Click for next page ( 16


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 15
CONCLUSION The Committee believes that evaluation of government programs is very much in the public interest and should not endanger the privacy of individuals who give information about themselves. It will not be easy to work out procedures and legal arrangements that balance the conflicting objectives of fostering accountability in government and protecting individual privacy. Special efforts must be made in two areas: (1) protecting evaluation research data from acci- dental release or malicious misuse by unauthorized persons; (2) protecting people who give information about themselves in the course of evaluation re- search or social experimentation from having that information used in a court or legislative inquiry. The Committee believes that these efforts should re- ceive high-priority attention in federal agencies, the Congress, and the re- search community. It is the Committee's hope that this Report and its appen- dices will contribute to informed debate on these issues. 15

OCR for page 15

OCR for page 15
APPENDIX A CONFIDENTIALITY-PRESERVING MODES OF ACCESS TO FILES AND TO INTERFILE EXCHANGE FOR USEFUL STATISTICAL ANALYSIS Donald T. Campbell, Robert F. Boruch, Richard D. Schwartz, and Joseph Steinberg* *The authors are indebted to Professor Tore Dalenius for a careful review of an earlier draft of this manuscript. The efforts of Donald T. Campbell and Robert F. Boruch have been supported in part by grants from the Russell Sage Foundation, by National Science Foundation Grant GSOC-7103704, and by National Institute of Education Grant NIE-C-74-0115. Portions of an earlier draft of this report have been used 'iy Boruch and Campbell as a basis for pages 261-269 of H. W. Riecken, R. F. Boruch, D. T. Campbell, T. K. Glennan, J Pratt, A. Rees, and W. Williams, Social experimentation: A method for planning and evaluating social intervention, New York: Academic Press, 1974. The resulting duplica- tion is retained with the permission of Academic Press. The effort by Joseph Steinberg is in his private capacity; no official support or endorsement is intended nor should be inferred.

OCR for page 15

OCR for page 15
CONFIDENTIALITY-PRESERVING MODES OF ACCESS TO FILES AND TO INTERFILE EXCHANGE FOR USEFUL STATISTICAL ANALYSIS INTRODUCTORY OVERVIEW Some examples: (1) Achievement test scores and parental interviews from a Head Start evaluation project are subjected to statistical reanalyses by other scientists, who correct and extend the original findings. (2) The Bureau of the Census releases a one percent sample of interview data for soci- ologists and geographers to use in scientific studies. (3) The success of graduates of a job training program is ascertained by using Social Security Administration records on their later earnings and unemployment insurance payments. These exemplify important uses of data archives, potentially of great value in improving our ability to solve social problems and to evaluate the usefulness of ameliorative programs. Such research utilizations of data files should be greatly increased if we are to optimally evaluate new social programs in the future. At the same time, if improperly done, such uses could violate the promises of confidentiality and increase the individual's risk of black- mail, invidious gossip, or arrest. This is the problem discussed in this paper. At present, there is great concern about invasion of privacy and confi- dentiality and about the threat to individual freedom represented by data banks. Such concerns are currently much stronger than are demands for increased objectivity in the evaluation of governmental programs. It is our belief, detailed in this paper, that both concerns can be reconciled—that data archive use for program evaluation can be achieved without increasing the dangers of invasion of privacy. But we also believe that the means of such reconciliation are too little recognized and that there is a real danger of ill-considered solutions to the privacy problem that would needlessly preclude the use of archives in program evaluation. At its worst, the privacy issue becomes a rationalization for evading meaningful program evaluation. The present analysis starts out by assuming the existence of administra- tive records and of archived statistical research data and then asks how both of these can be further used to generate non-individual statistical products without increasing the risks to individuals that are already implicit in these A-l

OCR for page 15
A-2 existing files.1 Such a focus is tangential to many of the main concerns in the discussion of the threats of data banks. While we favor a number of the current proposals for the reform of government data archiving, both present practice and these reforms are compatible with the recommendations in this paper. Thus, although not the focus of this paper, we join others in a number of recommendations on the management of data archives: 1. Administrative data collectors and evaluation researchers should refrain from collecting sensitive personal data not directly relevant to the government's legitimate concerns appropriate to the transaction at hand. 2. Identifying information should not be collected at all; that is, respondents should be kept anonymous from the beginning where this is compatible with the purposes of the evaluation. 3. For the purposes here represented, there would be no loss and perhaps some gain if individuals were given a copy of their own data at the time it is filed, with the opportunity to correct it if necessary and with the right to future access. (Much of this is already achieved for Social Security Administra- tion and Internal Revenue Service files.) 4. In regard to the desirability of restricting the uses of data to those which the individual anticipated and was agreeable to when providing the data, it would probably be desirable if the forms used for data collection (e.g., income tax returns) announced that the data would also be used for statistical summaries in which the individual would be unidentifiable. 5. Restrictions would be desirable to prevent any secondary use of archived data for "intelligence" or investigatory purposes, that is, for actions or descriptions targeted on the individual; all secondary uses of data would be restricted to statistical products in which individuals were unidentifiable. 6. For statistical research data files, individual identi- fiers should be replaced with code numbers for data processing ^^Statistical" has unfortunately acquired two meanings in recent privacy discussions, both of which we want to represent here. The Privacy Act of 1974 (Public Law 93-579 8552a(a)(6)) uses "statistical" to designate data collected originally for research rather than administrative purposes: "the term 'statis- tical record' means a record in a system of records maintained for statistical research or reporting purposes only and not used in whole or in part in making any determination about an identifiable individual ..." In the second usage, "statistical" designates averages, percentages, standard deviations, correlation coefficients, etc., created from the individ- ual data existing in either administrative files or research-generated records. Since these statistical products summarize individual data from many persons, their publication can usually be done in ways protecting individual privacy, by means discussed in this paper.

