Selection Practices for Web-Based Government Publications in State Depository Library Programs: Comparing Active and Passive Approaches

Transcription

1 Selection Practices for Web-Based Government Publications in State Depository Library Programs: Comparing Active and Passive Approaches Chi-Shiou Lin School of Library and Information Studies, University of Wisconsin-Madison, 4258 H.C. White Hall, 600 N. Park St., Madison WI 53706, Kristin R. Eschenfelder School of Library and Information Studies, University of Wisconsin-Madison, 4228 H.C. White Hall, 600 N. Park St., Madison WI 53706, This comparative case study examined the capture and selection practices used to populate electronic depositories of born-digital state government publications. The three case sites illustrate differences in collection building approaches, technological infrastructures, and statutory contexts. The findings reveal two basic modes of selection practices active selection and passive selection, and three selection models based on the loci of selection control library selection, liaison selection, and creator selection. Also, the findings suggest the power of defining and selecting government publications for state depositories is shifting from government agencies to state libraries in the active selection. The authors argue for the needs to attend to Web publications in non-traditional formats (e.g. an interactive HTML document) and to include common publications produced for lay citizens (e.g. brochures, fact sheets, FAQs, etc.) in the permanent collections in order to fully document the government activities for the historical record of a society. Introduction Over the last decade, there has been a significant shift in the United States government publishing from print publications to Web-based information. Encouraged by several legislative and administrative initiatives (Government Printing Office, 1996), governments at the federal and state levels are now publishing and disseminating massive amount of information predominantly via the Web. In 2003, nearly two-thirds of the information resources made available via the Federal Depository Library Program (FLDP) were born digital (California Digital Library, 2003). At the state level, a recent study estimated that 43% to 60% of the North Carolina state publications were published in digital format only (Martin, 2004), and the Arizona state agency Web sites contain more than 300,000 electronic documents at any given time (Pearce-Moses & Kaczmarek, 2005). For government agencies, Web publishing has several advantages over printing. It is considered a costeffective means of information dissemination, especially in times of budget cuts (Mullen, 2001; Martin, 2004); Web content is easy to maintain and update (Martin, 2004), and it potentially provides improved access (Mullen, 2001), enhances government-citizen interaction experiences (Clayton & Gregory, 2003; Reddick, 2005), and encourages civic participation (Larson & Rainie, 2002; Horrigan, 2004). For depository library programs, however, the shift from print to Web publications requires dramatic rearrangement of work processes. Many state libraries have initiated projects to collect and preserve Web-based government information. Several competitive models have emerged (Matthews, et al, 2003; California Digital Library, 2003; Pearce-Moses & Kaczmarek, 2005). This study systematically examined the collection development policies, statutory frameworks, and work arrangements for capturing and selecting Web-based state publications within three states Connecticut, New Mexico, and Texas. It identified the advantages and challenges associated with practices used in the programs and suggests important areas for future inquiry. Web-based information is changing the relationships between state libraries and government agencies. One area of change is who performs the work of reviewing and selecting information resources for permanent retention and how this work is performed. When printing was the predominant method for government publishing, government agencies decided what to publish into a tangible publication and what publications to submit to the depository library programs (Hernon & Relyea, 1995; Quinn, 2003). The state libraries, serving as distribution centers of the depository programs, were a passive recipient of government publications by the nature of the institutions. But in the Web environment, as we will show, the libraries are becoming active selectors for the initial inclusion of a Web publication in the digital depository. Further, in the two modes of selection practices we identified active selection versus passive selection, the library-agency relationships vary. While the limited data

2 limits any conclusions about the superiority of one model, analysis highlights advantages and disadvantages of each approach. Importantly, the study found that the current vocabulary in the literature is not sufficient to describe important differences between the case studies. Consequently, we introduce a new three part typology that distinguishes program arrangements in terms of capture methodologies, modes of selection, and the loci of selection control. Data from the three cases showed that publication type and file format are important cues for selection. This continued reliance suggests that selectors may overlook newer document genres specific to the web. The Web environment also contains many more ambiguous publication types (such as HTML-based brochures or FAQs) that require selectors to make critical selection decisions. Findings show that even in passive selection approach, the library must still provide collection development guidance to state agencies. Finally, the study considers possible positive and negative implications of use of the different models for the importance of document collections as a unique and important historical record (Hernon & Relyea, 1995). Literature Review An Overview of the Web-Based Government Publications Initiatives The preservation of born-digital Web-based information resources has received extensive research attention in library communities, such as developing models and infrastructures for digital repositories (Steenbakkers, 2000; RLG & OCLC, 2002), articulating electronic deposit workflows (Van der Werf, 2000), developing metadata schemes (OCLC/RLG Working Group, 2001) and software tools for capturing Web documents (e.g., Illinois State Library s PEP & CEP projects.) In the United States, the government has taken some action to preserve borndigital government publications. At the federal level, GPO has redefined FDLP activities to include the active discovery and selection of digital documents for inclusion in a master collection (Barnum, 2002). At the state level, progress in addressing the permanency of Web-based government information varies from state to state (Matthews, et al, 2003). Several states have initiated projects and employed technologies to collect and preserve Web publications (Matthews, et al 2003; California Digital Library, 2003). Existing state-level projects employ different collection building approaches. The California Digital Library report (2003) distinguishes between selective collecting and bulk collecting. Selective collecting requires that selectors have some a priori knowledge about the Web sites and the resources hosted in them. Bulk collecting uses automated crawlers to perform a global capture in a target domain or set of sites. Practices also differ in terms of the curatorial principles for the collections. The bibliocetric model aims at collecting publications ; similar to traditional library acquisition practices, selectors pick and choose Web content from the agency sites as if acquiring individual publications. A second approach, the archival model, envisions a Web site as an archival collection; this approach selects and manages hierarchy of aggregates, treating directories and subdirectories in a site as the series, and subseries of an archival collection. It handles aggregates of documents rather than individual documents (Pearce-Moses & Kaczmarek, 2005). State libraries that use OCLC Digital Archive (DA) service to develop their digital depository collections employ a selective collecting and bibliocentric model actively identifying individual documents for inclusion in the collection. Ten state libraries have contracted with OCLC to use the service (OCLC, 2004). To add a publication into the DA, the library staff must first identify the publication from an agency Web site. A cataloger then creates or locates a bibliographic record in WorldCat and instructs the system to harvest and ingest the target publication (OCLC, 2003). Texas, on the other hand, employs a home grown system that does not require active selection by the state library. The state is in the process of building an Electronic Depository Program (EDP) based on an existing Government Information Locator Service (GILS) called TRAIL (the Texas Record and Information Locator). The TRAIL/EDP also collects publications, but it employs the bulk collecting method. The state agencies are legally mandated to tag Web content at the publication level with metadata, and then the TRAIL/EDP automatically harvests, indexes, and stores those tagged publications. This collection building approach involves rarely active selection decisions from the state library (Hartman, 2001; Hartman & Condrey, 2004). The Challenges in Selection Web Publications For those states that employ a selective collection approach, selection decisions in Web publication preservation projects are complicated by a number of factors. Real preservation challenges stem from the short life of government information (California Digital Library, 2003) and the lack of agency efforts to restoring and

3 retaining retrospective information (Martin, 2004). Research shows that agency webmasters are usually more concerned with providing current and updated information and enhancing the user experiences by improving the looks and the functionalities of the Web sites rather than providing long-term access to superceded or noncurrent publications (Martin, 2004); a Web document may be removed as long as the agency considers it has served its current missions without thought to its continued value to the public. Previous research suggested that the average length of existence of a Web document in government agency sites is four months (Cho & Garcia- Molina, 2000). Further, many agencies do not have plans to retain and restore the removed born-digital information resources (Martin, 2004), and many states do not legally mandate such retention efforts by the agencies (Matthews, et al, 2003). If the state libraries fail to capture a Web publication in time, it may be lost forever. Challenges also stem from the fact that government information is often placed outside of the dot-gov domain (California Digital Library, 2003). Some agency web sites and publications produced by quasi-governmental organizations, oversight boards, or public education institutions may not have dot-gov Web URL addresses. This can cause confusion for both human selectors and automatic crawlers as it is may not be clear if information residing in the other domains is a legitimate resource for the collection. Also, estimates suggest the dot-gov domain contains a good deal of information in what is known as the Deep Web, such as information stored in databases that cannot be captured by crawlers (California Digital Library, 2003). The genres of the Web-based information resources also constitute a problem. Traditionally, government information resources are categorized into publications, documents, and records. Such distinctions, however, are blurred in the Web environment and can cause confusion in selection decisions (e.g., see Oregon State Library s discussions on item selection criteria available at Web publishing also creates new types of resources, such as organizational information (directories, job announcements, and press releases), homepages, and service-based online transactions (California Digital Library, 2003). The genre diversity of Web-based government information not only creates problems for selectors, but also challenges the traditional institutional responsibilities between libraries and archives. The statutory basis for collection and selection may also create challenges. While some state statues clearly define items to be included and the