College and Research Libraries JUDITH S. ROWE and MARY RYAN Library Service from Numerical Data Bases: The 1970 Census as a Paradigm This article discusses some of the problems of introducing machine- readable data bases intp the library service environment. The authors, a social scientist at a computer center, and a government documents librarian, describe the diverse approaches used in making tapes of the 1970 Census of Population and Housing available to users through the library. LARGE RESEARCH LIBRARIES have tradi- tionally been depositories for all of the maps and printed reports which are the products of each decennial census. Therefore it is a logical next step for them also to be the repository of these data in machine-readable form. First, this provides reference librarians with another resource for users whose needs are not satisfied by searching the printed materials, since the quantity of addi- tional data which can be stored com- pactly on magnetic tape has made it pos- sible for the Bureau of the Census to make available to the public at least ten times the amount of data available from any previous census. Second, re- search libraries are generally located at institutions which also have available large computers capable , of selecting, di- gesting and analyzing these data and, if tapes are available, it becomes unneces- sary for the user requiring a machine analysis to photocopy pages from re- ports and then keypunch the data. Rath- er, it becomes possible for a researcher to begin with the data already in rna- Judith Rowe is manager, Princeton-Rut- gers Census Data Project, Princeton Uni- versity Computer . Center; Mary Ryan is head, Public Affairs Service, University of California, Los Angeles, library. chine-readable form and to proceed im- mediately to work with these data. Thus, we now have more data in .a more usable format than ever before. But why should the library be in- volved? Why not just store the tapes at the computer center and let the com- puter people worry about them? By do- ing this, the library would be abdicating part of its role as an information cen- ter. It would be denying users the op- portunity of locating information at the one place we have trained them to look for it, the library. By the same token, since the acquisition of even a modest tape collection represents a sub- stantial financial outlay, it is important to ensure that the responsibility for de- cisions on acquisition, on bibliographic documentation and control, and on ref~ erence service be allied with an organi- zation with a commitment to public service and a continuity of collections and operation. Although many comput- er centers have a commitment to public service, computer centers are not li- braries and their staffs do not have the library skills necessary for such a proj- ect. On the other hand, librarians do not generally have the computer exper- tise necessary to exploit to the fullest this new information resource. What then is the solution? Actually /7 8 I College & Research Libraries • January 1974 there is no one best solution. Different plans are now in operation and many more are possible. In this report we will describe in detail two very different ap- proaches-at UCLA and at Princeton. As background, however, let us first pre- sent some of the questions which must be faced in arriving at a locally work- able solution. 1. Who will decide which data should be purchased? To some extent this is a matter of money, but other considerations are also involved, most of them similar to those in- volved in the building of any li- brary collection. How does one ser- vice all or most requests without the problems of supporting a huge amount of unused material? 2. What quantity and type of service will be provided? This depends on the availability of both funds and people. It also depends on the needs of users and on the nature of any other services of this sort already available in one area. 3. Where will the money come from for acquisitions, for computer time, for personnel, and for a myriad of miscellaneous expenses such as travel, duplicating, backup reels, telephones and clerical as- sistance? Many libraries are having great difficulties in maintaining even their traditional acquisitions and services. Assuming that one's total resources are not increased, does one reallocate resources and, if so, what criteria does one use, and just whose resources does one reallocate? It should be borne in mind that a complete collection of 1970 census summary tapes, pur- chased from the Bureau of the Census itself, would cost well over $100,000. 4. How is bibliographic control over machine-readable data bases to be exercised? Should records for these be interfiled in the card catalog? Given the complexity of these data bases and the volatile nature of machine storage, is the traditional catalog record sufficient? Are sup- plementary materials such as data description forms and content doc- umentation codebooks necessary and, if so, how are the records to be integrated? No fixed solution yet exists, but an ALA subcommittee has been established to recommend rules for the cataloging of ma- l'J chine-readable data files.