Lessons Learned from the Interagency GI Bill Data-Linkage Project

TABLE OF CONTENTS

CHAPTER

Executive Summary
Conceptualizing The Interagency GI Bill Data-Linkage Project For Success
Challenges And Obstacles The Interagency Project Encountered
Strategies The Interagency Project Used to Encourage Federal Agencies to Share Data
Interagency Data Sharing Also Encouraged by Federal Initiatives
Conclusions

APPENDIX

Research Questions and Key Findings from Five Reports of the Interagency GI Bill Data-Linkage Project
Individuals Contacted for This Report
Noteworthy Interviewee Comments on Interagency GI Bill Data-Linkage Project’s Challenges in Dealing with Agency Bureaucrats
Chronology of Selected Milestones of the Interagency GI Bill Data-Linkage Project and Evidence Act Initiatives, 2016 to 2025
Timeline for Datasets Analyzed by Interagency GI Bill Data-Linkage Project
Bureaucratic Challenges to Data Sharing Identified by 2017 Commission on Evidence-Based Policymaking
Statutory Ambiguity and Lack of Consistent Support for Evidence Building Identified by the 2017 Commission on Evidence-Based Policymaking
Key Provisions of the Foundations of Evidence-Based Policymaking Act of 2018
Comparison of Analogous Challenges: Data Sharing Working Group and Interagency GI Bill Data-Linkage Project
Synopsis of Comments on a Draft of this Report Provided by the Individuals Contacted

This report was prepared in part with support from the Gates Foundation. The views expressed here do not necessarily reflect positions or policies of the Foundation.

ABBREVIATIONS

CDO               Chief Data Officer
CFO                Chief Financial Officer
CIO                 Chief Information Officer
DOD               U.S. Department of Defense
ED                   U.S. Department of Education
FERPA           Family Educational Rights and Privacy Act
GAO               Government Accountability Office
IPEDS             Integrated Postsecondary Education Data System
IRS                  Internal Revenue Service
NCES              National Center for Education Statistics
NSDS              National Secure Data Service
OMB               Office of Management and Budget
PII                   personally identifiable information
VA                  U.S. Department of Veterans Affairs

Note: To minimize the use of abbreviations, we spelled out entities such as the Commission on Evidence-Based Policymaking and Data Sharing Working Group. When such organizations were referred to frequently in certain sections of this report, however, we adopted the following convention: Commission on Evidence-Based Policymaking (2017 Commission) and Data Sharing Working Group (Working Group). Moreover, the interagency GI Bill data-linkage project will be referred to frequently as the interagency project.

GLOSSARY

This glossary contains terms, positions, organizations, and statutes often referred to in this report, focusing on their relevance to the interagency project.

Advisory Committee on Data for Evidence- Building	Established by the Foundations for Evidence-Based Policymaking Act of 2018 (legislation enacted to promote the sharing of federal administrative data) this advisory committee laid out a roadmap for the creation of a National Secure Data Service demonstration.
Armed Forces Qualification Test	DOD administers this test to all military recruits prior to enlistment to help determine their military occupations. The test covers arithmetic reasoning, mathematics knowledge, paragraph comprehension, and word knowledge.[1] The interagency GI Bill data-linkage researchers used the test scores to measure academic preparedness.
Center for Administrative Records Research and Applications	The Center is an organization within the Census Bureau charged with the strategic re-use of administrative data from federal, state, and third-party providers. Through record linkage and statistical matching, the Center extends the Bureau’s demographic and socioeconomic research capabilities. In 2016, the Center approved more than 10 pilot projects to demonstrate the benefits of data sharing and linkages. The interagency GI Bill data-linkage project was one of those pilots.
Chief Data Officer	New position established at federal agencies by the Evidence Act of 2018 to foster the use of administrative records for policymaking. Chief Data Officers’ Council coordinates their activities across the federal government.
Chief Financial Officer	Statutory position at federal agencies created in 1990, reporting to the agency head, who is responsible for developing and implementing financial management systems. Chief Financial Officers’ Council coordinates their activities across the federal government.
Chief Information Officer	Statutory position at federal agencies created in 1996 responsible for IT management and security. Coordinates with Chief Financial Officers on their budgets. A Chief Information Officers’ Council coordinates their activities across the federal government.
Commission on Evidence-Based Policymaking	Commission established by statute in 2016. The Commissions 2017 report identified the obstacles to data sharing and laid out a strategy for using federal administrative data for policymaking.
Confidential Information Protection and Statistical Efficiency Act	2002 statute outlining the protections applied to data collected for statistical purposes by federal agencies such as the Census Bureau and the Bureau of Labor statistics. The Act included very narrow use exceptions for nonstatistical purposes. Title III of the Evidence Act of 2018 reauthorized the Act and contained provisions to expand access to statistical data containing PII for evidence-building while strengthening and preserving confidentiality protections.
Defense Manpower Data Center	DOD component that maintains personnel, manpower, training, financial, and other data. This data catalogues the history of personnel in the military and their families for purposes of healthcare, retirement funding, GI Bill eligibility, and other administrative needs. One of the datasets it maintains is the scores of new recruits who take the Armed Forces Qualification Test.
Evidence Act of 2018	Act’s three titles collectively direct federal agencies, under the coordination of OMB, to make agency data accessible and to use such data to support evidence-based policymaking. Creates new officials to further these objectives, including Chief Data and Evaluation Officers and a Statistical Official, and directs OMB to create an Advisory Committee on Data for Evidence Building. GAO must periodically report on progress and recommend actions to support agency capacity for evidence-based policymaking, and OMB must issue regulatory guidance necessary for implementing the Act. See Appendix H of this report for a summary of its key provisions.
Federal Student Aid	ED component that administers the federal student aid program.
Family Educational Rights and Privacy Act	Restricts the release of students’ educational records without their explicit consent but enumerates both exemptions and exceptions.
Integrated Postsecondary Education Data System	ED’s publicly available dataset that, among other things, reports aggregate data on enrollment and outcomes from colleges that participate in the federal student aid program.
Joint Statistical Project	Type of agreement that allows data sharing for the purpose of statistical research for evidence building between agencies.
National Center for Education Statistics	Research arm of ED that among other things conducts surveys of students attending schools that participate in federal student aid in order to report aggregate outcomes.
National Secure Data Service demonstration	Five-year demonstration project authorized by statute in 2022 with goal of developing, refining, and testing models to inform the full implementation of a data service as recommended by the 2017 Commission on Evidence-Based Policymaking. The vision for the demonstration is for a new entity that provides coordination and capacity-building services for data sharing.
National Student Clearinghouse	Nonprofit organization to which participating institutions voluntarily provide enrollment and outcome data on all students pursuing postsecondary education credentials, including veterans.
Office of Enterprise Integration, Data Governance and Analytics	VA component that is home to the National Center for Veterans Analysis and Statistics—VA’s hub for providing analytics on the current and future veteran population and usage of VA programs to support evidence-building for planning and decision-making. It owns no data but rather (1) integrates data assets from across the department to be used for enterprise program evaluation and reporting, and (2) provides data products focused on the veteran population and their use of VA benefits and services.
Predominate Purpose Statement	Type of agreement that allows data sharing for the purpose of improving agency operations.
Privacy Act	Enacted in 1974, the Privacy Act governs federal agencies’ access to and use/disclosure of information containing PII. The Act provides for exemptions for records not subject to the Act’s protections and exceptions applicable to information not requiring prior written consent from individuals prior to disclosure.
Title 13 of the U.S. Code	Describes the authorities of the Census Bureau to carry out its statutory responsibilities with respect to conducting the decennial census and spells out the privacy protections, exemptions, and exceptions that apply to the data it collects for that purpose.
Title 26 of the U.S. Code	Describes the authorities of the IRS to collect taxes and spells out the privacy protections, exemptions, and exceptions that apply to the data it collects for that purpose.
Veterans Benefits Administration	VA component that includes Education Service, which manages GI Bill education benefits’ programs.

I. EXECUTIVE SUMMARY

This report explores the lengthy interagency GI Bill data-linkage^[2] project, housed at the U.S. Census Bureau, which brought together administrative data from four federal agencies and the National Student Clearinghouse in order to analyze veterans’ outcomes under the Post-9/11 GI Bill. Hereafter, the project is referred to as simply the interagency project. The report also explores the contemporaneous congressionally directed effort to improve federal interagency data sharing.

The interagency project was able to demonstrate that when agencies share data policymakers can obtain answers to important questions that help to improve federal programs. Through careful collaboration across agencies, the interagency project was able to draw on multiple sources of data covering an unprecedented dataset size of 2.7 million veterans (every enlisted veteran who was eligible for GI Bill benefits who separated from the military as of June 30, 2018, and was age 65 or younger as of December 31, 2019). Data from the U.S. Department of Veterans Affairs (VA) and the Census Bureau were augmented with earnings data from the Internal Revenue Service (IRS) (the gold standard for researchers on earnings) and data from the U.S. Department of Defense (DOD) which enabled the team to study veterans of similar rank, military occupation and other characteristics, as well as service members’ academic preparation at the time of enlistment (an important “control” for the assessment of later student outcomes).