responsibilities of the parties involved (e.g., Texas), other state statutes are vague and leave room for disputes about what digital resources ought to be included and who has responsibility for selection or collection work (Matthews, et al, 2003). The collection building methods (selective collecting versus bulk collecting) both have benefits and drawbacks. Bulk collecting, while requiring few selection resources, can result in a huge collection containing numerous unwanted items. Further, many argue that the Web crawlers used in bulk collecting may fail to harvest important deep Web content. On the other hand, selective collecting is labor intensive and requires subject expertise; further, the strategy does not guarantee that selectors will collect all important Web information or be able to include deep Web content (California Digital Library, 2003; Pearce-Moses & Kaczmarek, 2005). Methodology This study used an exploratory multiple case study design with qualitative data collection methods. Qualitative studies are recommended when the dependent and independent variables in the phenomenon of interest are not well understood (Creswell, 2003) and when there is a tight interconnection between the phenomenon of interest and their organizational contexts (Yin, 2003). The three case sites included Connecticut, New Mexico, and Texas. The selection of the cases aimed to contrast two major collection building approaches in the existing Web archiving and preservation initiatives selective collecting and bulk collecting (California Digital Library, 2003). Connecticut and New Mexico use OCLC DA to build their digital depositories and they represent the selective collecting cases. The TRAIL/EDP in Texas, on the other hand, uses a typical bulk collecting, GILS-based approach, and employs a home-grown technological infrastructure. The investigation aimed to compare the methods used by the libraries to discover and identify Web publications, the selection criteria and collection development policies, and the work arrangement for item selection. The sources of data include semi-structured interviews with the librarians at the three case sites, internal documents supplied by the interviewees, and the information publicly available from the libraries Web sites. The initial interviews took place in December Two librarians from each case site participated in the telephone conference interview sessions. Each session lasted approximately one hour. The interview protocol is included as appendix A. Follow-up questions were answered mainly via s, except for one case site the

4 librarians agreed to receive a second interview in February Three librarians (a state document librarian and two catalogers) participated in the follow-up interview. Totally seven librarians from the three case sites participated in four telephone interviews. The interviewees included the state documents librarians, catalogers, and in one case site, the system librarian managing the electronic depository system. Totally seven librarians participated in four interview sessions. The interviewees were all actively involved in the digital state depositories. The state document librarians responsibilities included managing the state depository programs, developing digital depository collections, and providing public services on government information. The catalogers were responsible for cataloging and organizing state collections and were heavily involved in the overall collection development processes of the digital depositories. In one case site, a system librarian was in charge of the daily operations of the digital depository. Data analysis employed comparative tables (Miles & Huberman, 1994) to compare the legal contexts, technological infrastructures, collection development policies, selection practices, and methods used to discover and identify publications. The table of case comparisons is included as appendix B. Findings: Case Summaries Connecticut Although the Connecticut statutes do not explicitly include Web-based information resources in definitions of state publications, the state library has in practice extended the existing definitions to collect and preserve Web publications (Connecticut State Library, 2003a). The library is also proposing a revision to the state statutes to expand the definition of state publications to include those produced in electronic or other intangible format to encompass any publicly available publications issued or disseminated by a state agency. (Connecticut State Library, personal communication, January 10, 2005) The library began to identify and catalog born-digital publications in July 2000, and in July 2002, the library started to use OCLC DA. As of October 2004, the Connecticut Digital Archive contained over 4,000 state publications (Connecticut State Library, 2003b). The Connecticut Digital Archive collection development guidelines specify that a born-digital publication should be cataloged and preserved if it displays one or more of the following characteristics: 1) it provides access to information produced by one or more services inherent to the agency; 2) it is legislatively, judicially or administratively mandated; 3) it presents the results of a state task force, commission, or of committee work; 4) it is intended for public dissemination; or 5) it includes content that meets the criteria of substantiveness, comprehensiveness, and uniqueness (Connecticut State Library, 2003c). The library is trying to develop a systematic and effective methodology to discover and capture Web publications within agency Web sites. The librarians particularly look to capture and preserve publications required by state statutes, such as annual reports, financial reports, or statistical compilations. Currently, the librarians manually search in each agency site by using the agency search engine for new publications. For serial publications, OCLC DA allows the tagging of all cataloged serial titles so that they can run periodic reviews and harvest new issues serially. For monograph publications, searches are often limited to find PDF files because the librarians have observed that agencies tend to convert a print publication or produce a publicationlike document in PDF