~ \jl The whole question of staffing gives rise to many problems. What kind of staff and how much staff are needed and can be afforded? What background should they have? How are they to be trained? Under whose supervision should they be and how does one train the supervisor? Who will assume re- sponsibility for those problems which are general to the nature of machine-readable data files and those which are specific to the cen- sus itself? These include such things as the vast amount of data involved, the relation of these data to the printed reports, the complex- ity of the tapes, questions of geo- graphic coding, the necessity of maintaining and updating comput- er programs, errors and inconsisten- cies in the data, recalls of tapes, printed corrections to erroneous data on the tapes, problems in la- beling and identifying data sets, 0 ]ohn Byrum, head cataloger at Princeton University is chairman of the RTSD/CCS/ DCC Subcommittee on rules for cataloging ma- chine-readable data files. Other subcommittee members are: Lawrence W. S. Auld, Oregon State University, Henriette Avram; Library of Congress, Gerry Dobbin, University of British Columbia Library, Elizabeth Herman, U.R.L.- UCLA, Judith S. Rowe, Princeton University Computer Center. and a great many problems in con- nection with documentation, rang- ing from its quantity, combined with inadequate indexing and . cor- relation, right down to such mi- nute problems as some of the items most needed by users being printed in colors that photograph poorly. 6. Where should the staff be housed? Where should the data tapes and the supporting documentation be stored? At the library? At the com- puter center? Somewhere on neu- tral territory? Or should the opera- tion be separated, with reference service being provided at one loca- tion and data access and use at an- other? Here, as elsewhere, commu- nication is a matter of paramount importance. Computer program- mers, librarians, and social scien- tists often find it difficult to talk the same language. Is the an- swer an interdisciplinary specialist? There are also problems within the library itself. For every depart- ment in the library, from acquisi- tions to cataloging to reference, this new medium poses new prob- lems and we can testify from ex- perience that few precedents exist to aid in their solution. 7. What is the potential user commu- nity? Who will have direct or indi- rect access to the data? Is the ser- vice primarily designed for one's own campus users or for the out- side public? Will there be prefer- ential treatment, and, if so, on what basis will it be accorded? For example, would students,· faculty, researchers with outside grants, and profit-making firms all have equal access and would they all pay the same rate for any charges in- volved? 8. Will there be user charg-es and, if so, on what basis will they be de- termined? If user charges are to be Library Service I 9 instituted, how much demand for service will there be, especially in a nonmonopoly situation, and how much will charges have to run? If demand is sporadic, for example, would charges have to be unrealis- tically high? Should they be the same as those of profit-making or- ganizations providing similar ser- vices? What will one attempt tore- coup with user charges and what will be the public relations effects? Probably few people would ques- tion charges for computer time or for staff time spent on special pro- gramming, but what about staff time spent in reference, orienta- tion, and consultation? At precisely what point, for example, would a university library halt the free reference and information service it of- fers its faculty on its census resources and put it on a for-sale basis? ("Well, Professor Jekyll, I can answer questions on Part I of the Census User's Guide, provided they don't touch en the sum- mary tapes, but no questions about Part II," or: "You can look at the Census Bureau's fourth count documentation without charge, since it's depository, but it will cost you x dollars to use DUALabs' version.") At what point would conversations pass from the free orientation service stage and become priced consultations? Does one attempt to amortize the cost of the data base- and would there be public relations problems here? ("Well, Professor J eky II, the reason you must pay to use $1,000 worth of the library's census data, while Professor Smith can use $1,000 worth of the library's Sanskrit manuscripts without charge, is that we amortize tape data but not manuscript data.") What about amortizing the cost of supporting tools, such as maps? ("Well, Professor Jekyll, you may look at the commercial map of Butte County free, since we 10 I College & Research Libraries • January 197 4 didn't buy it to support computerized census, but you must pay a fee to look at the census map for that county." "Professor Jekyll, why are you begin- ning to look like Mr. Hyde?") Having surveyed some of the major problems, we will now look at two solu- tions. The University of California, Los Angeles, and Princeton approaches are very different in origin, scope, re- sources, and services, but they are simi- lar in that in both instances libraries and librarians are fully involved. THE UCLA PRoGRAM At UCLA there was no central social science data library and no history of library involvement in machine-readable numerical data acquisitions. However there was a large computer available supporting several major statistical soft- ware