This unprecedented interagency data sharing resulted in five important reports that presented the first comprehensive analysis of GI Bill usage and outcomes – including the characteristics of veterans using the Post-9/11 GI Bill and how their educational and labor market outcomes differ by demographic and military characteristics as well as by institutional characteristics, type of educational program, and field of study. The team found, for example, that veterans’ six-year graduation rates at the associate and bachelor’s levels are nearly double that of financially independent students (who are considered a comparable population for veterans, rather than 18 year olds whose parents are paying for college), and that an institution’s instructional spending levels had a high degree of correlation with veterans’ labor market success. The team was also able to conduct deep dives on specific population groups and regions of the country, as well as on the earnings of veterans who did not use their GI Bill.

VA has already indicated it is using the project’s Post-9/11 GI Bill outcomes data to enhance its delivery of benefits. Moreover, the interagency project helped to build “muscle” for future federal data sharing as evidenced by new data-sharing agreements, initiatives, and legislation that have occurred since late 2023 when the project’s draft reports were complete but still under review, and agencies had been briefed on their pending release.

The reasons for the interagency project’s ultimate success can be summed up in two words—persistence and perseverance. While the interagency project successfully delivered the first data on and analyses of veterans’ Post-9/11 GI Bill outcomes, it took about 8 years from conception to the release of its first three reports. The key challenges that confronted the interagency project were a bureaucratic culture at federal agencies, statutory privacy concerns raised by agency lawyers, and unpredictable situational delays such as agencies’ higher priorities and COVID. Privacy concerns raised by lawyers at the U.S. Department of Education (ED) ultimately doomed its participation in the project, even though ED had previously shared data with other interagency efforts when pressured to do so by the White House or congressional mandates. A bureaucratic culture led to protracted negotiations between the Census Bureau and VA, DOD, IRS and the National Student Clearinghouse (Clearinghouse), from which the interagency project purchased data on student records. That bureaucratic culture also caused delays in the data analysis and reporting phases of the interagency project. The challenging and time-consuming process of negotiating data-sharing agreements, however, was not unique to this groundbreaking endeavor. Many of the same obstacles were documented in reports about the challenges in linking federal data by the Commission on Evidence-Based Policymaking and the Data Sharing Working Group, released in 2017 and 2022, respectively.

Given that the analytical framework and knowledge base now exist for analyzing GI Bill outcomes, researchers hope the work of the interagency project will not be a one-time endeavor. Both policymakers and future cohorts of veterans would benefit from a periodic refresh of the data, which would require modest funding. A refresh model already exists in the automatic updating of earnings data on ED’s College Scorecard under a longstanding agreement between ED and IRS.

Because the interagency project’s lengthy timeframe paralleled congressional efforts to embrace federal data-sharing as the norm in policymaking rather than the exception, this report also examines the status of initiatives required by the Foundations for Evidence-Based Policymaking Act of 2018 (hereinafter referred to as the Evidence Act).[3] Progress under the Evidence Act is apparent in the development of new or proposed data-sharing tools and processes. Key challenges, however, are also apparent in the lack of dedicated resources for the new evidence-building infrastructure. Moreover, the inconsistent legal authorities that contribute to risk aversion by agency lawyers have never been addressed and continue to impede data-sharing projects. Evidence of the continuing impact of statutory ambiguity are new privacy concerns raised by ED about sharing even aggregate data that contains no personally identifiable information (PII).

Perhaps the most serious challenge to the Evidence Act’s implementation—and to future efforts at interagency data sharing—is agency cultures that focus on the perceived risks in data sharing and are not yet fully invested in the benefits of evidence-based policymaking for the public good. Changing agency cultures is likely to be an ongoing endeavor that requires strong leadership and oversight not only at the agency level but also at the Office of Management and Budget (OMB), which was assigned a major coordinating role by the Evidence Act but has been slow to issue the guidance required by the Act. An important step could be adopting the strategies used by the interagency project team to incentivize data sharing, such as identifying and including data elements that would enhance each agency’s mission and being patient but persistent while seeking the intervention of higher-level agency officials to break bureaucratic logjams.

II. BACKGROUND

The interagency GI Bill data-linkage project was initially conceived by current and recent federal policy officials, with several key goals: (1) strengthening the government’s evidence-building capacity, (2) answering long-standing policy questions on the educational and labor market outcomes of veterans who used the Post-9/11 GI Bill, a $100 billion investment in its first decade, and (3) examining how those outcomes varied by demographic group and military characteristics. Answering these questions through a traditional evaluation—conducted by researchers using surveys to collect data from individuals—would be cost-prohibitive and would not yield reliable results if survey responses were low. By contrast, a study that used linked administrative data from multiple sources could potentially produce more comprehensive and reliable results at lower cost and in less time.

In summer 2016, a meeting was convened that included researchers and data experts from the Census Bureau, VA, DOD, OMB, the American Institutes for Research, and the Clearinghouse to discuss the feasibility of linking data on veterans from different agencies and from the Clearinghouse (for student records). The Census Bureau suggested using its existing data-linkage infrastructure managed by its Center for Administrative Records Research and Applications to securely combine data provided by the other sources to produce aggregate statistics. The group agreed that was a feasible option and set the planning process in motion.

The interagency project’s planners also viewed it as a test case (“proof-of-concept”) to demonstrate the ability of linked administrative data to answer policy-relevant questions; the scalability of interagency data cooperation; and the usefulness of the platform at the Census Bureau to combine data from multiple agencies. It was hoped that the project also could inform the recommendations of the newly established, bipartisan Commission on Evidence-Based Policymaking, created by an act of Congress.

After 8 years, the interagency project published the first comprehensive analysis of postsecondary access and success for enlisted veterans using the Post-9/11 GI Bill. This analysis was made possible by the unprecedented sharing of administrative datasets among four federal agencies: VA, DOD, the Census Bureau, and IRS. The research team was able to link these federal datasets with enlisted veterans’ student outcomes purchased from the Clearinghouse. The interagency project team’s final reports:

contained detailed data on enlisted veterans’ usage of the Post-9/11 GI Bill and their educational and labor market outcomes—data that policymakers in both Congress and federal agencies previously lacked;
gave current and future beneficiaries (and taxpayers) compelling insights on the institutional sectors and career paths that provide the best payoff for their hard-earned educational benefits; and
provided data that could improve the delivery of the Post-9/11 benefit to veterans, both those who used their education benefits and those who did not.

From initiation to the publication of findings, however, the interagency project took longer than had been anticipated—roughly 8 years, from mid-2016 through mid-2024. Three reports were published in 2024 but the publication of two additional reports was delayed until February 2025 (see Appendix A for a summary of the five reports). As a result, the interagency project may not have succeeded in its ancillary goal of demonstrating that federal agencies can quickly share data to answer questions about programs’ outcomes. This report establishes a chronology for the interagency project that helps to identify the challenges encountered in accessing, combining, analyzing, and reporting on the datasets provided by the four agencies and the National Student Clearinghouse. Because the Clearinghouse was indispensable to producing outcome data, further references to agencies throughout this report include the Clearinghouse even though it is a nonprofit organization and not a federal agency.

Since the interagency project’s lengthy timeframe paralleled congressional efforts to improve policymaking through federal data sharing, this report also examines the status of initiatives required by the Evidence Act of 2018. Because of its many moving parts and its slow but evolving maturation, assessing overall implementation of the Act is daunting and perhaps impossible. The available reports we analyzed provide valuable implementation insights, but they tended to be narrowly focused on a single element in the new ecosystem created by the Evidence Act or on a subset of agencies involved in implementing the Act’s initiatives.

Appendix B identifies the individuals interviewed in researching this report.

III. LACK OF COMPREHENSIVE DATA ON VETERANS’ OUTCOMES

In 2009, VA implemented a more comprehensive and generous educational benefit—the Post-9/11 GI Bill.^[4] From fiscal years 2009 through 2018, almost $100 billion[5] was spent on eligible veterans and family members who had used the benefit. Researchers’ attempts to analyze the outcomes of veterans using the Post-9/11 GI Bill identified numerous shortcomings in the data available.[6]

Schools not required to report academic progress. The legislation establishing the Post-9/11 GI Bill did not require the collection and reporting of outcome data to demonstrate the returns from providing educational benefits to veterans or to identify needed improvements to the program. In fact, Congress did not provide VA with the authority to require schools participating in the GI Bill to report graduation rates until 2017.^[7]

2012 Executive Order requiring collection of student veteran outcomes never implemented. Executive Order 13607 (Principles of Excellence) issued in 2012 tasked several federal agencies with drafting student veteran outcome measures for eventual inclusion in what became VA’s GI Bill Comparison Tool, which was implemented in 2014.^[8] The draft outcome measures are still available through a link on ED’s College Navigator website.[9] An ED advisory group, however, concluded that requiring schools participating in both the GI Bill and federal student aid to report data on retention and graduation rates for veterans and servicemembers “was not feasible at this time and needs further study…. Further VA has plans to capture student outcome data in the future.”^[10] According to the advisory group’s report, minimizing the administrative burden on institutions was also a consideration in its decision.

GI Bill Comparison Tool still does not include veteran outcomes. Despite VA’s goal of providing a search tool to help veterans make an informed choice about where to use their educational benefits, the GI Bill Comparison Tool contains no veteran-specific outcome data. Rather, it focuses on attendance costs at, and veteran-focused services provided by, participating schools, allowing beneficiaries to search on a particular school or schools near where they live. In 2019, VA stopped showing College Scorecard graduation, retention, and earnings data for all students receiving federal student aid.^[11] VA attempted to calculate GI Bill graduation rates for schools that voluntarily reported beneficiary completions but stopped because of “the overwhelming demand from schools.”^[12] Reportedly, institutions were concerned that VA was undercounting graduations because it captured credential completion only for those individuals using benefits in the term that they earned a certificate or degree. In fact, all of VA’s administrative data reflects only those beneficiaries who are using GI Bill benefits. Veterans who exhaust their benefits before they graduate or veterans who enroll in a free community college program are missing from VA’s Comparison Tool dataset.