OCR for page 15
A-3 and computer storage, and the code key kept under tight security. For administrative data files, this may not always be feasible, but certainly should be required for any computer memory storage on a time-sharing basis. 7. Formal rules and guidelines should be promulgated to guarantee high standards of confidentiality and security management on the part of all data file staff. 8. For the data uses we advocate, a unified national data bank is not required. Such a data bank is feared because it multi- plies the power of a corrupt employee to blackmail or of the govern- ment to police the individuals on whom data are recorded. On the other hand, there are some recommended reforms of public record- keeping that would preclude the uses here advocated and that we regard as both needless and contrary to the public interest: 1. Abolition of the use of Social Security numbers for all but Social Security Administration files. Our recommendation is quite the opposite, namely, that Social Security numbers be recorded where possible. The abolition recommendation was designed to pre- clude merging files into larger data banks. Through the "mutually insulated file linkage" described in this paper, some types of file linkage can be achieved without merging, i.e., in a manner that pre- vents either file from acquiring identifiable individual information from the other file.^ This procedure does require, however, that common identifiers, such as names and Social Security numbers, exist in each file. We believe that were this procedure adopted, the reasons for the suggested prohibition on Social Security numbers would be eliminated. 2. The destruction of personal data files after a specified period, say five or ten years. Many social innovations call for longer-term statistical follow-ups that would be precluded by such a rule. 3. The requirement of a specific, separate permission state- ment and explicit informed consent for each separate statistical research use of an administrative file. We regard this recommenda- tion as contrary to the public interest as well as needless when no indivdually identifiable data are being released from the file, as in procedures described below. 4. Elimination of all individually identified files in any form. Needless to say, we regard this as impossible for adminis- trative files in their administrative use, unneeded for protection of individuals since this can adequately be done in other ways, ^"Merging files," means the complete combining of files so that all of the individual data from one file gets transferred into the other or into a new combined file in an individually identifiable manner. "File linkage" is the more inclusive term, covering not only file merging (the most complete form of linkage), but also other ways of relating the information in one file to the information in another. In the second part of this paper we discuss four modes of file linkage, short of merging, aimed at preserving privacy of individual records.

OCR for page 15
A-4 and seriously detrimental to our capacity for program evaluation. The requirements for safeguarding confidentiality are, of course, the responsibility of the administrators of each specific data file. The pre- cautions and procedures necessary will vary from setting to setting. This paper focuses on the most cautious and conservative approaches, not to recom- mend that they be required in all settings, but rather to emphasize that, for even the most sensitive settings, there are safe modes of access that will permit important statistical analyses. These conservative approaches do set limits to the degree of refinement possible in the statistical analyses, and, to avoid such costs, one should not use a more conservative approach than is called for by the requirements of the situation. The major sections of this paper discuss specific procedures under two main headings. The first is designated intrafile analysis by outsiders. In this category, all of the statistical analyses under consideration are to be done within a single data file. This category is exemplified by the U.S. Census Bureau's Decennial Census public use one percent and 0.1 percent sam- ples which are released for social science research purposes, or by the Social Security Administration's one percent Continuous Work History Sample (CWHS). In the experience of the Office of Economic Opportunity (OEO), such uses are encountered in the release of data for reanalysis from the evaluations of Head Start, Performance Contracting, and the New Jersey Negative Income Tax Experi- ment (NJNITE).3 Six classes of procedures are considered for intrafile anal- ysis by outsiders, concluding with the following recommendations: Deletion of known identifiers (name, Social Security number) is insufficient unless also accompanied by restrictions in number and refinement of data on variables that are publicly available elsewhere, or unless accompanied by error inoculation on public variables, especially by additive normal error; Microaggregated release is acceptable, albeit statistically costly; Best of all is in-file capacity as a public utility to run outsiders' statistical analyses, accompanied by randomized round- ing of frequency tabulations to prevent disclosure through compari- sons of sets of results. The secondary category of utilization considered is interfile exchange of confidential data, where one or both files are confidential and the objective is to relate variables across files in statistical analyses. (Most conserva- tively, such exchanges can be done without merging files, i.e., with neither file acquiring the other's confidential data.) Examples of this occur when Social Security Administration data are used to evaluate Job Corps training official title of this project, as evidenced by its final report, is the New Jersey Graduated Work Incentive Experiment (Watts and Rees, 1973). See also the special issue of The Journal of Human Resources, Spring, 1974. However, since it is much better known as the New Jersey Negative Income Tax Experiment, and since a number of publications will continue to use that name (e.g., Kershaw, 1972; 1974), it has seemed most convenient to retain that title. Because of the origins of COFAER and the original requests to it, the NJNITE is a major source of illustrations, hypothetical for the most part, in what follows.