packages, :as well as some data processing personnel involved in other projects within the library, a large gov- ernment documents collection, and a po- tential user community of several hun- dred. Many commercial operations in the Los Angeles area had early announced their intention to serve as Census Sum- mary Tape Processing Centers; in addi- tion, a nonprofit, self-supporting Cen- sus Service Facility had been established at the University of California, Berke- ley. This facility was operating under the joint auspices of Berkeley's Institute of Governmental Studies and its Survey Research Center, a group with a long history of service as a social science data archive. It had no connection with the library. It offered a wide range of ser- vices, emphasizing the production of standardized tabulations, as well as per- forming customized work tailored to in- dividual requests. Although it would serve private organizations and individ- uals, its efforts were particularly directed towards academic and governmental users; for UCLA to have duplicated Berkeley's services would have been needless and wasteful. However, it was obvious that students, faculty, and research personnel in many different fields would themselves need to be working directly with the machine- readable data on an individual basis. Unless a central source for the data and tools were available to them, various UCLA departments, schools, and insti- tutes would each have to obtain the tapes independently and, while there would then be much wasteful duplica- tion, there would still be no central source of information and no general availability of the data tapes. Foreseeing this situation, the library undertook the responsibility for serving as a central campus resource for census tapes and for information about them, including appropriate cataloging. The library system was already involved with several machine-readable data bases, in- cluding MEDLARS and MARC. In ad- dition, there was a Center for Informa- tion Services (CIS), funded by the Na- tional Science Foundation and then in · the second of four phases, which has the specific purpose of giving the li- brary the capability of acquiring, cata- loging, and providing services for rna- . chine-readable data bases, whether bib- liographical,_ numerical, or full-text.. CIS has as its .first priority the biblio- graphic files and its experimental ser- vices have included searching of CA Condensates, Compendex, CAIN, and the ERIC files. The census represented its first involvement with a numerical file. Although the need for a census ser- vice was very apparent, the resources available to the library were extremely limited. Had an attempt been made to recover costs through user charges, po- tential revenue would have been mini- mal because of the existence of Berke- ley's Census Service Facility, or would have been siphoned off from that facili- ty. UCLA has no specific budget alloca- tion for either census data acquisition, processing, or reference service. These are all paid for on an ad hoc basis out of the library budget, or by other par- ties., such as deparhnents, willing to con- tribute. Despite the administrative and finan- cial difficulties, it was decided that the UCLA library would attempt to offer service, within the limits of its resourc- es, to meet the most crucial campus needs. The course of action which seemed most appropriate, given the above framework of limited experience and stringent budgetary considerations, was for the library to join the START (Summary Tape Assistance, Research, and Training) Community organized by DUALabs (Data Use and Access Lab- oratories, Rosslyn, Virginia) under the sponsorship of the Center for Research Libraries, with aid from the Ford Foun- dation. Through this community, it would be possible to purchase tapes at a price substantially less than that of the Bureau of the Census and to take advantage of the programs already de- veloped by DUALabs to avoid incurring the heavy cost of original programming. Within the UCLA context, it was ob- vious that the logical library deparhnent to undertake the census tape service was the Public Affairs Service (PAS). Among PAS' key responsibilities is that of the library's government documents service. Thus, it receives the current printed reports from the Bureau of the Census, has a heavily-used reference ser- vice specializing in government docu- ments, and has had long experience with the census printed reports. Even more important, PAS, which incorporated several older services such as government documents, had been created in 1968 to offer a coordinated information service to those working in the fields of gov- ernment and public affairs, broadly in- terpreted. As a department of the re- Library Service I 11 search library, designed to supplement that library's more traditional resources, PAS was directed to place no limits on the kinds of material or forms of data that were acquired or used, so long as they were pertinent to the needs and in- terests of the clientele serviced. The census in machine-readable form is, in fact, a perfect example of the uncon- ventional library resource which the de- parhnent was created specifically to han- dle. With this mandate, census tapes for California, plus the necessary tools- programs, the MEDLIST, etc.