Federal ban on student unit record system inhibits outcome research. A provision in the 2008 Higher Education Act reauthorization prohibits ED from setting up a database that could collect and track individual-level data on all postsecondary students, also known as a student unit record data system.^[13]^,[14] Such a database could track the numerous students who attend several institutions before they drop out or earn a degree. Not only are such students difficult to track across institutions, but those who do not graduate likely end up being reported as noncompleters by some of the institutions they attended. A student-level data system would also benefit existing state systems—e.g., Texas, Virginia, and Minnesota[15]—which include only individuals who attend one of their state’s schools and are employed in-state. The inclusion of a student unit record system in the Higher Education was opposed on privacy and accountability grounds by the National Association of Independent Colleges and Universities. Because of the student unit record ban, a database maintained by the National Student Clearinghouse, a private nonprofit entity, is the only source of student outcome data, irrespective of whether students receive federal student aid, enroll in only one institution, attend multiple institutions, or ultimately earn a credential.

Attempts to overturn the ban on a student unit record system have been unsuccessful. A bill to reverse the 2008 ban, known as the College Transparency Act (Transparency Act) would create a longitudinal database that tracks all degree-seeking students throughout their postsecondary enrollments, including students who transfer to different institutions and part-time students. (Student transfers happen frequently, and the educational journey of such students is difficult to track.) This bipartisan proposal was first introduced in 2017 and was reintroduced in 2023.[16]^{, [17]} The Transparency Act also requires periodic matching with other federal data systems, including DOD, VA, IRS, and others. ED’s National Center for Education Statistics would become the repository for the database. One of the required data elements in the Transparency Act is military or veteran benefit status, which is defined as the receipt of federal military or veteran education benefits. A 2023 summary of the legislation highlights the following privacy protections: (1) using a secure, privacy-protected data network that relies on governmentwide, industry-supported security standards and data governance protocols; and (2) penalizing illegal data use, prohibiting the data’s use for law enforcement, and safeguarding PII.[18] A blog from the nonprofit policy think tank, Third Way, noted that the Act has the support of nearly half of senators and representatives from both parties and that about 200 organizations have endorsed the bill.[19]

Representative Virginia Foxx (R-VA), the former chair of the House Committee on Education and the Workforce, strongly opposes the College Transparency Act on privacy grounds.^[20] But, in 2024, Rep. Foxx modified her blanket opposition to a federal student unit record system when she introduced legislation, the College Cost Reduction Act, that would have created a student unit record system for students who received federal financial aid, DOD Tuition Assistance, or GI Bill educational benefits.[21]

Use of ED survey data to identify veteran outcomes has limitations. ED’s National Center for Education Statistics (NCES) does not have routine access through the Department to PII data on students who use federal student aid.^[22] Historically, NCES has fielded three surveys to collect outcome data on students who receive federal student aid. [23] The National Postsecondary Student Aid Survey conducted every 3-4 years is a snapshot of students at schools participating in federal student aid in the year the survey is fielded. The other two surveys are longitudinal and allow researchers to track the progress of a subset of the National Postsecondary Student Aid Survey’s students over 6 to 10 years: students who enrolled in postsecondary education for the first time (Beginning Postsecondary Students) or who have earned a bachelor’s degree (Baccalaureate and Beyond), capturing graduation, employment, earnings, demographics, and other variables. VA also shares PII with ED, which allows researchers to identify GI Bill beneficiaries in the survey sample. However, the surveys capture the use of benefits only during the year of the survey, not whether a veteran ever used benefits, and do not capture[24] the GI Bill benefit used.

College Scorecard data cannot be used to assess veteran outcomes. College Scorecard, ED’s postsecondary education search tool, allows prospective students and their families to compare schools that participate in federal student aid across a broad range of outcome metrics, including earnings data obtained from IRS. It does not, however, report any veteran-specific outcomes, even though ED’s NCES routinely receives PII data from VA, which it uses to identify and tag veterans in its survey samples.^[25]

IV. CONCEPTUALIZING THE INTERAGENCY GI BILL DATA-LINKAGE PROJECT FOR SUCCESS

The interagency GI Bill data-linkage project was a complex undertaking because of its ambitious objectives. However, it was conceptualized for success to help ensure the rapid negotiation of data sharing agreements and the release of reports as early as 2018. David Bergeron, a former ED official, laid out the challenges to such an endeavor in a 2016 paper published by the Center for American Progress.^[26]

“Breaking down the barriers between siloed information that the federal government already has about citizens who opt into (or out of) postsecondary education would improve policymaking, improve the services provided to citizens, and help potential postsecondary students make better choices about their education. But these barriers are perhaps the greatest challenge. Each federal agency has a unique reason for collecting and holding the data it has, and it has little incentive to share this information with other federal agencies.”

Ambitious interagency project scope. The project involved obtaining multiple datasets from VA, both for Post-9/11 enlisted beneficiaries who used their benefits and those who did not. The VA data were then matched to the longitudinal student outcome data from postsecondary institutions, housed at the National Student Clearinghouse,[27] to identify student veterans’ outcomes. The Clearinghouse then returned the matched data to VA, which provided the results to the interagency project’s team at the Census Bureau.

The Census Bureau has existing infrastructure to link incoming data from other federal agencies (for purposes of the decennial census). The interagency project’s team at the Census Bureau received VA data on veterans, which included outcomes obtained from the Clearinghouse; these data were then matched to DOD data on veterans’ military rank, occupation, tenure, service in hostile war zones, academic preparedness at enlistment, and other information;[28] and, finally, the data were linked to IRS earnings data. The project team sought but ultimately did not receive ED data on the federal student aid provided to Post-9/11 GI Bill beneficiaries, but did use publicly-available ED data on characteristics of educational institutions. Statistical tools were used to isolate the factors that appeared to be the most influential in shaping veterans’ outcomes, such as academic preparedness as measured by the Armed Forces Qualification Test upon enlistment, enlisted rank, or military occupation. PII was required to successfully link individual veterans in each of the datasets obtained, but all individual-level data were hidden using the Census Bureau’s Protected Identification Key discussed below. Table 1 summarizes the data provided by each agency to the interagency project’s team.

Table 1: Agency Data Used in Interagency GI Bill Data-Linkage Reports

Organization	Data requested
VA’s Office of Enterprise Integration, Data Governance and Analytics	List of all Post-9/11 GI Bill-eligible veterans; veteran demographic data from 2020 included in the U.S. Veterans Trends and Statistics data, and Veterans Benefits Administration’s Education Service files.
VA’s Veterans Benefits Administration	Veterans’ use of Post-9/11 GI Bill benefits through March 2020; a list of all payment records through fiscal year 2018; veteran demographic data from 2020 included in VA’s Benefits Administration’s Education Service’s files.
National Student Clearinghouse	Post-9/11 GI Bill-eligible veterans’ postsecondary enrollment and attainment records through June 2020.
DOD	Defense Manpower Data Center data from veterans’ Armed Forces Qualification Test (percentile upon activation), service experience (e.g., rank, military occupation), all activation and separation dates as of 2020.^a
IRS	W-2 and 1040 income and other data from tax years 2018 and 2019, including marital and dependents’ status, region, and zip code as of year of first separation.^b
Census Bureau	American Community Survey labor force participation from the 2019 American Community Survey, along with the Census Bureau’s crosswalk of Rural-Urban Commuting Area Codes and region for U.S. zip codes.
ED, Integrated Postsecondary Education Data System^c	Institution-level 2020 data on institution control and sector, as well as by-institution counts of students involved exclusively in distance education courses, merged with information on students’ institutions using the Clearinghouse’s Unit-ID Crosswalk Table. Institution-level data from 2009 to 2019 on institution sector; count of undergraduate and graduate degree seekers involved exclusively in distance education courses and total number of those degree seekers; and the total amount spent by the institution on instructional expenses divided by the total full-time equivalent for the institution. These variables were merged with information on students’ institutions using the Clearinghouse’s Unit ID Crosswalk Table and with the Unit ID field in the Veterans Benefits Administration’s beneficiary payments file.

Source: Excerpts from Radford, A.W, P. Bailey, A. Bloomfield, B.H. Webster, Jr., and H.C. Park (2024). A First Look at Post-9/11 GI Bill-Eligible Enlisted Veterans’ Outcomes. Washington, D.C.: American Institutes for Research, U.S. Census Bureau; and National Center for Veterans Analysis & Statistics, at the U.S. Department of Veterans Affairs, February. Radford, A.W., P. Bailey, A. Bloomfield, B.H. Webster, Jr., and H.C. Park (2024) Post-9/11 GI Bill Benefits: How Do Veterans’ Outcomes Differ Based on the Type of Education They Received? And How Are Veterans Who Have Not Used Their Education Benefits Faring? Washington, D.C.: American Institutes for Research, U.S. Census Bureau; and National Center for Veterans Analysis & Statistics, at the U.S. Department of Veterans Affairs, July. Bloomfield, A., A.W. Radford, P. Bailey, B.H. Webster, Jr., and H.C. Park (2024). Post‐9/11 GI Bill‐Eligible Enlisted Veterans’ Enrollment and Outcomes at Public Flagship Institutions, with a Focus on the Great Lakes Region. Washington, D.C.: American Institutes for Research, U.S. Census Bureau, and National Center for Veterans Analysis & Statistics at the U.S. Department of Veterans Affairs, July. These reports are available at https://www.air.org/project/study-post-911-gi-bill-student-outcomes. Appendix A contains a summary of key findings from each report.