OCR for page 15
A-5 programs, when the Census Current Population Survey data are related to IRS Statistics of Income data derived from income tax reports, or, conjecturally, if NJNITE interview data were to be related to withholding tax information or to PICA earnings. For such purposes, this paper emphasizes the most conserva- tive of several procedures: mutually insulated file linkage with random dele- tion of one respondent from each list. Link file brokerage, a widely discussed procedure, is considerably less conservative, but may be a useful strategy if adequate protection for confidentiality can be assured. There are also settings with confidentiality safeguards and legal protection from subpoena in which the still less conservative approach of direct file merging provides reasonable safeguards. Full merging has the advantage of preserving full statistical information on all relationships among all variables for any analy- sis or reanalysis. Where this approach is used, it would be very important for the merged file to have in-file capacity to run outsiders' statistical analyses (as discussed more extensively in the next section). INTRAFILE ANALYSIS BY OUTSIDERS Deletion of Identifiers It is customary in releasing data for reanalysis to delete names, Social Security numbers, and addresses from the data on individuals. In some set- tings this may provide sufficient protection in that it may not increase the respondent's loss of privacy or increase the risks of breach of confidentiality. In other settings, deletion of identifiers is an insufficient safeguard. Two features seem crucial: the number of items of information on each person and the availability of those items on public lists with names attached. For ex- ample, deletion of identifiers might be sufficient for a 0.1 percent sample of the 1970 Census because of the extremely scattered nature of the sample and the absence of parallel lists. However, even in this case, if census tract, age, and specialty are given for a low-frequency, visibly listed profession, such as M.D., individual identification could frequently be made and the other information on the record thereby identified with a specific person, thus mak- ing it possible for a corrupt user to infringe upon the M.D.'s privacy (Hansen, 1971). Where the research population is compact and where some of the variables are conveniently recorded with names on public or semi-public lists (here designated as public variables), the deletion of identifiers is less adequate. Thus, for a study conducted within a single school, even the date and place of birth are usually sufficient to reveal names. (Specific birth date in combination with birth place probably should always be treated as a personal identifier.) In the case of the NJNITE, data tapes are now being released to outside users through the Data Center of The Institute for Research on Poverty of the University of Wisconsin. It was decided in this case that a thorough deletion of identifiers provided adequate protection. This deletion covered names, addresses, exact birth places and birth dates, Social Security numbers, and, in addition, the names of doctors, teachers, etc. These data have been de-

OCR for page 15
A-6 leted not only from the released tapes, but also from the original interviews preserved in microfilm. Crude Report Categories for, and Restriction of, Public Variables For public variables in the confidential file (variables that are readily available elsewhere with names attached), cruder report categories should be used in the data released, to the level needed to prevent disclosure: e.g., county rather than census tract, year of birth rather than day or month, pro- fession but not specialty within profession, etc. For variables unique to the research project (unique variables), which therefore do not exist on other lists with names attached, this precaution is not necessary. Thus, for a multi-item attitude test or for an achievement test, individual item responses and exact total scores can be made available without jeopardizing confidential- ity. (Such data probably should be made available because of their relevance to the estimation of error for use in generating alternate statistical adjust- ments, an issue of ever-increasing concern.) Even with crude report categories on such public variables as geographic area, places and years of schooling, age, profession, etc., if there are enough such variables, combinations emerge in which only one or two persons occur and discovery of individual identity becomes possible. Thus, there should be a minimum lower bound restriction on the cell sizes of the full combination of public variables. For example, the rule might be adopted that there should be no combination of public variables yielding a frequency less than three persons. (Fellegi and Phillips, 1974; Hansen, 1971; Hoffman and Miller, 1970). Receding of variables using still cruder report categories, or complete deletion of public variables, should be done until the chosen criterion is achieved. Before a criterion is chosen, tests of the anonymity- breaking potentials of various criteria should be tried out using actual bodies of data and publicly available lists. These restrictions are obviously at the cost of some potential statistical analyses, particularly if some public variables have to be eliminated entirely. It might be thought that this could be avoided by releasing to each given user only some of the public variables, permitting the user to specify which public variables were of highest priority to the particular work. This strategy would suffice if there were only one user on one occasion or if users could be kept from sharing their data sets. But this seems impossible to guarantee, and such sharing would permit discovery of identity. For example, if user Alpha received public variables Pl and 1?2 plus all unique variables Uj through Un, while user Beta received public variables P3 and P4 plus all unique variabTes Ul through Un, they could easily employ the shared unique variables to achieve perfect matcTfes and thus generate a complete, merged deck with Pj_, ?2, PS> and P^ on each person. This full set of public variables might then be sufficient to identify individuals with the help of public lists. Random Subsample Release The last-mentioned problem of multiple users sharing differentially de- leted data sets and thereby gaining increased ability to disclose individual identities can be greatly attenuated by providing each user with a different