format. According to the interviewees, the site searching and inspection are not yet a fully formalized routine in their work practices. The tasks of identifying and selecting Web publications are scattered among several staff. Document librarians, reference librarians, and catalogers search agency sites and alert each other about the existence of publications potential for inclusion. The staffers who search agency sites compile lists of publications and send them to the unit head of the Government Information Services for selection review. The searches can yield between 1 and 50 publications from each agency, depending on the agencies and how the reviewers group the identified content. The unit head makes most of the selection decisions by applying the criteria outlined in the collection development guidelines. The time spent on reviewing each publication ranges from 1.5 to 10 minutes, depending on the content. The interviewees emphasized that they collect and catalog publications only. They do not attempt to preserve everything in the agency sites at this stage because that is out of the scope of the depository program, and they perceive that the OCLC DA is designed for digital documents at publication level rather than for a single webpage or an entire Web site. Further, staff perceived that agencies were less likely to voluntarily produce

5 publications not required by laws or by official obligations. As an interviewee commented, nobody in an executive branch is going to write a publication unless that is part of the official mission. Therefore, they capture and catalog only those publications the agencies are supposed to publish. The subject of the publication content is not a major selection criterion. When there is a need to evaluate the substantiveness of a publication, the decision is made on the professional judgment of the librarians. For example, a publication containing solid information about the state would be considered of more lasting value than one intended to present and advertise the image of a state politician. Review and selection of publications requires significant resources and the library would like to have the state agencies report their Web publications on an annual basis. Some agencies are very cooperative in reporting new publications, while others are less so. The librarians suspect that agencies with severe staff shortages are less cooperative. New Mexico The New Mexico s state statutes provide a basis for collecting digital state publications. Statutes define publications as any processed publication or other information funded in whole or in part by the state legislature or issued at the request of, or contracted out by, an agency regardless of physical format or characteristics. (NMAC ) Under the legal definition, the state library, in consultation with the 24 depository libraries, formulated a set of collection criteria to guide the selection of Web publications. In addition to meeting the legal definition, a Web publication for the NM Digital Archive needs to conform to one or more of the following conditions: 1) it continues a publication previously produced in paper or other traditional format but is currently available in electronic format only; 2) it is a born digital publication; 3) it is published both in paper and electronically, but is not distributed to all depository libraries in paper format; and 4) the publication must be produced in a format that OCLC DA can accommodate (New Mexico State Library, personal communication, December 6, 2004). The state library began to use OCLC DA in May The discovery and identification of the Web publications relies heavily on the state document librarian s knowledge of the agency Web sites. Due to her frequent use of agency sites in the reference work, she perceives that she can usually quickly identify new publications. In addition, the library also has some formalized searching activities integrated into work routines. For example, the library has a master list of all state serial publications used to continuously harvest the new issues. For monographs, the state document librarian and her assistants have tried to conduct searches on a quarterly basis although this practice has been difficult because of a recent loss of staff. The work of searching, identifying, and harvesting publications is spread out among the library staff. Tasks are mixed and scattered among the librarians involved, and as the participants described, everyone does a little bit everything. The state document librarian, public services librarians, and some library assistants do agency site searching to identify publications, but the state document librarian makes nearly all the major selection decisions, except decisions about publications voluntarily filed by agencies. Catalogers and others sometimes alert the document librarian about the existence of a new publication. The document cataloger and the librarian in charge of claiming print publications do publication harvesting and ingestion. Also similar to Connecticut, the major criterion for identification and selection is the publication type rather than subject area. When there is a need to assess the content of a publication, the decision is usually made on the selector s judgment taking into consideration whether it is of value to the state citizens, and whether it is of any long-term value. A noteworthy feature of the New Mexico selection decisions is that they do not limit the collection to born digital publications. They sometimes seek to collect the Web equivalents to print publications as substitute copies for depository distribution. This practice stems from agency failure to submit sufficient printed copies. Further, they sometimes find that the Web version of a publication provides more functionality for users for example, when an HTML publication provides interactivity; or, when the Web version contains colored illustration, while the print version is merely a black-and-white photocopy. The library is seeking to legally mandate the agency filing of electronic publications. They expect this change would eliminate any agency confusion about what ought to be filed and thus enhance the comprehensiveness of the electronic depository.