^aArmed Forces Qualification Test scores are used by the military services to determine enlistment eligibility and to assign applicants to specific occupational specialties, such as infantry, intelligence, and engineering.

^bIRS provided earnings data for 2018, 2019, and 2020. The 2020 data, an outlier because of COVID, was not used.

^cNo agreement was reached with ED to share data on the federal student aid received by veterans. IPEDS is in the public domain and did not require a data-sharing agreement with ED.

Interagency project viewed as test of ability to quickly negotiate agreements. An overarching goal of the project was to demonstrate that data sharing could be done quickly and to use that success to encourage other agencies to share data. The interagency project’s team was optimistic because (1) two different research teams at the U.S. Military Academy at West Point had secured data from VA on Army veterans who had used the GI Bill,[29] and (2) the nonprofit Student Veterans of America had earlier received de-identified, aggregate data on veterans’ use of the GI Bill.[30] All three efforts indicated VA’s willingness to share data that could be linked to Clearinghouse postsecondary outcomes. Moreover, the interagency project’s team sought early buy-in from federal agencies and received letters from several agency officials or verbal commitments to share data for the project.[31]

In addition, the Census Bureau already had data-sharing relationships with most, if not all, federal agencies for the preparation of the decennial census. The interagency project’s team also championed the use of interagency Joint Statistical Project agreements that were quicker and easier to negotiate. The interagency team’s approach was to minimize the number of agreements and to propose incorporating the interagency project’s data needs with ongoing Census Bureau data requests—that is, a single data delivery for multiple projects—which was successful at some but not all agencies. A prominent academic researcher expressed shock and surprise at how far the interagency project’s team had come as of July 2018 because several Joint Statistical Project agreements with VA components had already been signed. However, one of the interagency project’s team members commented that the strategy of demonstrating that data sharing could happen quickly may have actually prolonged negotiations with IRS.

Interagency project used an existing data-sharing structure with robust privacy protections. The interagency project was conceptualized for success because it used the Census Bureau’s privacy-preserving safeguards. The interagency project was housed at the Center for Administrative Records Research and Applications, an evidence-building component at the Census Bureau. In 2016-2017, this component sponsored more than 10 pilots to demonstrate the benefits of data sharing, and the interagency GI Bill data-linkage project was one of the pilot projects approved by and housed at the Census Bureau. The decision to locate the project staff at the Census Bureau was made to underscore a commitment to protecting the PII needed to successfully link federal databases.

The Census Bureau has significant infrastructure in place to help ensure the confidentiality and protection of federal data that includes PII. Researchers undergo background checks and mandatory training before receiving “special sworn status,” which essentially treats them as temporary Census Bureau employees with legal obligations to safeguard sensitive information.[32] All work is performed exclusively within a Census Bureau-designated Federal Statistical Research Data Center on secure servers. The Census Bureau anonymizes PII data prior to any analysis using its Person Identification Validation System to assign a unique identifier to each individual’s record.[33] No PII data is ever provided to researchers for analysis.^[34] Finally, the Census Bureau incorporates a “disclosure avoidance” process to maintain the confidentiality of any research using PII: All analytical reports undergo a review by the Census Bureau’s Disclosure Review Board before being released to the public to ensure that the data cannot be used to reveal individual identities.[35]

Key phases of the interagency GI Bill data-linkage project. The interagency project had three distinct but overlapping phases from 2016 through 2024, when its first three reports were released:

2016 and 2017—scoping the interagency project and seeking agency cooperation;
2016 to 2022—negotiating data-sharing agreements with federal agencies; and
2021 to 2025—data cleaning, analysis, drafting, review and clearance, and reporting.[36]

Appendix D provides a more detailed chronology of selected interagency project milestones from 2016 through 2025, when two additional reports were released.

V. CHALLENGES AND OBSTACLES THE INTERAGENCY PROJECT ENCOUNTERED

The interagency GI Bill data-linkage project encountered challenges from the negotiation through the reporting phases of the project. This section begins with an overview of several interrelated factors that prolonged data-sharing negotiations and then focuses on factors that contributed to delays during the data-matching and reporting phases. We then move on to a more granular examination of the specific challenges at each of the agencies from which the interagency project sought data. Finally, we report on several, more recent data-sharing initiatives, which were likely inspired by the activities of the interagency project.

A. Three Factors Prolonged Data Sharing Negotiations.

Negotiations with several agencies began in the fourth quarter of 2016 and ended in February 2022 when the last data-sharing agreement was signed—an addendum to a 2018 agreement. The principal contributors to the more than 5 years of negotiations and delays were a combination of (1) bureaucratic agency culture, (2) statutory privacy concerns, and (3) unpredictable situational delays. According to one interviewee, getting the layers of bureaucracy on the same page—senior leadership, administrative staff, technical staff, and lawyers—was challenging. Moreover, the process of negotiating data-sharing agreements was ad hoc, i.e., there was no standardized process within or across agencies and there were no established forms for data sharing.^[37] See Appendix C for a compilation of comments from interviews conducted during the course of this research that highlight the challenges in dealing with agency bureaucracies. Interviewees’ comments were lightly edited for clarity.

1. Bureaucratic culture. The most common characterization by interviewees of the negotiations was that they were plagued by bureaucratic delays, particularly when negotiations moved from concept to specifics. One interviewee highlighted the culture of federal agencies, where the number and layers of government staff involved in the negotiations who could say, “no,” or do nothing created multiple veto points that stalled the interagency project.

Lawyers. While there was conceptual agreement that data sharing would be beneficial, there was often hesitancy from federal agency lawyers when discussions turned to the specifics of the data being requested. The inevitable involvement of lawyers in the negotiations was often cited by interviewees as a major challenge.[38] As explained by one interviewee, ambiguity in the statutes leads to hesitation, and this empowers what he referred to as privacy advocates, within agencies, including agency lawyers. During negotiations with one agency, the Census Bureau and agency lawyers disagreed about the permissible statutory purposes for data sharing. One interviewee characterized the perspective of agency lawyers as risk-averse. The lawyers, he said, believe that they stand only to lose from data sharing because any leaking of PII data could potentially jeopardize their careers. In commenting on a draft of this report, a different interviewee suggested that the most effective antidote to lawyers’ risk aversion would be for a very senior official to direct the lawyers to find a way to say, “yes,” thereby ensuring minimal risk and reducing lawyers’ potential fear of career jeopardy.
Staff turnover. Because the interagency project spanned so many years, staff turnover was not uncommon. Any new agency face at the negotiating table required the interagency project’s team to reestablish a personal rapport, re-explain the scope and purpose of the interagency project, and respond to any concerns.^[39] The interagency project’s long timeframe also resulted in the expiration of several agreements, requiring the negotiation of extensions.
Point of contact. The federal agencies with which the interagency project’s team negotiated are complex organizations with multiple components. One interviewee noted that the lack of responsiveness encountered at one agency may have been due to the difficulty in identifying which office at the agency to approach with the data-sharing request. At VA, the lack of a central point of contact was a particular challenge because multiple VA components owned the data and had to sign off on data sharing. Some components were unresponsive and the interagency project’s team had to consider finding an alternative data source or abandon the use of certain variables.
Exacerbating staff turnover, lower-level agency staff did not assign as high a priority to data sharing as their senior leadership, as evidenced in the difficulty in scheduling meetings or eliciting insights as to what “concerns” might be contributing to delays. At one critical agency, a lower-level staff member simply lost track of the data-sharing request, thereby causing delays. At another, agency staff failed to keep their leadership in the loop. Whenever the interagency project sought assistance from senior agency officials, those officials helped to break stalemates and get the negotiations back on track; but, in later follow-up and execution, the interagency project’s team still had to deal with the same lower-level staff.
Additional duty. Data sharing is often an additional duty for agency staff involved in the negotiations. As a result, they are more likely to be focused on their day-to-day responsibilities rather than the longer-term benefits to their agency’s mission of data sharing. Furthermore, agency staff often have no incentive or reward for sharing data, which can be costly in terms of staff time and the manpower needed to prepare the data.
Access to technical experts. One interviewee explained that negotiations typically involved agency administrative staff and lawyers but not the technical staff who are the data experts. The interagency project’s team seldom had contact with such technical data experts, which led to the wrong data being sent by one agency. At a different agency, an interviewee stated that the technical staff didn’t really understand what the interagency project was requesting, have a way to share and extract the data, or have the resources to do so. At this agency, technical barriers were more serious than legal issues.

2. Statutory privacy concerns.^[40] Statutory privacy concerns raised by agency lawyers were a major contributor to the delays in the interagency project’s data-sharing negotiations. Although legal issues arose during negotiations with the Clearinghouse and IRS, negotiations involving ED lawyers were the most challenging and ultimately resulted in no data being obtained about veterans’ receipt of federal student aid.