OCR for page 15
A-7 randomly selected subset of the data. This approach is obviously most usable for files containing a large number of individuals. It provides most protec- tion where the sample is a small portion of the total available population and when the public and unique variables are few enough in number and crude enough in categories so that many persons end up with identical patterns and individ- ual identification is precluded. Microaggregation Feige and Watts (Feige and Watts, 1970; Watts, 1972) have developed a technique of microaggregation for the release of census data on firms, as a substitute for the release of individual data. This approach has been recom- mended as a general approach to the release of confidential data. The idea of microaggregation is to create many synthetic average persons and to release the data on these rather than on individuals. Thus, instead of releasing in- dividual data on the 1,200 participants of the NJNITE, as has been done, hypo- thetically one might group the data into 240 sets of five each and release average data on every variable for each set (probably with within-set variance as well as mean). Outside users could then do all of their secondary analyses on these 240 synthetic persons. Feige, Watts, and their colleagues have done such analyses on Federal Researve Board data on banks and have been able to compare microaggregate analyses with individual data analyses. Their conclusion is clear that such microaggregation is much more useful than no release of data at all. It re- sults in a loss of statistical efficiency, but does not necessarily bias the statistical estimates. For most conceivable grouping variables, anonymity and confidentiality are preserved at the individual level. The actual acceptable basis for microaggregation must be thought through in detail for every specific body of data. The following preliminary sugges- tions hypothetically illustrate the problem for the NJNITE. For most purposes, aggregation should probably keep intact the experimental design,' that is, aggregation should be done within treatments and in comparability across treat- ments. (There would be many more such aggregates in the 640-person control group than in one of the experimental groups that average around 80 persons, but each one of the experimental group aggregates should be identifiable as parallel to certain control aggregates, etc.) In the Feige and Watts discus- sion, local region is a preferred basis of aggregation. In the NJNITE, cities differ considerably in time of initiation of the experiment and in attrition rates. Therefore, city should be used as a basis for aggregation, and possibly region within city. Ethnicity would be wanted for some uses, but with an initial sampling model assigning as few as 16 cases per experimental treatment per city, comparability would be hard to maintain for any variable not blocked on initially. (Aggregating by initial blocking seems a reasonable rule in most social experiments.) If complete data cases are aggregated separately from those lost through attrition, comparability is jeopardized because the attrition in the NJNITE is differential, being greatest in the control and low payment treatments. Possibly, attrition could be handled by reading out for each variable not only mean and variance but also the number of persons on which data were computed, basing the mean on those cases providing data.

OCR for page 15
B-79 "Statutes: Existing and Proposed" at B-19ff, supra for a discussion of such statutes: 21 U.S.C. § 1175 (Supp. 1972), N.Y. CIVIL RIGHTS LAW § 79-j (McKinney 1972), and MD. CODE ANN. art. 35 § 101 (1971). 213. Contra, a number of proposed statutes for journalists before the 93rd Congress define the term "information" so as to preclude limiting coverage to specific ways in which information is stored. E.g., S. DOC. NO. 158, 93rd Cong., 1st Sess. (1973); H.R. DOC. NO. 1813 § 4(3)-(5); H.R. DOC. NO. 2187, § 4(3)-(5); H.R. DOC. NO. 2200 § 4(3)-(5), 93rd Cong., 1st Sess. (1973). 214. Proposed Statute, § 5(5). 215. See following discussion rejecting the absolute-qualified dichotomy often used to describe privilege conferring statutes. 216. See generally McCormick, EVIDENCE (1954) [hereinafter referred to as McCORMICK], § 91 at 187-91 (presence of a third party); § 95 at 194-200 (consultation in furtherance of Crime of Fraud). 217. 68 Misc. 2d 355, 326 N.Y.S. 2d 434 (1971). See also Baker v. F & F Investment. 339 F. Supp. 942 (S.D.N.Y. 1972), where a federal district court stated that in interpreting the New York newsman privilege statute, N.Y. CIVIL RIGHTS LAW § 79-h, no exceptions were consistent with the policy of the Illinois newsman statute. The Illinois statute provides an exception to the privilege in actions for defamation and that the privilege may be denied if all other sources of information have been exhausted and disclosure is essential to the public interest. 218. N.Y. CIVIL RIGHTS LAW § 79-h (McKinney 1972). 219. 21 U.S.C. § 1175 (Supp. 1972). 220. 21 U.S.C. § 1175(b)(c) (Supp. 1972). 221. E.g., N.M. STAT. ANN. § 20-1-12.1 (Supp. 1970) considers the following factors in determining whether or not a privilege should be granted: (1) the nature of the proceeding; (2) the merits of the claim or defense; (3) the adequacy of the remedy otherwise available, if any; (4) the possibility of establishing by other means that which it is alleged the source requested will tend to prove. 222. See "Current Judicial Protection" at B-16ff. supra, 223. See NEJELSKI and LERMAN, supra note 1 at 1087-89. 224. The power of the grand jury to compel testimony has been consistently recognized by the U.S. Supreme Court. See Branzburg v. Hayes, 408 U.S. 665, 688 (1972), citing Blair v. United States, 250 U.S. 273 (1919), for the proposition that the grand jury's authority to subpoena witnesses is historic, 250 U.S. at 279-81. The Branzburg court regarded the power as "essential." 408 U.S. at 688. 225. The power of congressional committees to compel testimony was recognized by the U.S. Supreme Court in Watkins v. United States, 354 U.S. 178 (1957). Furthermore, the power can be enforced by Congress pursuant to the contempt of Congress statute. 2 U.S.C. § 192 (1970). 226. On the federal level, Congress has passed legislation granting the power of compulsory process to a variety of administrative agencies. E.g., 15 U.S.C. § 1312(a) (1970), (Anti-Trust Division of the Department of Justice); 47 U.S.C. § 409(e) (1970), (F.C.C.); 29 U.S.C. § 161 (1970), (N.L.R.B.). 227. Defendants in criminal trials have a constitutional right to compel