6 Texas The TRAIL/EDP program in Texas operates in very different legislative, organizational, and technological contexts from Connecticut and New Mexico. First, agency cooperation with the service is mandated. In 1995, the state created TRAIL, an indexing and locator service, to facilitate the access and discovery of state government information. Agencies have been required to include metadata in their Web publications and designate an agency liaison to support the TRAIL operations. In 2000, the state launched the Electronic Depository Program (EDP) built on the existing TRAIL service (Hartman & Condrey, 2004). Although the EDP component was not released to the public at the time of manuscript preparation, the collection had accumulated over 172 Gigabytes of state publications at the end of 2004 (Texas State Library, personal communication, December 7, 2004). Texas also differs from the other cases in that the TRAIL/EDP program and the state depository library program for print publications are separate, although the state document librarian works closely with both the print and the electronic depository programs. A fulltime librarian is in charge of the TRAIL/EDP daily operations with some assistance from the library s information technology staff. While Connecticut and New Mexico provide access to both print and Web publications via their library catalogs, the Texas State Library does not intend to integrate TRAIL/EDP records into its library catalog; rather, TRAIL/EDP will work as the access service only for Web publications. Also unlike Connecticut and New Mexico, TRAIL/EDP relies on bulk collecting of the Web publications; therefore collection building involves little human intervention from TRAIL/EDP staff, but requires significant human intervention and decision-making from the agencies who must apply metadata to documents in order to trigger the harvest. The interviewees explained that publication-by-publication review and selection by the state document librarians was impossible in Texas due to the huge amount of government information produced by over three hundred state agencies. Because the bulk collecting approach relies heavily on the input from the publishing agencies, the library created a collection development plan to aid agency publication selection (Hartman, 2001). According to its methodology, agencies are required to include appropriate metadata at the publication level for all Web publications. In larger agencies, the task of tagging Web publications is often designated to the agency librarians. In other agencies, it is unclear who has responsibility for tagging documents. Because the collection development plan outlines the framework for agency staff to tag publications, it serves more as an inter-agency protocol than a selection policy. A Web crawler from the TRAIL/EDP system automatically identifies and harvests a Web publication based on the combination of the publication type (e.g., annual report, budget) and the MIME types (the file format, such as PDF, HTML) in its metadata. Currently, the plan lists 30 publication types and 8 MIME types for automatic harvesting. The plan also formulates schedules to periodically capture the snapshots of all state publications in order to preserve the versions/editions of the publications (Hartman, 2001). Due to the huge size of the collection, the documents librarians do not inspect the metatags of the harvested publications. Librarian review happens only when an anomaly occurs, and such review typically examines the overall tagging practices of an agency rather than focusing on a particular publication. For example, an anomaly in harvesting that produces too few publications, or feedback from library users about search results in TRAIL, might lead the librarians to discuss metataging practices with agency liaisons in order to suggest improvements. But in general, according to the interviewees, they do not police around for the adequacy of metatags. Although the interviewees perceived a good level of agency cooperation at this stage, they perceived that maintaining compliance to ensure the quality of the services would remain a continuing challenge. TRAIL/EDP seeks to preserve information of permanent value for researchers in addition to collecting publications that agencies are legally required to produce. However, due to the automatic harvest and the huge size of the collection, the librarians can rarely assess the substantive content of potential materials. Therefore, unlike the other state depositories, there might have been some unwanted information accumulated in the entire TRAIL/EDP collection. Theoretically, the librarians can discard unwanted information harvested into the collection if it doesn t meet the collection development criteria; but this is rarely done. According to the interviewees, the risk of storing unwanted digital information is low since storage space is not an issue. Discussions: A Comparison across the Cases Digital Repository Terminology First, the current terminology available in the literature was not adequate to describe the complexities of the

7 selection and capture methodologies and the relationships between depository programs and state agencies. Previous work differentiated between selective collecting and bulk collecting (California Digital Library, 2003), but this terminology does not fully reflect the range and variation in observed organizational and technological arrangements. Our data suggest that it is important to distinguish between (a) capture methodologies, (b) modes of selection, and (c) loci of selection control. The three cases employed two main types of capture methodologies which we refer to as bulk harvesting and dispatch harvesting. The Texas TRAIL/EDP employs a bulk harvesting approach that automatically seeks out tagged publications across a number of Web servers at scheduled intervals with little human intervention. In contrast, OCLC DA employs a dispatch harvesting approach in which selectors must identify the URLs and instruct the system to capture a specific publication at that known URL. We identified two major modes of selection from our data, which we refer to as active and passive selection. Dispatch harvesting requires active selection, in which the digital depository staff actively identify and select items for inclusion. Bulk harvesting involves passive selection, in that publications are automatically added to the collection based on metadata specifications. Staff rarely review the incoming items. Our analysis also identified three different loci of selection. For example, library selection is where the state library staff do most of the selection for Web publications and control the collection development of the state digital depositories. In