In general, this report describes any legal issues that arose on an agency-by-agency basis in section D of this chapter. Here, however, we provide a brief overview of some of the ambiguities and inconsistencies in the statutory authorities that govern the use of privacy-protected data. These ambiguities amplified the role of risk-averse agency lawyers in data-sharing negotiations, which many individuals interviewed for this report said contributed to long delays. This analysis also draws on the findings of the Commission on Evidence-Based Policymaking, which are discussed in more detail in Chapter VII and Appendix F and G of this report.^[41] Finally, this section briefly discusses some of the legal challenges that ED itself encounters when it either successfully or unsuccessfully requests data from other federal agencies.

Privacy statute restrictions. The privacy restrictions in the Privacy Act and the Family Educational Rights and Privacy Act are the most challenging for data-sharing because they require agencies to obtain an individual’s written consent to disclose records containing PII.^[42] However, both statutes contain exemptions and exceptions to disclosure.

Privacy Act. The Privacy Act exemptions and exceptions are enumerated in a report by the Congressional Research Service.^[43] The Act’s 10 exemptions identify records that are outside the scope of the Privacy Act protections, including data collected for civil legal actions and material required by statute to be maintained and used solely as statistical records. The Congressional Research Service notes that there is some ambiguity as to what constitutes a statistical record because some administrative records containing PII may also be used for statistical purposes. The Privacy Act also lists 12 exceptions that allow the use of PII without consent, including for the Census Bureau to carry out its mandated activities. Several of the exceptions are broad and thus open to interpretation, including (1) “routine use,” which the Justice Department characterizes as the most controversial exception because of its potential breadth; and (2) “for statistical research” that is not used in making individual determinations, an exception which has been litigated and also appears to be open to interpretation.[44]
Family Educational Rights and Privacy Act. Under the Family Educational Rights and Privacy Act referred to (FERPA), which governs student records at ED, the terms “exemption” and “exception” refer to different ways in which the general rule of requiring consent for the disclosure of student education records might not apply. While sometimes used interchangeably in common language, their technical distinctions in FERPA are important. In essence, a FERPA exemption[45]means the records are not covered by the Act in the first place because they are not considered education records, while a FERPA exception[46] means the record is covered by FERPA, but disclosure without consent is permitted under specific conditions.

Thus, under FERPA’s exceptions, data collection contractors for surveys fielded by ED can receive student PII for survey participants. Similarly, the Clearinghouse also has access to PII-level data provided by institutions of higher learning in order to help them comply with federal student aid enrollment reporting requirements and to conduct research to help in the administration of federal student aid. The Clearinghouse, however, does not receive data from institutions on the federal student aid that many students receive.

Nevertheless, ED’s lawyers insisted that ED was unable to share data with the interagency team despite the evaluation (research) exemption in FERPA and the statistical research exemption in the Privacy Act.

Agency-specific data-sharing authorities. Agency authorizing statutes may identify permissible purposes for data sharing that includes PII such as to (1) measure a program’s impact (statistical research for evidence building) or (2) identify administrative improvements to a program’s operation. For example, VA has the authority to share data for statistical/research purposes to measure impact (evidence building), which allowed several VA components to sign agreements establishing Joint Statistical Projects with the Census Bureau. As discussed later in this chapter, interviewees commented that the Census Bureau and IRS lawyers disagreed over the purposes for which data could be shared, with the former arguing that data sharing was permissible for evidence-building and the latter believing that data could be shared only to improve agency operations.

ED has shared PII with several agencies, according to a 2017 paper,^[47] to help ensure the implementation of statutory requirements such as the collection of student loan debt (using data from the Department of Health and Human Services New Hires Directory) and the enforcement of the Gainful Employment regulation (using data from the Social Security Administration). Gaining approval for the latter required White House intervention. While there is a statutory exception that allows ED to gain access to another agency’s administrative data for the sole purpose of debt collection, its research component (NCES) encountered a statutory barrier with respect to Social Security Administration data for its trio of surveys that collect information on students who participate in federal student aid. The Center had sought an agreement with the Social Security Administration that would generate plausible values for wages using students’ PII when such data were not available from the student interviews. However, because express legislative authority to do so had not been found, the effort as of 2017 was unsuccessful.

Although the lack of express statutory authority may be an impediment to data sharing, congressional requirements to share data do not always guarantee that obtaining a data-sharing agreement will be quick and easy. One interviewee said that a congressionally mandated data-sharing agreement with DOD took 4 years to negotiate. This individual commented that two sets of DOD lawyers, its Privacy Office, and its Cybersecurity office offered seemingly “endless” objections without identifying solutions. Moreover, as discussed in Section D of this Chapter, an ED-Census Bureau data-sharing agreement that was also mandated by Congress took more than 10 years to negotiate.

3. Unpredictable “situational” delays. Several unpredictable events contributed to delays. First, the arrival of COVID in March 2020 had a ripple effect on the interagency project. For several months, COVID shut down Federal Statistical Research Data Centers used by the project’s research team at the Census Bureau and other locations, and reliable remote access was initially problematic. The pandemic made in-person meetings impossible and resulted in slippages in project milestones. In addition, key personnel on the interagency team and at agencies became sick, which further delayed the project. COVID also had a tertiary effect because the interagency project decided to use only 2018 and 2019 earnings data from IRS because the 2020 earnings might have been skewed by the pandemic and not be representative of veterans’ experiences. At the same time, the high priority the Census Bureau attached to meeting the decennial census deadline, despite COVID, led to unavoidable delays in the interagency project, as did a less significant but congressionally mandated ED-Census Bureau data-sharing agreement.

B. Data Quality-Related Delays Occurred During the Matching Phases, Holding Up Data Analysis

The data files provided by agencies had to be “cleaned” before they could be linked to the other data sets being used to analyze veteran outcomes. Data cleaning was particularly important because the project’s analysts would be matching the data across multiple datasets in order to explore the relationship between students’ outcomes and various demographic and military service characteristics. By “fixing or removing incorrect, corrupted, incorrectly formatted, duplicate, or incomplete data within a dataset,” cleaning ensures that the data is accurate and consistent enough for analysis.^[48]

According to members of the interagency team, the quality of the data provided by the Veterans Benefits Administration (Benefits Administration) caused major delays and had to be cleaned twice, first by the Clearinghouse and then by the interagency project’s team. Data provided to the Clearinghouse by VA’s Data Governance and Analytics Group also required data cleaning.[49] The two VA components began sending records on veterans’ use of benefits and other data to the Clearinghouse in April 2020. In addition to providing duplicate records, the various components within VA that owned the data provided by the Analytics Group used different PII identifiers for the same veteran. Moreover, some of the data provided was old and had to be resent. The time devoted to data cleaning is evidenced by the approximately 10 months between the arrival of the VA data at the Clearinghouse (April 2020) and the Clearinghouse’s completion of matching and then returning the matched outcome data to VA (late January 2021).[50]

Data problems continued with the matched VA and Clearinghouse data sent to the Census Bureau. Interviewees also noted that the data from VA required a significant amount of “cleaning” once it arrived at the Census Bureau, a task they characterized as substantially more time-consuming than the actual data analysis. For example, interviewees reported that the Veterans Benefits Administration had failed to document the columns and provide a data dictionary. Moreover, the data included duplicate records; mixed up data between school and veteran; omitted files; omitted values for requested files; and file format issues. Numerous calls and months passed as the interagency team sought to receive clean data from the Veterans Benefits Administration.

Several interviewees said that the messiness of the VA data and the time spent data cleaning were not unexpected because the VA data were originally created for internal administrative purposes, not statistical analysis. One interviewee explained that departmental datasets typically consist of operational records used to deliver benefits, which are “moving at the speed of light.” Operational data is not curated for analytics, and the priority of the program managers is not to go back to fix errors in the records but rather to keep the benefits flowing, explained one interviewee. The program managers work with one record at a time, but the entire database is massive, magnifying the impact of errors in individual records. Another factor contributing to data-quality delays, according to several interviewees, was the challenge the researchers faced in speaking directly to the individuals who managed the data being requested. Such meetings were difficult to schedule and never occurred at one agency despite repeated requests.

According to the interagency team, data from other agencies also had missing files and required clarification of columns and a data dictionary. Thus, cleaning of the IRS data that was received in July 2021 was not completed until January 2022.

C. Report Publication Delayed by Challenges in Obtaining Comments and Clearances

In October 2022, about 7 months after the receipt of the last dataset, a draft of the first interagency project report was sent for review and approval to DOD, VA, and Census Bureau staff not associated with the interagency project.^[51] The anticipated publication date for the report was November or December 2022. However, significant delays occurred because the agencies did not provide comments in a timely manner, and the first of the three reports was not published until February 2024.

According to interviewees, agency review and clearance was slowed down by what was characterized as the low prioritization given to the request by agency staff. When one agency was asked to review a draft report, its employee responded by providing an online tutorial on how to write a report, suggesting the report (prepared by PhD researchers with extensive report-writing experience) had not been opened or reviewed. Possibly, comment and clearance delays can be attributed to the additional duties of individuals providing comments. An interviewee at one agency explained that he was given assignments that were a higher priority than the interagency project, which therefore was relegated to a back seat. This individual was told by his supervisor that he nevertheless “would be able to handle it all.” As a result, several other interagency project reports sent for review and clearance in June 2024 were delayed for about 6 months and were not published until February 2025.