OCR for page 15
B-80 testimony, U.S. CONST. amend. VI; FED. R. CRIM. P. 17(e) (1973). Prose- cutors, as well, have the power to compel testimony, FED. R. CRIM. P. 17 (e). 228. E.g., FED. R. CIV. P. 4(e) provides civil litigants in federal court the power of compulsory process, though limited by a territorial restriction of 100 miles. However, the power of civil litigants to compel testimony is not constitutionally mandated. 229. E.g., the Drug Abuse Prevention and Control Act of 1970 and the Compre- hensive Alcohol and Alcoholism Prevention, Treatment and Rehabilitation Act of 1970 confer protection to researchers in "civil, criminal, legis- lative or other proceedings;" the A.C.L.U. Privacy Committee Memorandum on a Testimonial Privilege for Researchers (April 6, 1973, mimeographed) also confers protection to a researcher "in any court, legislative or other body having the contempt power." 230. E.g., N.Y. CIVIL RIGHTS LAW § 79-j (McKinney Supp. 1972) makes records of the multi-state information system for psychiatric patients "not sub- ject to examination in the court or by agencies of this [N.Y.] state." In the news area, the "general" statutes do not mention specific pro- ceedings, only specifying the dimensions of the privilege. E.g., MD. CODE ANN. art. 35 § 2 (1971); MICH. STAT. ANN. § 38.945(1) (1954); N.J. STAT. ANN. 2A:84A-21 (Supp. 1969). The New Jersey court has held that the New Jersey statute can be asserted at discovery, as well as at trial, despite the fact that specific proceedings are not listed. Beecroft v. Point Pleasant Printing & Publishing Co., 82 N.J. Super. 259, 197 A. 2d 416 (1964). 231. Professor Vince Blasi, drafter for the National Conference of Commis- sioners on Uniform State Laws of the Uniform Information Source Protec- tion Act, is the main proponent of this approach in the news area. A uniform law has not yet been approved by the Commissioners on Uniform State Laws. The statute is still in draft form, the latest draft being that of April 1, 1973. 232. The April 1973 draft defines adjudicative proceedings as "any proceed- ing in which a judicially enforceable final judgment may be rendered, but does not include any proceeding of a grand jury or any other accusa- torial or investigative proceeding." See generally Hearings on H.R. 717, Before Subcomm. No. 3 of the House Comm. on the Judiciary, 93rd Cong., 1st Sess. 1611 (1973) (statement of Vince Blasi) [hereinafter referred to as STATEMENT OF BLASI]. 233. Id. at 142-43. 234. Id. at 143. 235. Id. at 142-43. 236. Testimony of Professor Vince Blasi at ABA Committee Newsman Shield Law Hearings, June 10, 1973. Rather than advocating specific qualifications, Professor Blasi proposed a flexible standard by which a court would balance the need for the contents of the source communication with the detriment to the journalist. 237. Proposed Statute, § 2. 238. ALAS. COMP. LAWS ANN. § 09.24.160(b) (Supp. 1970). The court may deny the privilege if it "results in a miscarriage of justice." 239. S. DOC. NO. 1311, 92nd Cong., 1st Sess. (1972), provides an exception for threat to human life or espionage. See also H.R. DOC. NO. 2101,

OCR for page 15
B-81 93rd Cong., 1st Sess. § 4 (1973). Another group of proposed statutes provides for an exception to the privilege if the following tripartite test is met: (1) there is a probable cause that the protected person has information clearly relevant to a specific probable violation; (2) the information cannot be obtained by alternative means; and (3) there is a compelling and overriding national interest in the information. E.g., H.R. DOC. NO. 265, 93rd Cong., 1st Sess. § 3 (1973). This reflects the test suggested by a number of the petitioners in the Branzburg v. Hayes litigation as well as the view taken by Justice Stewart dissenting in Branzburg, 408 U.S. at 743. 240. 408 U.S. at 690-91. 241. Testimony of Robert Dixon, Senate Hearings on Newsmen's Privilege, supra note 44. 242. Thirteen of the 19 state journalist statutes do not set forth any qualifi- cations for law enforcement purposes. None of the researcher statutes specifically make an exception for law enforcement purposes. 243. See "Case Histories" at B-12ff. supra. 244. E.g., H.R. DOC. NO. 3369, 93rd Cong., 1st Sess. (1973). See note 239 supra. 245. Exceptions to the confidentiality provisions of the Social Security Act are made in cases of national security and for information concerning aliens as requested by the Attorney General under § 209(c) of the Immigra- tion and Nationality Act of 1952, 8 U.S.C. 1360(c) (1970). 246. See Korematsu v. United States. 323 U.S. 214 (1945), in which the Court justified the internment of Japanese aliens on the West Coast for national security purposes. 247. Draft Report of the President's Commission on Federal Statistics, U.S. Government Printing Office, Wash., D.C. Vol. II (1971). 248. See BLASI REPORT, note 19 supra at 247. 249. E.g., drug use, possession, and sale; prostitution; gambling; sexual deviance; pornography use, possession and sale; and generally all crimes that principally harm the individual involved rather than society at large. Many prominent criminologists, in fact, advocate decriminaliz- ing this type of activity. See generally BLASI REPORT, note 19 supra, citing H. Packer, THE LIMITS OF CRIMINAL SANCTIONS (1969); N. Morris & G. Hawkins, THE HONEST POLITICIAN'S GUIDE TO CRIME CONTROL (1970); J. Kaplan, MARIJUANA—THE NEW PROHIBITION (1971). 250. It is convenient to include in "political corruption" criminal activity that involves recipients of public benefit programs usurping their rights under the program, information likely to be revealed in the course of research to evaluate the public benefit programs. 251. Letter to the authors from George Van Hoomissen, March 3, 1971, quoted in NEJELSKI and LERMAN, supra note 1 at 1101 n. 64. Cf. House Hearings, supra note 42 at 453 (testimony of Hon. William Cahn, District Attorney for Nassau County, N.Y., and immediate past president of the National District Attorneys Association, testifying on newsman's privilege). But whatever the extent of the suppressive effect, [from compelling newsman to testify] if the content of news and its depth and quality are thus diminished or impaired in the short range interest of law enforce- ment, the ultimate result may be long range disadvantage

