(3) What proposals are there for a more integrated statistical system?

(4) Would further integration and particularly a national data center present a threat to personal privacy?

(5) Is the present administrative machinery adequate for the tasks at hand?

TO WHAT EXTENT IS THERE COORDINATION AND INTEGRATION?

The statistical programs of the Federal and State Governments, although good by historical or relative standards, are not sufficiently integrated or coordinated. Indeed, we do not have "a national statistical system" but rather several subsystems and some rather unrelated pieces. The present situation has evolved over many years, whereby new statistical series were initiated in response to the changing forces in our economy and the changing concerns of the time. The various bodies of data were collected with different objectives in mind and by different agencies of Government. Different objectives were often unavoidable and even desirable, but they should not be allowed to lead to duplication or the neglect of general high priority statistical needs, as has too often been true.

Government statistical programs focus attention on measurements of particular economic, social, or demographic phenomena. The programs themselves are oriented to the production and publication of statistical series, and the primary device for information retrieval and dissemination is the printed publication.

In general, the bodies of data do not mesh according to any overall system and there is much inflexibility which often prevents fitting the micro data to behavioral models. Reasons for the incompatibility include the following:

(1) differing definitions, classifications, and timing of respondent reports when uniformity is needed;

(2) difficulties of access to original or source data;

(3) differing qualities of data and inconsistent documentation.

In regard to the coordination among producers and users of data, the present system appears to be pragmatic and informal. In many cases, the statistical programs are connected with the administration of particular programs, and decisions bearing on statistical matters (for example, questions concerning the publication of series and changes in definition), tend to be highly influenced by the groups most closely affected by the operating programs. In some cases, the agencies have advisory groups and in other instances, the agency contacts with users tend to be more informal. To the extent that statistical decisions are influenced by particular groups, there may be a tendency for general needs to go unmet.

The decentralized organization of the statistical system, like the programs themselves, has evolved without a detailed, overall plan. This decentralization has imparted a dynamic character to the system and has enabled it to meet many emerging needs. At the same time, the coordinating efforts of the Bureau of the Budget have prevented many of the major adverse effects of decentralized efforts. However, under present circumstances, much more needs to be done and can be done to further integration-without sacrificing the benefits stemming from the basic decentralization of the data producing processes. As

one of the witnesses, Professor Richard Ruggles, stated with reference to the present system:

**The term "decentralized," while applicable, may be somewhat misleading *** it has been the result of a jungle-like growth of statistical activities by different government agencies having widely differing purposes. The result has been extensive duplication and lack of coordination. *** Although the Office of Statistical Standards has led to considerable improvement and rationalization, the Federal statistical system continues to operate on a highly decentralized and uncoordinated basis.

WHAT ARE THE IMPLICATIONS ON THE EFFICIENCY OF
OUR STATISTICAL SYSTEM?

On the basis of testimony, the subcommittee concludes that the present methods of operating our statistical programs have become obsolete in the sense that the need for integration and data availability have outstripped progress. At the present time, statistical programs require radical improvement to take advantage of modern technology and to meet the new and rapidly expanding needs of policy and analysis.

Although the situation described in the previous section was not ideal before the advent of the computer, it was then infeasible to relate large bodies of data either at one period of time or over long periods of time. However, computer technology has drastically increased the speed and efficiency of handling data and has made possible entirely new kinds of analysis. It has also greatly simplified the problem of data storage and retrieval.

The statistical series produced by Government agencies serve reasonably well the traditional tasks of management. But they are totally inadequate to meet the changing policy needs of our timesto aid in the planning of modern corporations and in the design and operation of new Government programs and the improvement of existing ones. Our rather isolated pieces of information do not fit into a general system. But this is a minor part of the difficulty since every general system has its limitations and inflexibilities. The real difficulty with the present system is its lack of flexibility, which is needed to interrelate different bodies of data. One witness, Dr. Edgar Dunn, expressed the present situtation as follows:

* We are still living in identical statistical "rowhouses," so to speak, when we have the technology and means to adopt the architecture to the size and interests of the family. The information contained in the statistical publications cannot be easily manipulated or related. Moreover, the micro data that were used to produce the aggregated series have frequently not been maintained in usable form, nor are the various bodies of basic data compatible due to a lack of coordination in the original production processes. The detailed data, then, which are being collected at great cost and inconvenience are not being fully utilized to produce the wide variety of aggregate statistics needed or to study micro relatioships. While some efforts are now going forward to put new additions to the stock of data into machine readable form for improved accessibility, much more effort on a Government-wide basis is needed.