contrast, liaison selection is where some party an agency librarian, liaison, or other highly trained agency information professionals select for an entire agency. In the third type, creator selection, the agency staffer that creates a publication assigns the metadata, thereby selecting the document for inclusion in the repository. The capture methodologies, modes of selection, and loci of selection control are deeply intertwined. Further, the three are also shaped by the technological capacity and statutory support available to the digital depositories. Certain combinations require certain technologies, and certain technologies dictate certain arrangements. For example, using off-the-shelf solutions like OCLC DA allows state libraries to start collecting Web publications right away without having to invest on technologies or to negotiate for agency cooperation. But the externally-acquired solutions usually result in the combination of dispatch harvesting and active selection by library selectors that demands intensive selection resources and manpower from the libraries. Reliance on Publication Type to Guide Selection Another finding is that, in all the three case sites, selectors relied on publication types to identify publications from the entire content of an agency site. This finding is noteworthy because the continued reliance on these traditional publication types for selection decisions may overlook valuable Web-based information embodied in innovative digital genres, such as homepages, databases, interactive maps, applications, and multimedia documents. Also, selectors may more reliably identify certain publication types, such as annual reports, financial reports, or statistics, while overlooking other informational content that could have been published in lesser forms such as brochures, pamphlets, and fact sheets. The use of file formats to sensitize the manual searching processes or guide automatic harvesting may reinforces the bias. Data from this study suggest that the manual site searching may favor the discovery of publications in PDF format. But other research has shown that agency content creators are increasingly seeking to convert publications into HTML format to improve their usability (Eschenfelder, 2004a). Conversion to HTML may inadvertently cause a publication to be passed over in a selection scheme favoring PDF or word processing file types. Reliance on publication types and file types may result in a collection that favors the official, bureaucratic publications. Although these bureaucratic publications may provide a good deal of the organizational history, the common publications written for the lay citizens, such as brochures and pamphlets, may provide important historical evidence of the ways government agencies informed and educated the public (Ham, 1993). More research is needed to determine how much of such Web content is selected or harvested if they are not published in PDF or word processing formats. The Changing Relationships between State Libraries and State Agencies Perhaps one of the most significant changes brought about by new Web publication selection practices is the shifting role of document librarianship from passive receipt of publications to active selection of government information. In the past, when printing was the predominant means of publishing, government agencies had the control over the submission of their publications to depository libraries. Because agencies had the power to determine if an information resource comprised of a publication for depository distribution, some official publications became fugitive from the depository programs by the ways agencies defined them (Hernon &

8 Relyea, 1995; Quinn, 2003). In some models in the Web environment, this power has shifted from the publishing agencies to the state libraries. While the selection of government publications is primarily driven by legal requirements, all the selection criteria reviewed in this study left room for selector discretion particularly, for documents that fall outside of standard publication genres. The implications of this shift of selection power and the selection discretion some state libraries now enjoy are currently unknown. Initial data suggest that the change can positively result in more government publications in the state collections. Taking Connecticut as example, between 1998 and 2002, the state library depository program took in less than 200 publications each year (Connecticut State Library, 2003d). However, from 2002 to 2004, the library digitally archived more than 4,000 publications (Connecticut State Library, 2003b). But the shift in selection power from agencies to librarians could also lead some agencies to reduce the amount of discretionary information available on government Web sites, particularly if there is no mechanism by which agencies can request removal of a non-legally mandated publication from a depository. If agency managers wish to better control what discretionary publications about the agency are saved for perpetuity, they may do so by centralizing control over the production of content for Web sites, and reducing the amount of Web content produced (Eschenfelder, 2004b). This could have the unfortunate effect of reducing the overall amount of information available to the public via agency Web sites. The change also raises important questions about what should be selected. Data from this study suggest that subject topic was not a selection criterion. But given their selection discretion, some might argue that library selectors should seek to include more publications from controversial or otherwise important government programs because publications from these programs might attract more attention from future scholars. Further, it is not clear if state libraries desire this discretionary selection power. Differences in Who Selects The cases differed in terms of who selected documents for inclusion in the depositories. In Connecticut and New Mexico, librarians identify and select Web publications for the digital depositories. In contrast, in Texas the agency liaisons or publication creators define and tag a Web publication for the automatic inclusion in the state depository. These differences raise important questions about the possible consequences of librarians or agency liaisons/creators as selectors that future research should investigate. First, it is unclear what type of selection will result in a superior collection. On one hand, the content creators or the agency liaisons may have more the intimate knowledge necessary to judge the permanent value of a publication; this suggests that the passive selection model with a creator locus of control (in which agency staff create the publications and apply metadata) will