OCR for page 15
B-82 to law enforcement and social health. 252. H.R. DOC. NO. 3369, 93rd Cong., 1st Sess. § 7(c)(3) (1973), provides an exception to the privilege in proceedings involving "murder, forcible rape, aggravated assault, kidnapping, and airline hijacking or when a breach of national security has been established . . . In no case, how- ever, shall the application [for an exception] be granted where the crime at issue is corruption or malfeasance in office. " See also Hearings on Newsmen's Privilege, note 18 supra (testimony of William Cahn) at 453 and BLASI REPORT, note 19 supra at 251. 253. The researcher studying criminal behavior in this way may use unethical practices to elicit incriminating responses from subjects, for example, by paying the subject for candid responses. Because the statute is de- signed in part to protect the subject's privacy interest, it should serve to offset the unethical behavior of the researcher in eliciting the information by allowing the researcher to resist compelled disclosure of the information. 254. See "Case Histories" at B-12ff, supra, 255. 42 U.S.C. § 242a(a) (1970); 21 U.S.C. § 872(c) (1970). 256. Knowledge of future drug transactions is, as well, knowledge of crimes to be committed in the future. A number of old cases limited the future crimes exception of the attorney-client privilege to intended crimes involving "malum in se" or moral turpitude. See McCORMICK, note 216 supra, § 199 at 200 n. 56 citing Bank of Utica v. Mesereau, 3 Barb. On. 528, 598 (1848); Hughes v. Boone, 102 N.C. 137, 9 S.E. 286, 292 (1899). Modern authorities, however, view such a limitation as inappropriate because the exception to the privilege merely has the consequence of opening the door to the evidence at trial. McCORMICK § 199 at 201. 257. But consider that the common law historically recognized a duty to report felonies. It prescribed criminal liability for failure to do so under the crime of "misprison of a felony." Convictions for such an offense, however, were rare. F. Perkins, CRIMINAL LAW, 512 (2d ed. 1969). While today it is questionable whether a citizen has a legal affirmative duty to report a crime, the Supreme Court has stated that "concealment of crime and agreements to do so are not looked upon with favor." Branzburg, 408 U.S. at 697. Morality may, however, dictate that the researcher re- veal future crimes to law enforcement authorities. This may be done whether or not a privilege exists. 258. Cf. the psychotherapist- or psychiatrist-patient privilege, argued to be analogous to the researcher-subject privilege, for both depend on develop- ing a working relationship with the respondent. NEJELSKI and LERMAN, supra note 1 at 1146. State legislatures have not agreed on whether to divest this privilege in the context of future crimes. The Connecticut legislature consciously did not provide a future crimes exception. Goldstein & Katz, Psychiatrist-Patient Privilege: The GAP Proposal and the Connecticut Statute. 118 AM. J. PSYCHIATRY 733, 738-39 (1962). Contra, MASS. ANN. LAWS Ch. 233 § 20B (1969) making an exception to the psychotherapist-patient privilege where the psychotherapist believes a present "threat of imminently dangerous activity" by the patient towards himself or herself or others. 259. The Sixth Amendment compulsory process applies equally in federal and state proceedings. Washington v. Texas, 388 U.S. 14 (1967).