-and the needed documentation, were ordered. A specialized census reference service is now offered which includes extensive personalized orientation and a limited amount of consultation. General ques- tions about the tapes are answered at PAS' regular reference desk, but this specialized service is a separate and dis- tinct service within PAS. There were several reasons for this. First, the ref- erence desk is an exceedingly busy place . where questions must be answered ex- peditiously, or suggestions made to en- able the user to start on his own search, so that the next reader waiting can be helped; since the typical initial census tape orientation takes at least one hour, it could not be handled as part of the regular service without causing that ser- vice to break down. Furthermore, there are in all nine librarians and seven oth- ers who are scheduled at the Public Af- fairs reference desk, and, given the time needed for someone to become trained in the census service and to keep up with the continuing How of docu- mentation, it was not economically fea- sible for all the staff to participate in the service. This service is purely a reference and orientation service. It is not a produc- tion operation. Programs are available for the clientele to use, but the librari- ans do not themselves use them. Nor do \., 12 I College & Research Libraries • January 1974 librarians offer such services as printing out data from the tapes for people, manipulating data for them, or doing any data processing. Instead, they make the data and the tools available to the user, so that he may do his own work. If he wants the processing done for him, he may obtain this service from Berkeley or another Summary Tape Processing Center at their stated rates. The reference and orientation service is available to any inquirer and iri fact the UCLA library serves as a User Con- tact Site of the Clearinghouse and Lab- oratory for Census Data, operated by DUALabs under contract with the Cen- ter for Research Libraries. Copies of the tapes themselves and of the programs are kept at UCLA's Campus Computing Network ( CCN) with access restricted to those who have received prior authorization from the library. Access is authorized by the li- brary as a matter of course for anyone who has CCN computer time and he may then use the tapes and programs in accordance with CCN' s standard proce- dures. This authorization system is de- signed to provide data on demand and usage to aid in planning. It enables the reference staff to ensure that potential users are acquainted with the documen- tation and are aware of the printed re- ports before they start their work with the tapes. It enables a degree of security to be maintained over the tapes, since it reduces the chance of damage caused by a completely uninitiated and untrace- able user. In addition, insofar as users are willing -and this is done only with their consent-it permits the library to act as a clearinghouse to alert users to similar projects already underway on campus. At the present time, no charge is made by the library for its census reference service or for access to its tapes and pro- grams, though this may well change. CCN of course applies its standard pro- cedures in relation to the computer time. If a user is unable to obtain CCN computer time, or prefers not to, CIS will sell him copies of any of UCLA's census tapes or programs, so that he may use them at his own computer facility. At one time, tape lending was consid- ered, but the problems inherent in such a procedure made this impractical. At present, therefore, the library's service is completely designed for "do- it-yourselfer.s." In some ways, it is analo- gous to typical academic library service on books in foreign languages. For ex- ample, bibliographies of German books, the German books themselves, diction- aries, indexes of translations, and direc- tories of translators and translation centers are made available to the reader. He is offered reference service, helping him to identify these items-but not a translation service for German books. There is a group of services ranging from reference through programming and keypunching, to which the user of census data, or of any data in machine- readable form, must have access. How- ever, the decision as to which of these services will be supplied directly by the library and which will be handled else- where, with or without library involve- ment, is a matter of local option. These services are supplied somewhat different- ly at Princeton than at UCLA, largely because the origins of the Princeton Census project, and its financing, are different. THE PRINCETON PROGRAM When the Princeton Library was ap- proached by the Center for Research Li- braries about acquiring the census tapes, two important precedents already exist- ed: a tradition of cooperation with Rutgers for the purpose of avoiding duplication of special collections, and a prior acceptance of numerical rna- chine-readable data files as a legitimate library resource. Since it was obviously less expensive to combine forces, an agreement was signed which created the Princeton-Rutgers Census Data Project For the Princeton library this was a logi- cal sequel "to other similar steps. For ex- ample, Princeton is a member of the Inter-University Consortium for Politi- cal Research