result in better collections. However, one can also argue that resource limits within an agency may preclude adequate training of staff to correctly select and tag publications, and this may negatively influence the inclusion of, and the long term access to, the agencies publications. On the other hand, some may argue that librarian selection will result in superior collections as librarians have specialized training for collection development, understand the community needs, and appreciate the missions of the collections as historical record (Lane, 1981; Hernon & Purcell, 1982). This would suggest the superiority of library selection or certain liaison selection in which an agency librarian selects for the entire agency. The current data suggest some drawbacks to the active selection/librarian locus of control arrangement. The librarians in Connecticut and New Mexico expressed a desire for agencies to take more responsibility to identify and report Web publications. This suggests that state libraries may wish to retain their more traditional role. Librarians also expressed concern about their ability to keep up with and identify all relevant publications. They saw agency identification of publications as important to improving the quality of the collection. Another venue for further investigations is to examine selection practices within the theorizing of archival appraisal. Archival appraisal theory has a long history of debate about the roles that archivists and document creators play in the selection of archival materials (Ham, 1993; Craig, 2004). The application of such framework, however, needs to recognize the fundamental differences in the library and the archival professions e.g., the different conceptualization of the materials and the collections they manage and the perceived role they play in the society. The study results also pointed out a number of concerns related to the passive selection approach employed by Texas that deserve further inquiry. First, the effectiveness of the passive selection (employing liaison or creator selection) depends on the agencies understanding that the purpose of metadata implementation is to trigger the harvesting of a subset of webpages that qualify as publications for inclusion in a repository rather than simply compliance with the state metataging requirements. Texas librarians expressed concern that

9 metataging requirements would lead some agencies to simply tag all webpages with valueless tags in order to comply. Selectors in any of the models may be well trained or poorly trained thus affecting the quality of the resulting collection; but, it is likely that poor training would be more common in liaison or creator selection models. Previous research found that only a few Web content producers in government agencies were even aware of the state metadata requirements, much less trained or motivated to apply effective metadata (Eschenfelder, et al., 2002; Eschenfelder, 2004a). More research is needed to determine to what extent the differences in selection models affect collection quality. Conclusion This paper presented the major findings of an exploratory case study to investigate current selection practices in three state libraries. Growing agency use of the Web for distribution of information suggests that it is time to rethink the roles and responsibilities of government information librarianship in creating and maintaining access to digital government information. This study identified major challenges facing state libraries employing different collection development models, including both active selection and passive selection. The analysis of the three case sites created new vocabulary to better describe the differences between state programs, described the continuing reliance on publication type for selection, detailed the changing relationships between state libraries and publishing agencies, and raised important implications about the differences and the social consequences from who selects publications for inclusion in the state depository collections. ACKNOWLEDGMENTS The authors thank Louise Robbins, Ciaran Trace, and two anonymous reviewers for their constructive comments on the earlier versions of this manuscript. The authors also thank the three state libraries and the participating librarians for their time and input. REFERENCES Barnum, G. (2002). Availability, access, authenticity, and persistence: creating the environment for the permanent public access to electronic government information. Government Information Quarterly, 19(1), California Digital Library. (2003). Web-Based Government Information: Evaluating Solutions for Capture, Curation, and Preservation. Oakland, CA: California Digital Library. Retrieved September 15, 2004 from Cho, J., & Garcia-Molina, H. (2000). The evolution of the Web and implications for an incremental crawler. In Proceedings of the 26th International Conference on the Very Large Databases (pp ). San Francisco: Morgan Kaufmann. Clayton, T.J., & Gregory, S. (2003). The new face of government: citizen-initiated contacts in the era of e-government. Journal of Public Administration Research and Theory, 13(1), Connecticut State Library. (2003a). Definition of Electronic Connecticut State Documents. Retrieved January 18, 2005 from Connecticut State Library. (2003b). Connecticut Digital Archive. Retrieved January 18, 2005 from Connecticut State Library. (2003c). Guidelines for selecting Connecticut electronic state documents for description and preservation. Retrieved January 18, 2005 from Connecticut State Library. (2003d). Connecticut State Depository Library Program. Retrieved January 18, 2005 from Craig, B. (2004). Archival Appraisal: Theory and Practice. Munchen: K.G. Saur. Creswell, J.W. (2003). Research Design: Qualitative, Quantitative, and Mixed Method Approach. Thousand Oaks, CA: Sage. Eschenfelder, K.R. (2004a). Behind the Web site: an inside look at the production of Web-based textual government information. Government Information Quarterly, 21, Eschenfelder, K.R. (2004b). How do government agencies manage the text content on their Web sites? A framework to compare Web content management practices. Library and Information Science Research, 26(4), Eschenfelder, K.R., et al. (2002). State Agency Web Content Management: Current Practices, Best Practices, and Policy Issues. Madison, WI: University of Wisconsin-Madison, School of Library and Information Studies. Retrieved June 30, 2005 from Government Printing Office. (1996). Study to Identify Measures Necessary for a Successful Transition to a More Electronic Federal Depository Library Program. Washington, D.C.: GPO. Ham, F.G. (1993). Selecting and Appraising Archives and Manuscripts. Chicago: American Society of Archivists.

