Federal agencies. For the long-range development of sound statistical information on social processes, however, it is essential.

The first step in achieving this situation is the enactment of a clear legal status as "protected enclaves" for selected statistical and research agencies in the major departments, and for other clearly identified statistical and research units within other agencies. The enclave must be insulated from intervention and from unauthorized access to data. Employees must be subject to strict ethical standards established with respect to data handling and to penalties for voluntarily releasing identifiable data contrary to law. Once an individual business or other person has agreed to provide information for statistical purposes, there should be a mechanism for transferring identifiable data among agencies with a continued assurance of legal protection. At a minimum, this would require:

a.

A statement at the time of data collection about the general character of potential statistical uses;

b. A review agency such as the Office of Federal Statistical Policy and Standards, with oversight responsibilities for Federal statistics should be given the power to authorize transfers (OMB has restricted authority in the Federal Reports Act to authorize transfers among agencies under its jurisdiction);

c. A clear set of criteria that specifies when transfer of identifiable data would qualify as being of sufficient public interest to justify the transfer; and

d. A set of procedures to provide for removal of identifiers or destruction of the basic data files after the purposes of the transfer have been achieved.

4. Administrative data sets should be accessible to statisticians and researchers in "protected enclaves" for some statistical uses unrelated to the purposes of the original data collection.

In certain cases, as mentioned above, agencies need to use administrative records for establishing sample frames, for verifying the total universe characteristics, and for other statistical or research. purposes. Identifiable data extracted from administrative records for statistical purposes should be held confidential by the agency which receives them in the same manner that data collected directly from the respondent are held confidential. In essence, this suggests the creation of a "protected data set"

composed of those items derived from administrative sources for use in the "protected enclave." Thus, under this procedure, subpoena and other access to the identifiable data through the original administrative submission would not be restricted, but such access would not be permitted from the statistical or research data file. At the same time, the controlled exchange of data extracted from administrative records among statistical agencies would not be restricted further than the process defined in the third principle stated above would imply.

The above principles would establish statistical data as a special class of information. To summarize, it must be made clear from the outset in the laws which would implement these principles that: (a) statistical and research data may not be disclosed in identifiable form or used for determining the benefits, rights, and privileges of individuals or of businesses, and (b) the sole use of these data is for determining relationships and preparing reports or anonymous microdata sets which are statistical or research oriented in character. Such protection of these data would be uniquely strong. Once this protection is established, a controlled exchange of statistical data could appropriately be encouraged to improve the accuracy and comprehensiveness of the various measures employed, as well as to assure reduced costs of data collection and minimum reporting burden.

Further Developments Are Needed

In conjunction with the principles discussed above, further work is needed on alternative methods for protecting the identity of individual respondents or data sources. Several options exist for increasing the amount of detail of individual records which can be released to the public without identifying a particular data subject. The traditional means for reducing disclosures, of course, is the grouping of data into celIs which represent three or more respondents. The Census Bureau has developed disclosure analysis techniques which take other factors into account as well. Additional research should explore the use of camouflaged data elements in public use microdata records. One such method is random rounding, in which an error term, with zero mean and known variance, is added to the responses." The data would

16 M.S. Nargundkar, "Random-Rounding to Prevent Statistical Disclosure," Proceedings of the American Statistical Association Social Statistics Section, 1972, pp. 382-385.

Next Steps

The principles of confidentiality and the strategy for their implementation discussed in this chapter should be interpreted as overall requirements for achieving improved confidentiality for data used for statistical and research activities, while simultaneously enhancing the quality of the end product of these activities. There are many specific problems yet to be resolved, and the various activities. that are described above will all contribute significantly to this effort. Several agencies are engaged in examining their own needs for improved confidentiality laws and practices, and some have embarked on the difficult task of drafting possible legislation, even as the issues are being explored and answers are being sought.

The best approach to this problem is for each agency to study its own situation with respect to data confidentiality and to develop solutions which fit its unique problems within the general principles suggested above. As agencies begin to formulate their plans, the Office of Federal Statistical Policy and Standards will review their progress. This process will ensure that the legal proposals will serve the purposes of each agency while enhancing the capability of the U.S. Federal Statistical System, when viewed as a whole, to address the important informational needs of the day. We must first demonstrate the ability to maintain statistical and research data as "protected data." Only as improved legal protection of confidentiality is secured throughout the system will the ability to engage in increased data sharing under controlled conditions to achieve those important informational goals be available to us.

Conclusion

The protection of confidentiality of statistical and research information and the appropriate balance to be achieved with the conduct of the Government's information gathering and analytical activities will continue to be the subject of concern and public debate. Consideration of the public's right to know and the burden on the public of Federal requests for information will also be part of this debate. The Privacy Act addresses some of these issues but leaves important questions unresolved. While one approach

Definition and Purpose of the Federal-State

Local Cooperative Systems

The Federal-State-Local Cooperative Statistical Systems include those federally initiated or sponsored statistical programs in which State and local agencies participate in the collection or compilation of nationally standardized statistics. The cooperative systems are undertaken for the mutual benefit of the participants, involve multiple States (and local jurisdictions), and contain data of a recurrent nature which are intended to have broad applicability.

Three factors have contributed historically to the development and implementation of the FederalState-Local Cooperative Statistical Systems:

1. The Congress through legislation and the executive agencies through program efforts have increasingly emphasized the responsibilities and roles of State and local agencies in the resolution of national, as well as State and local, problems.

2. The Federal Government has shown an increased interest in improving and extending statistical measures of the status of subnational economies; the composition and characteristics of the population; and the nature, scope and effects of programs designed to promote the general welfare of the Nation and its people. The recent trends toward decentralization, revenue sharing, and block grant funding have served to increase the need for comparable, consistent information on State and local areas, both for assessing relative need and for effectively targeting program resources.

3. Federal and State governments alike have recognized the need to work jointly on information collection and production activities to reduce respondent burden and to improve the scope, uniformity, quality, and timeliness of statistical data in the most economical and efficient manner.

More recently, added emphasis has been given to the cooperative statistical systems governmentwide

through the issuance of a Presidential directive which stated:

Agencies should establish cooperative data collection programs with State and local governments wherever practical to eliminate duplicative reporting of similar data by more than one level of government, so long as no legal prohibition against this exists.'

Nature of the Cooperative Systems

Generally, the Federal-State-Local Cooperative Statistical Systems include the specification of federally required information and the definition of data elements included in the programs. At varying levels, attempts are made to ensure the quality and comparability of data within and across the States. By definition, each of the cooperative programs involves coordination and joint participation of the Federal and State agencies, as well as involvement of other relevant data producers.

Outnumbering these common features are the cooperative systems' characteristics which vary from program to program. For example, some of the cooperative systems in existence today involve all States equally (e.g. the Current Employment Statistics Program of the Bureau of Labor Statistics), while others, such as the Cooperative Health Statistics System, involve the States at variable levels. Reasons for the variability in State participation in the several cooperative systems include inadequate resources at the Federal level to support nationwide participation, inability of some States to meet the requirements of joint participation, and the existence of previously established systems of data collection and compilation involving parties other than the State agencies.

Likewise, the scope of the several cooperative systems varies substantially. Some programs, such as the Census Federal/State Cooperative Program for Local Population Estimates, deal with a single,

'Jimmy Carter, "Cutting Federal Red Tape for State and Local Grant Recipients," Memorandum for the Heads of Executive Departments and Agencies, September 9, 1977.