In the area of national economic policy, improved integration would provide valuable information on productivity, prices, and incomes. For example, meaningful discussions of price-wage guideposts require quantitative information on productivity, prices, costs, and inputs for the total economy and by industry. Yet, at the present time, the output, input, price, and income data are estimated from different samples and by different statistical bureaus. The result is that we often cannot be sure whether a small computed change in unit-labor costs in a particular industry is due to sampling variability, differing sampling units, or differing classification procedures.

A second example where a more integrated system of data might significantly improve economic policy is in the area of fiscal policy. What determines the level of investment and thus full employment? Is it capacity utilization, corporate profits, previous investment, the cost of borrowing, or some other variable? Answers to questions of this type often involve bringing together financial data collected for enterprise units with nonfinancial data collected for establishment units. Significant progress in answering these questions might be made if researchers could relate relevant financial and nonfinancial data over an extended period of time or compare the situations of many diverse firms at one point in time.

In the field of human resource programs, further integration would help to improve the design and administration of government programs. For example, in the study of the causes of poverty, it would be advantageous to merge Census data with data from the operation of the Social Security Administration, and with income data from the Internal Revenue Service. The ability to combine data sets and to compare data over time would also help to refine the dimensions of poverty. Efforts to relate some of these bodies of data have been and are being made sporadically but these efforts must not only be increased but made a regular and organized feature of our statistical system.

The integration of State and local data with Federal data would aid in the design and operation of government programs which provide for special educational and training resources, or special aids to businesses. It would also be valuable to State and local government planning and to businesses in their planning of investment and sales. Here, too, efforts have begun, but they require far greater scope and support.

In addition to the loss of analytical capabilities stemming from the present disorganized system, there is unnecessary duplication of effort by government agencies and duplicative reporting by respondents. In part, these duplications arise because of the absence of a master directory of business respondents available to all statistical agencies.

In addition to duplication, the present dispersal of activities means that economies of scale in the use of automatic data processing and technical experts are not being fully realized. This, also implies that the information is being produced more slowly than would be required by modern processes.

WHAT PROPOSALS ARE THERE FOR A MORE INTEGRATED SYSTEM?

The subcommittee heard proposals designed to improve the operation and particularly the integration of our statistical programs.

The most complex and potentially far reaching of the proposals was that of establishing a national statistical data center (sometimes referred to as a national statistical servicing center). Little agreement was expressed on specific aspects of such a center; however, its proposed general functions would include the following:

(1) To establish and maintain an index of all sets of data available within the Federal system and usable for statistical purposes.

(2) To assemble the data from selected sources, integrate them to the maximum feasible extent, and preserve the data in usable and highly accessible form by use of large scale automatic data processing equipment. (With modern technology, it may be unnecessary to physically locate the data in the center, in view of computer hookups and computer consoles.)

(3) To function under uniform disclosure standards, which should also apply to other Federal statistical agencies, in order to insure maximum protection of personal privacy.

(4) In cooperation with users in and out of government and with the collecting agencies, to formulate standards for further collection efforts in order to promote maximum integration and minimum duplication without undue sacrifice in terms of specific functions of the agencies.

(5) To define the accessibility and cost of various sets of data to potential users inside and outside of government. Special tabulations could also be performed by the center, with due consideration for disclosure restrictions and with costs assessed against the users.

The potential benefits from such a center are difficult to estimate since it would create an entirely new dimension in statistical servicing capability. In general terms, the center would provide for a more efficient use of collected information by providing greater flexibility.