D. Some Data-Sharing Challenges Were Agency-Specific

The interagency project negotiated data-sharing agreements with four agencies plus the National Student Clearinghouse. The negotiations resulted in 11 agreements. The first data-sharing agreement was signed in April 2018 and the last one in February 2022. This section describes the negotiations with each agency, beginning with the privacy protections enumerated in the Census Bureau’s agreement with the interagency project and ending with the unsuccessful negotiations with ED.

According to interviewees, most agencies—including VA, DOD’s Defense Manpower Data Center, and IRS—were open to data sharing even if the negotiations were prolonged and encountered bureaucratic delays. However, negotiations with ED never resulted in a data-sharing agreement because of ED lawyers’ statutory data privacy concerns. While the National Student Clearinghouse was an early interagency project advocate and ally that agreed to participate even before the project had been funded, even the Clearinghouse was nevertheless a source of delays; its lawyers required three agreements, each of which necessitated months of negotiation and review. The critical agreement that initiated the flow of PII data on enlisted beneficiaries from the Veterans Benefits Administration to the Clearinghouse for matching to the Clearinghouse’s student records was not signed until January 2020, more than a year after data-sharing agreements were reached between the Census Bureau interagency team and VA components. One interviewee cited the Clearinghouse agreements as surprisingly challenging to negotiate because of all the legal issues raised by its lawyers.

Census Bureau agreement with embedded researchers (one agreement)

This 2018 agreement articulated all of the privacy protections inherent in positioning the interagency GI Bill data-linkage project at the Census Bureau, including (1) bringing in external analysts as special sworn status temporary Census Bureau employees; (2) using its secure research data centers with secure servers for all analytical work; (3) anonymizing all PII before it reached researchers; and (4) ensuring a final, independent review to ensure that any publicly reported data could not be used to reveal individual identities.

Situating the interagency project at the Census Bureau was a strategic decision because the Census Bureau regularly obtains data from other federal agencies for purposes of the decennial census. As a result, the data-sharing pathways and infrastructure already existed, which theoretically would support the success of the interagency project. The Census Bureau also employed several high-level staff who were personally invested in the importance of interagency data sharing.

In addition, the project and its placement at the Census Bureau was conceptualized amidst a congressional effort to promote data sharing for policymaking. In 2016 legislation, Congress established the Commission on Evidence-Based Policymaking (Commission). The expectation was that the Commission’s report would recommend the creation of a National Secure Data Service within the Census Bureau, which already had the infrastructure and personnel.

Anticipating that the Census Bureau would house the proposed new Data Service, the Bureau established an evidence-building initiative in 2016.[52] The initiative was placed in the Bureau’s Center for Administrative Records Research and Applications with more than 10 pilot programs to demonstrate the benefits of data sharing and linkages.^[53] The interagency GI Bill data-linkage project was one of those pilots. According to one interviewee, the pilots were intended to demonstrate the Census Bureau’s ability to become the new Data Service. The pilots all had the same rigorous procedures to safeguard the integrity of any PII requested.

In 2017, the evidence-building initiative at the Census Bureau fell victim to budget cuts, streamlining, and reorganization, and, in the process, fell off the radar of senior Census Bureau leadership. According to one interviewee: “In a period of 6 months, the Center’s pilots, including the interagency GI Bill data-linkage project, went from high profile to a fly on the wall” and team members felt that no one at the Census Bureau was willing to expend capital and approach agencies such as ED, DOD, VA, and IRS to advocate for data sharing on the interagency project’s behalf. One example of the lower priority assigned to the interagency project was noted earlier: the review and release of final project reports experienced long delays because of competing Census Bureau priorities.

U.S. Department of Veterans Affairs (four agreements)

The interagency project’s negotiations with VA for data on veterans confronted bureaucratic delays and the lack of a centralized process for obtaining permission to share administrative datasets owned by disparate VA components. Ultimately, agreements authorizing Joint Statistical Projects were signed in the summer of 2018 with two VA components—the Veterans Benefits Administration (Benefits Administration) and the Data Governance and Analytics group (Data Analytics) in the Office of Enterprise Integration. Additionally, two ancillary memoranda of understanding were initiated in 2018 or 2019.

In September 2016, the Deputy Under Secretary for Economic Opportunity in the Veterans Benefits Administration pledged to work with the interagency project, identifying the Director of the Benefits Administration’s Education Service as the primary contact point. The Education Service administers the GI Bill benefits programs and maintains a database of veterans who have used any GI Bill benefit, including the Post-9/11 GI Bill. However, other VA components controlled the release of additional data needed to report on veteran outcomes, such as demographic data, disability ratings, and the identity of eligible veterans who did not use their GI Bill benefits. Negotiating the release of these data were challenging, in part because of the number of VA components involved. VA’s Data Analytics group is the centralized repository for much of the department’s data but does not own any data and had to negotiate for approval to release data from various VA components.^[54] The Data Analytics group did have agreements with the Census Bureau to provide some of the same data on an ongoing basis, but that did not include the authority to release it to the interagency project’s team.^[55]

Although VA’s Deputy Under Secretary for Economic Opportunity had expressed early support for the project, one interviewee commented that the Education Service was “missing in action,” that is, not as involved in the interagency project as it should have been. Members of the interagency project singled out the Education Service as a contributor to significant delays throughout the project and messy records when the data were delivered. Within the Education Service, staff turnover and the failure of lower-level staff to keep their superiors informed about the status of data-sharing negotiations led to delays, according to interviewees. One Education Service staff member assigned to shepherd the data-sharing agreement was reportedly removed for failing to provide updates on the project to his superiors and, according to several interviewees, his replacement’s questions were characterized as “back to square one,” the equivalent of starting the negotiations all over again. To break these logjams, the interagency project had to reach out to senior-level officials, including the Under Secretary for Benefits, who oversaw the component managing the GI Bill and who intervened on behalf of the interagency project.

In contrast to the Veterans Benefits Administration, VA’s National Center for Veterans Analysis and Statistics, reporting to its Data Governance and Analytics group, was an early proponent of the project with a representative on the interagency project’s team and serving as a co-author of reports, and frequently intervened to shepherd data from various VA components and even personally cleaned “messy” data from other VA offices.

One interviewee said that VA has made progress in developing a data-sharing infrastructure since the enactment of the 2018 Evidence Act, which will facilitate and “streamline” data sharing going forward. For example, a data linkage-type project would utilize a centralized intake process through a new data governance council, which is composed of VA representatives of the components that own administrative datasets. The co-chairs of the council are VA’s Chief Data Officer and the department’s Deputy Chief Information Officer.[56] However, several interviewees shared a different perspective on the governance council’s activities. According to an interviewee, one component was not aware of the new centralized intake process and was approaching agencies independently with data-sharing requests. A different interviewee indicated that the new centralized intake process was more cumbersome for the renewal of agreements because the VA components on the new governance council now must sign off on what should be a routine approval.

National Student Clearinghouse (three agreements)

A major question that was vigorously debated and that contributed to a fair amount of delay in the early years of the project was the flow of PII data, specifically whether the Clearinghouse should send matched records with PII data directly to the Census Bureau. Two factors contributed to this lengthy debate. One factor was the involvement of two Clearinghouse lawyers with different viewpoints as to what constituted “re-disclosure.” The Census Bureau’s lack of an existing relationship with the Clearinghouse was the second factor in the hesitancy of the Clearinghouse to send matched records containing PII to the Bureau. During the course of the negotiations, however, the Clearinghouse became more comfortable with the role of the Census Bureau, including its PII safeguards. Eventually, it was determined that the Clearinghouse would return the matched records to VA’s Education Service, which could provide them to the Census Bureau on VA’s terms, obviating the need for a separate Clearinghouse-Census Bureau memorandum of understanding.

Nonetheless, three data-sharing agreements were needed to proceed with the data matching at the Clearinghouse: two regarding the need to maintain confidentiality for VA data and one regarding the flow of PII data from VA’s Education Service to the Clearinghouse, back to VA’s Education Service, and from the Education Service to the Census Bureau, with explicit acknowledgement of other privacy safeguards built into the interagency project.^[57] First, consistent with Census Bureau procedures, the PII in the Clearinghouse-matched data would be assigned an anonymous identifier before it was disclosed to the interagency project’s researchers. Second, there would be no disclosure of institutionally identifiable data (the name of the institution attended by a veteran) without explicit Clearinghouse permission.^[58]^,[59] Finally, the Clearinghouse data would become part of a “Combined Dataset” that included de-identified data from DOD and other agencies.

DOD’s Defense Manpower Data Center (one agreement and one amendment)

The interagency project sought a data-sharing agreement with the Defense Manpower Data Center (Manpower Center) for veterans’ military rank, occupation, tenure, service in hostile war zones, other characteristics of their military service, and, importantly, scores from the Armed Forces Qualification Test. That test measures arithmetic reasoning, mathematical knowledge, paragraph comprehension, and word knowledge of incoming service members, and thus provided the interagency team with a snapshot of veterans’ academic preparedness at the time they enlisted. The interagency project’s researchers viewed these data as an important “control” to hold constant the comparisons of veterans across other variables; in other words, these data enabled the interagency team to compare similarly situated veterans during regression analyses. In December 2016, the interagency project was on the brink of a data-sharing agreement for these data with the Manpower Center. However, the Manpower Center requested some changes to the agreement, which in hindsight the interagency project’s team believed it should have accepted because further negotiations dragged out the project. Because of bureaucratic delays, a data-sharing agreement was not signed until April 2018. When negotiations with the Manpower Center were stalled, a senior-level DOD official intervened to help get the negotiations back on track. One interviewee commented that Manpower Center officials may not have understood why the data were needed or how they would benefit from data sharing. The data-sharing agreement successfully incorporated the strategy of a single data delivery for both the decennial census and the interagency project. The interagency project did not receive the Manpower Center’s data until October 2019 because of the priority attached to the decennial census, which also relied on Manpower Center data.