OCR for page 15
B-83 260. 388 U.S. at 19. 261. See generally J. Brennan, The Criminal Prosecution: Sporting Event or Quest for Truth? 1963 WASH. U.L.Q. 279, 285-88 (1963). 262. Rule 17 (d) of the Federal Rules of Criminal Procedure was amended in 1966 to make it less difficult for poor defendants to subpoena witnesses, pro- viding for court payment of subpoena fees upon defendant's showing of "inability to pay" rather than "indigency." The revisions also make the defendant's application for subpoena fees an ex parte proceeding so that the application would not serve to provide strategic information for the prosecution. The 1966 revision also decreases the defendant's burden of proof for the issuance of a subpoena from materiality of the testimony to necessity to prepare an adequate defense. Finally, upon an adequate showing by the defense the new rule provides that the court "shall" rather than "may" issue a subpoena. 263. Branzburg v. Hayes, 408 U.S. at 705. 264. State v. Groppi. 164 N.W. 2d 266, 271, 41 Wis. 2d 312 (1969), rev'd on other grounds, 400 U.S. 505 (1971). The court rejected the attempt of Father Groppi to subpoena witnesses. 265. United States v. Beye, 445 F. 2d 1037 (9th Cir. 1971). 266. 354 F. Supp. 208 (D.D.C. 1972). See "Current Judicial Protection" at B-16ff. supra and note 107 supra. 267. 354 F. Supp. at 216. 268. Id. at 216. 269. Id. at 215. 270. Adopted from H.R. DOC. NO. 2651, 93rd Cong., 1st Sess. § 3 (1973). 271. Branzburg v. Hayes, 408 U.S. at 690. 272. See "Current Judicial Protection" at B-16ff. supra discussing Baker v. F & F Investment and Democratic National Committee v. McCord, note 40 supra and accompanying text. 273. Contra, see "Statutes: Existing and Proposed" at B-19ff. supra, which pro- vide for judicial balancing. These statutes make it possible for courts on a case-by-case method to accommodate law enforcement interests by divesting the privilege when necessary. 274. See Hearings on Newsman's Privilege, note 44 supra (testimony of Sanford Smith and Arthur B. Hanson) at 262, stating that about half of the quali- fied journalist privilege bills presently before the 93rd Congress provide such an exception. 275. E.g., Minnesota Free Flow of Information Act. 276. 376 U.S. 254 (1964). 277. In ordinary libel cases, the defendant has the burden of proof to show the truth of the published story. The defendant, therefore, has the choice of either revealing the source or protecting the source and thereby risking an adverse judgment if truth can be proven in no other way. See Note, Reporters and their Sources: The Constitutional Right to a Con- fidential Relationship, 80 YALE L. J. 317 (1970). 278. 376 U.S. at 279-80. The Court, thus, shifted the burden of proof as well as the burden of coming forward to the plaintiff in libel actions. Defendants in libel actions under the New York Times rule are stripped of the traditional presumption of malice against them as well as the neces- sity of proving an affirmative defense of truth. This was done to foster the First Amendment interest in brisk public discourse.

OCR for page 15
B-84 279. 390 U.S. 727 (1968). 280. 390 U.S at 731. 281. "A defamatory communication usually has been defined as one which tends to hold the plaintiff up to hatred, contempt or ridicule or to cause him to be shunned or ridiculed." Prosser, LAW OF TORTS (4th ed. 1971) [here- inafter referred to as PROSSER] § 111 at 739. 282. Recoveries against the press, however, have been made. See Goldwater v. Ginzberg, 414 F. 2d 324 (2d Cir. 1969), cert• denied, 396 U.S. 1049 (1970), reh. denied, 397 U.S. 978 (1970). 283. In addition to any living person, business entities can also be defamed. While a corporation, a partnership, etc., has no personal reputation, it does have "standing" and "prestige" in the business community, and language which disparages its reputation may be actionable. PROSSER, supra note 281, § 111 at 745. 284. A communication to a third party is essential for an action for defama- tion. The technical term for this is "publication" though there is no requirement that the material be published in the conventional sense. PROSSER, supra note 281, § 113 at 766. 285. 17 Cal. App. 3d 621, 95 Cal. Rptr. 175 (1971). See also B. Thorton, White v. State of California and Code Sections 11120-11127: A Pre- Computer Privacy Case and the Legislative "Answer." 47 L.A. BAR J. 320 (1972). 286. CA. CIV. CODE § 47 (1954). 287. Actual malice is defined as "that state of mind arising from hatred or ill will toward the plaintiff." CA. CIV. CODE § 48(a)(4) (1954). 288. A number of absolute privileges have been established immunizing certain communications from any defamation action. E.g., statements made in the course of judicial proceedings; communications of executive officers of the government. See also Justice Black concurring in Rosenbloom v. Metromedia, 403 U.S. 29 (1971), stating that the First Amendment does not permit recovery of libel judgments against the news media under any cir- cumstances. 289. Perhaps the best generalization is that a communication is so protected if it was "fairly made by a person in the discharge of some public or private duty, whether legal or moral, or in the context of his own affairs in the matters where his interest is concerned." Toogood v. Sprying, 1 C.M. & R. 181, 149 Eng. Rep. 1044 (1834), quoted in PROSSER, supra note 281, § 115 at 786. For a general discussion of communications so protected, see sources cited in PROSSER, supra note 281, § 115 at 786 n. 75. 290. See PROSSER, supra note 281, § 115 at 791 n. 41. 291. Id. at 792. 292. E.g., Carey v. Hume. Senate Hearings on Newsmen's Privilege, note 44 supra. 293. See Cervantes v. Time Inc., 464 F. 2d 986 (8th Cir. 1972), cert• denied, 409 U.S. 1125 (1973). 294. Taken from Minnesota Free Flow of Information Act § 4. 295. See NEJELSKI and FINSTERBUSCH, note 1 supra at 14 in which a hypotheti- cal is furnished to point out the potential conflict. A teachers union sues the Office of Education in order to obtain the raw data from a study of contract teaching. The basic analysis had not been compiled by the Office of Education. The teacher's union argues that it should have an opportunity to evaluate the data compiled, for its members will be