and the Consortium Mem- bership Fee has for several years been part of the library budget, although re- lated computer time and personnel ser- vices are funded by the Computer Cen- ter. When this decision was first made at Princeton, it was innovative. Today several other schools belonging to the Consortium have followed the Prince- ton pattern and more are considerit:lg doing so. However, it is not enough merely to foot the bill. Someone must undertake responsibility for the acquisi- tion, the storage, and the use of the data. These responsibilities had been as- sumed for many years at Princeton by what is now known as the Social Science User Services ( SSUS) section of the · Computer Center and therefore it was logical for this group to perform the same functions in relation to the census tapes. The advisory committee involved in the organization of the Census Project included, in addition to librarians, rep- resentatives of the Computer Center and of several departments in which - there were potential census users. This committee confronted immediately the fact that funds would be needed for data acquisition and storage, computer use, and programming support. Contri- butions were therefore secured not only from the Princeton and Rutgers li- braries but also from the budgets of several large research projects which al- ready contained allocations for the ac- quisition of census data. The directors of these projects were all pleased with Library Service I 13 an arrangement which would afford them access to the data tapes without the burdens of acquisitions, biblio- graphic recording, physical control, or software development and at less cost than if individual purchases were to have been made. Practically, how does the project function? The Princeton-Rutgers Cen- sus Data Project is recognized by the Bureau of the Census as a Summary Tape Processing Center and by the Clearinghouse and Laboratory for Cen- sus Data as a User Contact Site. This means that inquiries at the project of- fice, which is located at the Princeton University Computer Center, frequently come from nonuniversity sources. How- ever, on the four major campuses served, people are accustomed to look- ing for printed census data in their li- braries and the project encourages them to continue doing so. With the aid of the Census Packet (a monthly acquisi- tions list covering both printed and ma- chine-readable materials, which also in- cludes lists of suggested readings, de- scriptions of maps and other geographic aids and miscellaneou,s training materi- als ) , Part II of the Census User's Guide and a one-day orientation, most of our reference librarians are aware of the potential of the machine-readable data. The librarians who are directly respon- sible for the Bureau's printed reports, particularly the Public Administration Librarian at Princeton and the Govern- ment Documents Librarian at Rutgers, can answer questions concerning the specific tables which are included in the machine-readable files and the geograph- ic areas which are more adequately cov- ered by these and, as a result, it is now quite common for the project to receive a call from a library user asking, for ex- ample, "Can I get Table 29 in the Sec- ond Count for all minor civil divisions in four New Jersey Counties?" 14 I College & Research Libraries • January 1974 Reference librarians are still doing census reference work but with expand- ed resources. Those users who come di- rectly to the project office, but without such a specific request, are first referred to the printed reports, samples of which are available for use there. Frequently, the next step is to send them to the near- est library census collection, but in those cases where it is evident that they will need machine-readable data, two op- tions are open to them. They may have the necessary retrieval done for them, and for this there is a charge of $8.50 per hour for programming time with a $25.00 minimum per request or they may themselves use the available pro- grams and access the necessary data tapes in which case they receive relevant orientation and documentation without charge. In either case they would pay for computer time but not for consul- tation or for use of the tapes. In administering the project the So- cial Science User Services section of the Computer Center has provided a bridge between the library and the computer. The services performed by this group are similar to those provided by com- parable centers throughout the country -sometimes at computer centers but more often attached to research insti- tutes or social science departments. However, the Social Science User Ser- vices section is unique in that it main- tains and has maintained for years an active relationship with both the refer- ence and technical services staff of the Princeton library so that in many ways it functions as a special library. Its regu- lar services include responsibility for ac- quisition and control of much of the social science and literary data in ma- chine-readable form obtained by or g~­ erated at Princeton University, and then released for public use. This involves providing reference service not just for the data tapes but for all of the sup- porting documentation, including code- books and records of physical and logi- cal characteristics. SSUS