Issues with the Management Center’s 2019 dataset were not apparent until the interagency project’s analysts attempted to match this DOD component’s data with other datasets that started arriving in the spring and summer of 2021. The analytical team discovered that the data it had received from the Manpower Center in 2019 were not the data it thought it had requested: the data were incomplete because they contained data for only one year, and the academic preparedness scores were not from the service members’ date of enlistment. Moreover, an additional dataset from the 2019 Defense Enrollment Eligibility Reporting System had been requested but was not provided. In response to an emailed question about the wrong data being sent in 2019, the response from a Management Center employee was, “I am not sure what you mean by wrong data being sent.”

The interagency project’s team attributed the receipt of the wrong data to its inability to communicate directly with the individuals at the Manpower Center who were pulling the data. Meetings between the project’s analysts and the technical experts at the Manpower Center were requested but never took place. To receive the necessary data, the Manpower Center required an amendment to the 2018 agreement. The amendment was drafted in October 2021, approved in February 2022, and the corrected data were received in April 2022—a lengthy delay.

Internal Revenue Service (one agreement)

According to members of the interagency team, although IRS verbally agreed in 2017 to share tax data with the interagency project for the purpose of evidence-building, approval to share data ended up spanning several years of negotiations. According to interviewees, the delay was because of lawyers’ disagreement over the statutory data-sharing authorities and the appropriate data-sharing vehicle.[60] Approval to share data was finally reached in early 2021.

Title 13 § 8 authorizes the Census Bureau to engage in Joint Statistical Projects by acquiring data from federal agencies and other nonprofits as long as the Census Bureau protects the privacy of individuals identified in the datasets.[61] Use of this authority was the interagency project’s team preferred data sharing mechanism.^[62]^,[63] In contrast, IRS shares data with the Census Bureau under Title 26 §6103, which allows the Census Bureau to use tax data for statistical purposes. [64] Eventually, in February 2021, the interagency team abided by IRS’ preferences and submitted a Predominant Purpose Statement. It was promptly approved the following month. In retrospect, one interviewee said that the interagency project should have initiated work on a Predominant Purpose Statement much earlier (in 2017), which could have shaved several years off the timeline to obtain IRS data. The IRS data-sharing agreement was the last to be signed, but the W-2 and1040 income and other data arrived quickly (July 2021).

U.S. Department of Education (no agreements negotiated)

Negotiations between the interagency team and ED began in 2017, and the last meeting was held in early 2024. Initially, according to one interviewee, ED was nonresponsive to the interagency project’s overtures on data sharing, which may have been related to “not knowing who to talk to.” The lengthy ensuing negotiations with ED involved three sets of Departmental lawyers, including two privacy lawyers, one who had concerns that FERPA did not permit data sharing and another who had Privacy Act concerns. One interviewee noted that dealing with these two ED lawyers increased the challenges because, even if the concerns of one lawyer could be assuaged, the other lawyer could still say no. Another interviewee pointed out that the likely participation of ED’s Privacy Officer and Cybersecurity Office would have reinforced any privacy concerns raised by ED’s General Council’s lawyers. He characterized this group of ED lawyers as a “circular firing squad.”

However, somewhat competing narratives emerged from interviews conducted for this report—that the privacy concerns primarily emanated from ED’s Federal Student Aid office but that the Department itself could have been persuaded to share PII with the interagency project’s team. On the one hand, several interviewees painted ED with a broad brush and attributed the stalemate to “ED lawyers,” who did not believe data sharing is permitted by statute, an attitude that they believed had become part of the culture at ED. In effect, they said that staff at ED are “trained” to say “No” to data sharing. On the other hand, a different interviewee offered a less monolithic, more nuanced picture of the data-sharing landscape at ED, commenting that the Department’s Federal Student Aid office has its own more restrictive mindset about data sharing and has long wanted to maintain a big wall between itself and the rest of the Department. Consistent with that more nuanced perspective, several interviewees involved in the negotiations suggested that even though ED had long limited the use of data on federal student aid to identifying operational improvements for aid delivery, high-level support within the Department and the White House could have led to sharing PII with the interagency project’s team. One of these interviewees even questioned why VA had not reached out to press the case for the importance of obtaining this information, which this individual said would have caught the attention of ED officials.

Finally, at one point in the negotiations, ED proposed a data exchange where the Census Bureau would send student veterans’ PII to ED, and ED would match the PII and provide aggregated data on the cohort’s receipt of federal student aid, e.g., median student loan debt, the percentage in default on federal loans, median Pell Grants, etc. This proposal, however, was a nonstarter because the Census Bureau is prohibited by Title 13 from sharing any privacy protected information with another agency. Moreover, the interagency project needed individual-level financial aid records in order to match those records to individual-level records being provided by VA, DOD, IRS, and the Clearinghouse. The interagency project’s team did discuss having VA send PII to ED in return for aggregate data but decided it was not worth the effort. Ironically, as discussed in the next section of this report, VA is now pursuing such an agreement with ED.

There are examples of ED’s sharing PII data with other federal agencies, including the Census Bureau, suggesting that data sharing by ED is possible:[65]

A congressionally mandated ED-Census Bureau data-sharing agreement that had been in the works for more than 10 years was signed in August 2023, about 6 months before the release of the first interagency project’s report. These negotiations had taken precedence over the interagency project’s attempts to secure similar data from ED. According to ED officials, the data-sharing agreement that was approved with the Census Bureau protected privacy by anonymizing PII, the same safeguard the interagency project’s team had offered. We were unable to obtain a copy of the agreement, but we did confirm that (1) it involved ED’s sending November 2021 data containing PII to the Census Bureau in 2023, (2) ED subsequently stopped providing such data, and (3) it was unclear why ED stopped sharing data under the agreement. The interagency project’s team was unaware that this agreement had finally been signed.
After White House intervention, ED was able to obtain earnings data from the IRS for the launch of ED’s “College Scorecard” search tool in 2016.^[66]

E. Data-Sharing “Spring Roots” Inspired by the Interagency GI Bill Data-Linkage Project

Several initiatives appear to have been inspired by the interagency project. Many of the initiatives involved data sharing to inform policymaking and several led to the building of infrastructure, which should benefit future data-sharing efforts.^[67]

Census-Clearinghouse data sharing initiated. Over the course of the interagency project, the Clearinghouse became more comfortable engaging with the Census Bureau in other data-sharing initiatives involving PII. For example, one such data-sharing exchange involving PII examined the role credentials play in the earnings and career prospects of students and workers. According to one interviewee, this collaboration would have taken longer had a comfort level not been established by the interagency GI Bill project. An additional bonus is that the Clearinghouse and the Census Bureau now have templates to use in negotiating additional agreements.
OMB infrastructure building. Shortly after the release of the interagency project’s first report, OMB requested copies of the project’s agreements with federal agencies (as described in Section D above) to use as models for future data-sharing projects. This request was a testament to the importance of the interagency project and its contributions to the building of infrastructure to support future evidence-building initiatives.
VA initiates new data-sharing agreements. In early 2024, the Veterans Benefits Administration reached out to several federal agencies and the Clearinghouse to initiate negotiations for aggregate outcome data that could be shared with veterans, potentially on the GI Bill Comparison Tool.^[68] According to one interviewee, the interagency project opened VA’s eyes to the benefits of including outcome data on the Comparison Tool website.[69] As of January 2025, no agreements had been finalized, but those with ED and the Clearinghouse were reportedly the closest to signature. ED negotiations, however, reportedly have been on hold since the summer of 2024 pending a decision from ED lawyers on whether or not to proceed.It is unclear if an agreement could overcome a new ED privacy issue that has emerged concerning the release of aggregate data on veterans’ earnings, an offer ED had previously made to the interagency project’s team.^[70] According to one interviewee, ED already spends considerable resources on obtaining such data for its College Scorecard on students who receive federal student aid, and there is concern about the public reporting of two similar but not identical estimates of average student loan debt, that is, one for all students and one for veterans. While another interviewee acknowledged that releasing two different estimates on average student loan debt is always a concern in the research community, he believed that concern could be overcome by the use of privacy-masking tools.
VA use of outcome data. Two VA components appear to be considering or actively engaged in efforts to use and build on the Post-9/11 outcome data included in the interagency GI Bill data-linkage reports. One interviewee said that his office plans to come up with new projects based on the insights from the interagency project’s reports. An interviewee from a different VA component said that his office was using the reports’ findings to identify gaps in services provided to GI Bill beneficiaries and actively considering contacting beneficiaries who have not used their benefits.
New congressional requirement for data sharing. Section 215 of the Senator Elizabeth Dole 21st Century Veterans Healthcare and Benefits Improvement Act of 2024 required VA to negotiate data-sharing agreements with ED, Treasury, and other agencies as appropriate in order to post outcome data on the GI Bill Comparison Tool.[71]

VI. STRATEGIES THE INTERAGENCY PROJECT USED TO ENCOURGE FEDERAL AGENCIES TO SHARE DATA

The interagency project’s team employed several strategies to advance negotiations when faced with bureaucratic culture. The team’s strategies can be characterized as a combination of persistence coupled with a “carrot is better than a stick” approach.