OCR for page 15
B-85 adversely affected by results revealing that innovative teaching methods are better than those traditionally used. 296. 44 U.S.C. § 3501 et seq. (1970). 297. 5 U.S.C. § 552 (1970). 298. 14 U.S.C. § 3507 (1970). 299. 44 U.S.C. § 3508(b) (1970). 300. The Act requires that federal agencies make available to the public in the Federal Register information about their organizational structures, operating methods and procedures, rules, policies and interpretations. Also available to the public for copying are the agencies' final adjudi- catory decisions, precedents, staff manuals that affect the public, and indices for all matters required to be published. These requirements are conformed by a rule that agencies cannot rely on the above informa- tion unless it is indexed and published or made available to the public. 301. This exemption could be interpreted to prohibit disclosure of informa- tion made confidential by a federal researcher's privilege statute. On the other hand, the provision is more likely to be applied to exempt from the Act only data that are specifically exempt rather than data exempt by a general researcher's privilege. 302. 301 F. Supp. 796 (S.D.N.Y. 1969), dismissed as moot, 436 F. 2d 1363 (2d Cir. 1971). 303. The court reasoned that confidentiality was not assured because the requests stated that the information could not be withheld from other federal agencies. The other federal agencies could then release the information to the public. 304. Note that the disclosure obligation applies only to federal research. See Nathanson, Social Science and Administrative Law and the Informa- tion Act of 1966, 21 SOC. PROBS. 21 (1973), for a discussion of the availability of information to the researcher through the Freedom of Information Act. 305. See, e.g., Getman v. National Labor Relations Board, 450 F. 2d 670 (B.C. Cir. 1971). (Two labor law professors and a psychology professor were granted access to names and addresses of employees eligible to unite in a union election in the possession of the N.L.R.B. These researchers were studying the impact of campaign techniques on the union election.) 306. But see Davis, ADMINISTRATIVE LAW TREATISE (1972), concluding that even if the Freedom of Information Act were "fully obeyed, the information that the Act opens up that would otherwise be closed is minimal." 307. See Consumers Union of United States v. Veterans Administration, 301 F. Supp. at 806-08, stating that suits to compel disclosure under the Freedom of Information Act are in equity and, thus, the ultimate decision will be based on balancing the competing interests of privacy and the public's right to know. 308. Proposed Statute § 2(1) and (2). 309. Proposed Statute § 5(1). 310. But consider the procedures in obtaining a grant of confidentiality under the Drug Abuse Prevention and Treatment Act of 1970, which requires that researchers get approval from the Department of Health, Education, and Welfare that they are involved in bona fide drug research. This effec- tively means that researchers must sustain the burden of proof.

OCR for page 15
B-86 311. ARK. STAT. ANN. § 43-917 (1964) seems to place the burden upon the person seeking the information. 312. See "Coverage: People" at B-31ff. supra. 313. In order to divest the privilege the government must "(1) show that there is probable cause to believe that the newsman has information that is clearly relevant to a specific probable violation of the law; (2) demonstrate that the information sought cannot be obtained by alternative means less destructive of First Amendment rights; and (3) demonstrate a compelling and overriding interest in the information." 408 U.S. at 743. 314. 21 U.S.C. § 1175(b)(2) (Supp. 1974). 315. E.g., NEV. REV. STAT. § 49.385 (1971); N.J. STAT. ANN. 2A:84A-21 (Supp. 1969). 316. N.Y. CIVIL RIGHTS LAW § 79-h (McKinney 1971). 317. 69 Misc. 2d 256, 329 N.Y.S. 2d 291 (1972), aff'd. per curiam, 39 App. Div. 2d 837, 333 N.Y.S. 2d 299 (1972). In affirming the Supreme Court, the Appellate Division implied that the reporter's case would have been improved if he had shown that the manuscript contained information that was deleted from the article. 318. 342 N.Y.S. 2d 731, 41 A.D. 2d 687 (1973), app. dismissed, 344 N.Y.S. 2d 955, 32 N.Y. 2d 764 (1973). 319. See also Application of Howard 136 Cal. App. 2d 816, 289 P. 2d 537 (3rd Dist. 1955) and In re Taylor, 412 Pa. 32, 193 A. 2d 181 (1963). While both the California and the Pennsylvania journalist privilege statutes do not contain waiver provisions, both cases indicate that waiver will be implied when statements of the informer are publicly disclosed or published. 320. § 3: Waiver. 321. McCORMICK, supra note 216, § 93 at 194. 322. Id. at 218. 323. See note 216 supra. See also BLASI REPORT, note 19 supra at 285-89, discussing arguments for the subject to control the privilege but conclud- ing that the arguments in favor of the journalist controlling the privi- lege are stronger. Contra, Note, Reporters and Their Sources: The Con- stitutional Right to a Confidential Relationship, 80 YALE L.J. 317, 369-70 (1970). 324. Other parties, however, have a stake in the privilege. See "Introduction1 at B-lff. supra. 325. Accord, NEJELSKI and LERMAN, supra note 1 at 1145. 326. E.g., observational or survey techniques. 327. This provision does not, however, apply if the researcher knows the subject's identity but agrees not to reveal it. 328. Contra, People v. Wolf, supra note 317. Publication of a stay in the Village Voice was held to effectively waive the protection conferred in the New York privilege. 329. See, e.g., McCORMICK, note 216 supra, § 91 at 187-92 (attorney-client) and § 101 at 216-18 (doctor-patient).

OCR for page 15

OCR for page 15

OCR for page 15