also main- tains three major statistical packages (OSIRIS, SPSS, and Data-Text) for analyzing data on the computer, and has access to a fourth, the Princeton- produced P-Stat. Consultation is avail- able for any computer-oriented research project in the social sciences or the hu- manities; this process can cover all phases of methodology from question- naire design to analysis. All of these services are provided without charge to members of the university community; users pay only for special programming and keypunching. This then is how census tape service has developed at UCLA and at Prince- ton-Rutgers. Many other solutions are possible within a library framework, but whatever the approach finally adopted by a library, it is certain that there are many potential users not only of census data in machine-readable form but also of the vast array of oth- er machine-readable data resources which are becoming publicly available. In spite of the obvious technical diffi- culties, these are clearly significant in- formation resources and as such should not be ignored by libraries. Machine- readable information resources are now available to a greater or lesser extent at virtually every research-oriented college and university and in government agen- cies at all levels, but the number of in- stances in which the libraries at these institutions are involved or even aware of these resources is sadly small. No li- brary, regardless of its lack of technical expertise, should completely surrender its responsibility as .an information cen- ter to an academic department, a re- search group, or a computer center. It is not necessary that librarians hold these data physically in the library or that they process or even know how to process them, but it is necessary that reference librarians have enough knowl- edge of these data to advise users of their existence, their general contents, and of the means by which they may be accessed. Failure to do so is likely to re- sult in proliferation of competitive- ser- vices in an area in which costs can be- come extremely high. FINANCING For those libraries which have the de- sire and the capability to become more heavily involved in providing data ser- vices, the question of financing may seem insoluble. Once a library has de- cided how far along the continuum of possible data services it wishes to move (and this decision will inevitably be col- ored by the nature of any existing ser- vice at one's location ) , the question of how to finance these new activities must be confronted. Assuming that a library's budget cannot be increased to meet this new demand, three alternatives are available, all of which may seem at odds with traditional library policy, but all of which have already been implement- ed by libraries. The first alternative is the re-allocation of existing library re- sources. The second is to institute user charges. These may be applied directly to each individual user or may be paid by departments, agencies, or research groups in a lump sum determined by usage. User charges might well be ap- plied to all computer-related work but probably not to basic reference service in most libraries. It is certainly simpler and perhaps more acceptable to users to charge for keypunching, programming, and machine time than for orientation and basic consultation. The other alter- native which could be implemented ei- ther separately or in combination with user charges is that of outside subsidy~ Traditionally, libraries have provided out of their general fund for acquisi- Library Service I 15 tions requested by departments without expecting reimbursement from depart- ments. Since, research budgets often contain provisions for machine-readable acquisitions, it seems entirely appropri- ate that when possible these funds be funneled through a central channel, the library, especially since a contract or grant may actually require that any such data become the property of the institu- tion rather than of the individual. Al- though it must not be overlooked that many granting agencies specifically pro- hibit the purchasing of library resources on their funds (since the preexistence of adequate library facilities was a ba- sic reason for awarding the grant to that institution in the first place) this seems an administrative problem capa- ble of solution. All things considered, it seems logical that, before embarking on the major new activities that service from machine-readable data bases rep- resents, a library might well solicit con- tributions from potential user groups, whether on-campus or off-campus, and employ any such contributions as seed money for initial acquisitions and proc- essing. To summarize, no library can com- pletely abdicate its involvement in ma- chine-readable data resources, unless it elects to abdicate part of its responsi- bility as an information center. How- ever, since other nonlibrary centers may by default have assumed many of the functions involved, the degree of li- brary activity must take account of the existing situation. At the very least, com- munication with the nonlibrary center would always be desirable, as would en- tries for machine-readable data files in the public catalog and basic reference service. This minimal activity must be- come an integral part of the service of every research library.