10 Hartman, C.N. (2001). Collection Development Plan: Selecting Texas State Documents to be Included in the Electronic Depository Program, Version 1.0. Austin, TX: Texas State Library. Retrieved January 18, 2005 from Hartman, C.N., & Condrey, C. (2004). TRAIL: from government information locator service to electronic depository program for Texas state publications. DTTP: Documents to the People, 32(2), Hernon, P., & Purcell, G.R. (1982). Developing Collections for U.S. Government Publications. Greenwich, CT: JAI Press. Hernon, P., & Relyea, H.C. (1995). Government publishing: past to present. Government Information Quarterly, 12(3), Horrigan, J. B. (2004). How Americans Get in Touch with Government. Washington, D.C.: Pew Internet & American Life Project. Retrieved June 30, 2005 from Illinois State Library (n.d.). Capturing E-Publications (CEP) for Public Documents. Retrieved June 30, 2005 from Illinois State Library (n.d.). Preserving Electronic Publications (PEP). Retrieved June 30, 2005 from: Lane, M.T. (1981). State Publications and Depository Libraries: A Reference Handbook. Westport, CT: Greenwood. Larson, E., & Rainie, L. (2002). The Rise of E-Citizen: How People Use Government Agencies Web Sites. Washington, D.C.: Pew Internet & American Life Project. Retrieved June 1, 2004 from Martin, K.E. (2004). Publishing trends within state government: the situation in North Carolina. Journal of Government Information, 30(5/6), Matthews, R.J., et al. (2003). State-by-State Report on the Permanent Public Access to Electronic Government Information. Chicago: American Associations of Law Libraries. Retrieved October 1, 2004 from Miles, M.B., & Huberman, A.M. (1994). Qualitative Data Analysis. 2nd ed. Thousand Oaks, CA: Sage. Mullen, A. (2001). GILS metadata initiatives at the state-level. Government Information Quarterly, 18(3), NMAC Retrieved January 18, 2005 from OCLC. (2004). OCLC Annual Report 2003/04. Dublin, OH: OCLC. Retrieved January 18, 2005 from OCLC. (2003). Learning to Use Digital Archive Web Archiving Tools. Retrieved January 18, 2005 from OCLC/RLG Working Group. (2001). Preservation Metadata for Digital Objects: A Review of the State of the Art. Retrieved June 30, 2005 from: Oregon State Library. (n.d.) Criteria for Selecting Items for Electronic Archiving. Retrieved June 30, 2005 from Pearce-Moses, R., & Kaczmarek, J. (2005). An Arizona model for preservation and access of web documents. DTTP: Documents to the People, 33(1), Quinn, A.C. (2003). Keeping the citizenry informed: early Congressional printing and the 21st century information policy. Government Information Quarterly, 20(3), Reddick, C.G. (2005). Citizen interaction with e-government: from streets to servers? Government Information Quarterly, 22(1), RLG, & OCLC. (2002). Trusted Digital Repository: Attributes and Responsibilities: An RLG-OCLC Report. Available at: Steenbakkers, J. (2000). Setting Up a Deposit System for Electronic Publications: the NEDLIB Guidelines. Amsterdam: Koninklijke Bibliotheek. Van der Werf, Titia. (2000). The Deposit System for Electronic Publications: A Process Model. Amsterdam: Koninklijke Bibliotheek. Yin, R.K. (2003). Case Study Research: Design and Methods. 3rd ed. Thousand Oaks, CA: Sage.

11 Appendix A: Interview Protocol 1. Please describe how your library identifies potential contents from state agencies web sites for inclusion in the state government information repositories. If it is done by automatic harvesting, please describe how it works. (1) Scope of harvesting? (2) Frequency of harvesting? (3) Depth of harvesting? (4) Formats included/excluded? (5) Other factors? 2. Please describe the current selection criteria or guidelines for selection in your library, if there s any. (1) What do the selection criteria address? (e.g.: individual user needs / community interests / historical value / subject approach / legal requirements / others?) (2) What are/were the major concerns when the criteria/policies were being developed? (3) How do the selection criteria/policies work so far in your experiences? What are the advantages and limits of the current selection criteria/policies? (4) What additional criteria are being / might need to be considered, if there s any? 3. Please describe the work arrangement and procedures of content selection/evaluation: (1) Who are the primary selectors in the current project/program? (2) Who else are involved in the selection of contents, if there s any, and how they are involved in the selection process? (e.g.: the publishing agencies / bibliographers / catalogers / collaborating libraries / others?) (3) How do the primary selectors assess the value or the significance of contents? (4) What types of information are easy to evaluate? What types of information are difficult to evaluate? (5) What are the major factors that facilitate or interfere the current practices of content selection (e.g.: technological, policy issues, inter-organizational, etc.?) 4. In summary, what do you considered successful in your experiences with building up the electronic collection? What are the major problems, difficulties, or challenges related to collection development that you ve come across (or are working on) in addition to the questions discussed above?