First, the center would provide a staff and other resources that would be ready, willing, and able to service special requests for data and for information about data from users both inside and outside of Government. This would greatly increase the efficiency of data users since it would save hours of searching for the data, negotiating to obtain the data, and readying the data for use.

Second, the center's operation would gradually improve the quality and increase the quantity of data available, while either reducing the burdens on respondents or at least not increasing them. In part, this would arise because the basic micro data from the separate programs and agencies would be permanently stored, coded and made usable under standards of the national statistical center. Thus these micro data could be related to each other and/or recombined into new aggregates. At the same time these efforts would reinforce the other major function of the center-namely the improved coordination of the processes of collection, processing, storage, and retrieval of data by all the Government agencies. Since the operations could be carried out inside the center, manipulation of the data (for example, matching operations) could be carried out without violating disclosure principles or threatening personal privacy.

The financial cost of a Federal statistical servicing center cannot be satisfactorily estimated because the staff work essential to the specification and costing of program options has not been done. Realistic cost estimates can only be developed gradually as experience is gained. Mr. Raymond Bowman estimated that an initial budget of about $2 million per year for the first 2 years would be required to start a

statistical data center. He went on to estimate that in the following 3 years, the costs would be about $5 million per year, exclusive of computer hardware costs.

Although the center would incorporate several functions designed to further integration, some of these possibilities were mentioned as separate possibilities.

A master list of firms and establishments was proposed to aid in the design and operation of sampling and in the classification of data, for example by area and industry. Under the present situation, the Standard Industrial Classification System does not guarantee that a particular firm or establishment will be classified in the same industry by each of the different agencies, particularly in cases where the firm has several establishments, one or more of which produce products in different industries.

Several witnesses also testified as to the need for an index describing the statistical data collected by the Federal Government. The index would indicate such characteristics as where the data are stored, their accessibility, and sampling characteristics.

6

The subcommittee probed for suggestions on ways that the Federal Government could encourage cooperation among statistical programs at the Federal, State, and local levels of government. Numerous witnesses stressed the need for Federal aids to promote an integrated network of statistics for the United States. Suggestion was made that operating programs involving two or more levels of government should have provisions to insure comparable and meaningful statistics.

WOULD FURTHER INTEGRATION AND PARTICULARLY A NATIONAL DATA CENTER PRESENT A THREAT TO PERSONAL PRIVACY?

Further integration can and should be attained without sacrificing principles of personal privacy. Modern technological capabilities for data association, storage, and retrieval, like most technological changes, present new possibilities for use or misuse. The rational approach to these new capabilities is not fear to act, but rather action to control technology for man's use."

From the standpoint of the national data center as it relates to the issue of personal privacy, relevant considerations include (1) the determination of what information might go into the center, and (2) the design of adequate safeguards to prevent misuse of information. In regard to the first point, it should be pointed out that the proposed national data center differs fundamentally in design and purpose from central files designed to hold information about individual people or individual firms. The purpose of the data center under discussion is to furnish information primarily about the "public face" of groups of people, households, firms, etc., and their interrelationships-never about the individual units themselves. This would involve

The Supplement to Economic Indicators, prepared for the Subcommittee on Economic Statistics by the committee staff and the Office of Statistical Standards, Bureau of the Budget, describes some of the most commonly used series in the economic field. Major Federal statistical agencies publish catalogs, price lists, and other types of guides to their statistical publications. Agency catalogs and other guides to statistical publications are listed in and complemented by the bibliography of "Principal Statistical Publications of Federal Agencies" published by the Bureau of the Budget in its booklet, Statistical Services of the United States Government.

The issues concerning personal privacy have been the subject of investigation not only in the recent hearings of this subcommittee, but also hearings of the Special Subcommittee on Invasion of Privacy of the Committee on Government Operations, U.S. House of Representatives; and of the Subcommittee on Administrative Practice and Procedure of the Committee on the Judiciary, U.S. Senate.