed. 2d 747 (1976) (per curiam). There are several novel considerations in issuing sanctions against government officials for failure to permit discovery: (1) the significance of an abuse of discretion is magnified by the importance of the issue and the status of the noncomplying party; and, (2) issue-related sanctions are ordinarily sufficient. See In re Attorney General of the United States, 596 F.2d 58, 65 (2d Cir. 1979). As the opinion of the court emphasizes, the unique nature of this case bears out the importance of limited and necessary sanctions. Nonetheless, the district court found that a preclusion order was appropriate in light of appellants' "knowing and wilful" noncompliance with the discovery order. There is no contention, however, that appellants' reticence was a matter of bad faith, and the district court did not so find.

I agree with the rest of the panel that the preclusion order was overly broad. First, there was ample evidence already made available by appellants that contained essentially the same information as contained in the FARs. The Bureau was producing Form 388s, which contained the total population and housing unit figures from the address registers for each enumeration district. A.352. The 388s also provide totals of vacant housing units. Id. Second, the order effectively determined several issues despite contrary evidence not derived from the FARs or D-160s. For example, there was independent evidence that the undercount in rural areas was greater than the undercount in urban areas, A.635, that could not be offered at trial to rebut the plaintiff's case as a result of the preclusion order. This was a matter of historical fact unrelated to issues of the present undercount in New York and the claims of mismanagement, but tending to disprove appellees' claim. Third, the FARs and D-160s were not necessary to prove several of the claims that were included in the preclusion order. Paragraphs 47(a) and 47(b) of the complaint, for example, allege that inadequate staffing and procedures constituted mismanagement that exacerbated the undercount. The extent of the staffing and the adequacy of procedures are not proved or disproved by anything in the FARs or D-160s.

The court's opinion implies that the district court did not adequately consider the interests of other states as claimants to both the representational and budgetary pies. Clearly the case requires consideration of the interests of other states. But this appeal concerns primarily the issue whether the decision concerning discovery and the subsequent preclusion order as a sanction for failure to permit court ordered discovery was tailored to the facts of the case. I do not agree with the court's opinion that "(t)he question presented by litigation such as the one now before us is whether one state can be granted such an increase (in congressional representation) without full consideration having been given to the effect on other states." At 736. The interests of other states are not, strictly speaking, implicated until different types of relief are being considered, although a decision on the merits regarding both the claimed undercount and mismanagement is highly desirable. From a practical standpoint, I am not at all sure whether the court's opinion purposes to consider other states' interests in a manner that is any different from the adopted course of the district court. The entire panel apparently believes that it is possible to conduct the trial in this case, even though only one state is involved, so long as there is no prejudice to the rights of others. The district court order provided that any appropriate relief should be formulated by the Census Bureau in a manner that safeguards the absent interested parties. A.41-42. The Bureau, and not the district court, must initially bear the responsibility of formulating equitable relief if such relief is warranted.

In sum, I find the express language of 13 U.S.C. § 9(a), as well as its legislative history and case interpretation, to support a narrower privilege than that advanced by the appellant and accepted by the majority. However, the district court's disclosure order did not meet the command of the statute. I also find that the preclusion order was not tailored to the conduct or information at issue. Accordingly, I concur in the judgment.