The “lessons learned” strategies identified included:

Play nice by behaving positively, rather than being aggressive when dealing with seemingly intransigent agency staff, because they control the data.
Be creative and willing to take “yes” for an answer even when it is a lesser “yes.”
Offer to collaborate with an agency on other projects by taking into consideration the agency’s data priorities.
Tailor research questions to meet agencies’ interests because each agency is devoted to its mission. Data sharing needs to be viewed as more than just research endeavors but also as products that meet each agency’s needs.
Be patient but persistent and ask for help in breaking stalemates in the negotiations.
Engage very high-level agency officials to ensure higher prioritization by lower-level staff.

Examples of these strategies included:

The interagency project’s team learned a lesson when negotiations with DOD broke down in December 2016 over changes to the draft data-sharing agreement that DOD had requested but which were not dealbreakers and which the interagency team wishes they had embraced immediately to head off delays.
The interagency project’s team appended its data-linkage requests onto larger, ongoing interagency projects to speed data delivery.
At the outset, the team asked the VA Data Governance and Analytics group what its data priorities were and then offered to run reports based on several topics it identified– specifically, rurality, gender, and disability. This collaborative approach helped secure early cooperation with the Data Analytics group, which was a consistent ally in delivering data (the Data Analytics group is a VA component separate from the Veterans Benefits Administration).
The interagency project’s team asked senior officials to intervene when negotiations languished or hit roadblocks because they were not a priority for lower-level staff, an approach that worked at both VA and DOD. Members of the research team were able to engage an agency deputy secretary, a deputy undersecretary, assistant secretaries, and agency chiefs of staff. Each time, the interagency project’s team received robust and immediate support, sometimes with the high-level staff expressing shock that the lower-level staff had allowed such delays in the interagency project. One interviewee noted that “interagency project delays were caused by staff turnover (both supervisors and analysts) and the lack of ownership by those who replaced them.”

Although the interagency project’s team obtained help from senior-level agency officials, they did not ask for such help from the White House or OMB. There were conflicting views within the team about the usefulness of seeking such high-level interventions among interviewees. On the one hand, some interviewees believed that high-level support for data sharing at the White House and OMB could have sped up the data-sharing negotiations. They cited the success of ED’s project to create the College Scorecard higher education search tool, which was launched in 2015. However, several interviewees attributed this success directly to the President’s personal involvement in forcefully nudging ED to obtain IRS earnings data.

On the other hand, many interviewees did not see obtaining White House-level support as a sustainable long-term approach to existing bureaucratic hurdles to evidence-building projects. Aside from Presidential support, one interviewee commented that the White House Domestic Policy Council is not structured to bring together agencies for data-sharing because it is too stove-piped. Several interviewees doubted that the interagency project’s timeline would have been shortened by going to the Domestic Policy Council or OMB because it is not sufficient for such officials to indicate that data sharing is a priority—such support needs to be sustained, and officials need to follow up with directives and enforce them. While such senior-level officials can help to break stalemates to get negotiations back on track, data-sharing projects still must be implemented and operationalized by lower-level staff. There was a consensus among those interviewed, however, about the benefit of intervention by senior-level agency officials to ensure that data-sharing negotiations were not stuck with lower-level staff.

VII. INTERAGENCY DATA SHARING ALSO ENCOURAGED BY FEDERAL INITIATIVES

Concurrently with the initiation of the interagency GI Bill data-linkage project, Congress was considering initiatives to improve interagency data sharing. Two individuals involved in the interagency project supported those initiatives in testimony before a Congressional Bipartisan Commission on Evidence-Based Policymaking in October 2016.[72] The Bipartisan Commission’s efforts resulted in enactment of the Foundations of Evidence-Based Policy Act of 2018 (Evidence Act), the implementation of which overlapped with the interagency project. The Evidence Act is sprawling in scope: a complex network of new officials, new responsibilities, new councils, new processes, and a demonstration project.^[73] Although realization of the Act’s long-term goals remains a work in progress, its implementation status sheds light on the challenges to interagency data-sharing encountered by the interagency GI Bill data-linkage project and on the potential impact on similarly ambitious data-linkage projects.

A comprehensive evaluation of the status of all the Evidence Act initiatives was beyond the scope of this report, but insights were available from three relatively recent reports as well as from interviews with individuals involved in the interagency GI Bill data-linkage project, data sharing experts, and agency staff involved in implementing the Act’s initiatives. Each report highlighted the challenges and progress made during the Act’s implementation: (1) a 2022 study conducted by the Data Sharing Working Group (Working Group) on the obstacles to evidence-based data sharing; (2) annual surveys of the Chief Data Officers (CDOs) conducted by the Data Foundation,[74] the most recent of which covers 2024; and (3) the 2024 annual report to Congress on the implementation of the National Secure Data Service (NSDS) demonstration. We also reviewed reports by GAO that were mandated by the Evidence Act. Finally, many interviewees shared their impressions of the progress made through early 2025.

A. Background on the Evidence Act Initiatives

The Evidence Act was the culmination of years of conversations between Representative Paul Ryan and Senator Patty Murray about the need to develop a strategy to more effectively use government data.

2017 Bipartisan Commission report identifies obstacles to data sharing. Draft legislation creating the Bipartisan Commission on Evidence-Based Policymaking (Commission) to develop such a strategy was introduced in 2014 and signed into law in March 2016.^[75] A little over a year later, the Bipartisan Commission released a 128-page report containing a detailed account of the bureaucratic and legal challenges in accessing and linking federal data for policymaking.[76]

Bureaucratic challenges. Throughout its report, the Bipartisan Commission identified significant obstacles to data sharing between federal agencies and external researchers, including (1) a complex array of processes, protocols, and approaches; (2) cumbersome and onerous procedures that were characterized as idiosyncratic and that differed across agencies; (3) the compounding of challenges when researchers seek to access multiple agencies’ data, and (4) resource challenges and researcher burdens that can impede evidence-based studies. Appendix F of this report contains extracts from the Bipartisan Commission’s report on these bureaucratic challenges.

Ambiguous and inconsistent legal authorities. The Bipartisan Commission’s report describes the “complex web of statutes, regulations, and implementing guidance (or absence thereof) [that] drives risk aversion at agencies, causes frustrations for the evidence-building community, and limits the value of data for statistical purposes.” “In effect,” the Bipartisan Commission concluded, “the existing legal environment limits the government’s ability to steward data responsibly as a valuable resource for the American people and for policymakers.” To mitigate the risk attributable to inconsistent legal authorities, the Bipartisan Commission called on Congress to amend statutes as appropriate to allow the statistical use of data for evidence building.^[77] Appendix G of this report contains extracts from the Bipartisan Commission’s report on ambiguous and inconsistent legal authorities.

Bipartisan Commission’s conclusions and recommendations. The Bipartisan Commission’s report also articulated a vision for using federal administrative data to help inform policymaking, rejecting the assumption that its routine use would threaten privacy: “The Commission believes there are steps that can be taken to improve data security and privacy protections beyond what exists today, while increasing the production of evidence.” The report made 22 recommendations to improve both the privacy protections afforded to the American public and the availability of rigorous evidence to inform policymaking by improving data access, modernizing privacy and confidentiality protections, strengthening evidence-building capacity, and establishing a Data Service to support governmentwide evidence-building. OMB was envisioned as playing a coordinating role across federal agencies, and Congress was expected to provide sufficient resources to support evidence-building activities.

Evidence Act of 2018 initiatives. The Bipartisan Commission’s blueprint resulted in the Evidence Act of 2018, which was signed into law in January 2019.[78] Many of the Bipartisan Commission’s recommendations were incorporated, including the establishment of new advocates for data sharing within each agency—Chief Data and Evaluation Officers, Statistical Officials, and coordinating councils—as well as a standard application process for data-sharing projects. The Evidence Act also instructed OMB to establish an Advisory Committee on Data and Evidence Building (Advisory Committee) tasked with developing a vision and framework for the Data Service.^[79] The Evidence Act did not tackle the issue of inconsistent legal authorities. In other words, it created new evidence-building advocates and mechanisms without addressing the statutory barriers to data sharing. Nor did Congress provide additional funding to implement this complex set of activities. Appendix H of this report summarizes the key provisions of the Evidence Act of 2018.

B. Data Sharing Working Group’s 2022 Report Describes a Familiar Set of Federal Data Sharing Challenges

A 2022 report by the Data Sharing Working Group[80] identified challenges to interagency data-sharing that were similar to those encountered by the interagency GI Bill data-linkage project, including a slow and decentralized process, characterized by stove-piping within agencies; a lack of mechanisms to end stalemates; agency risk aversion concerning PII; a lack of data-quality standards; and the absence of a data inventory that could be shared. Several challenges encountered by the interagency project, however, were not discussed in the Working Group’s report, including staff turnover and competing organizational priorities. Appendix I summarizes the specific data-sharing challenges identified by the Working Group and compares them to analogous challenges encountered by the interagency project.

Background. The Chief Data Officers’ (CDO) Council established the Working Group in October 2020.[81] The goal was to help the council better understand the ever-evolving nature of data-sharing needs as well as the data-sharing challenges facing all federal agencies. After examining these issues, the Working Group developed recommendations for improving access and data sharing within and between agencies, including initiatives to:

Expedite data agreements.Problem: Federal data sharing requires agreements between the owner and/or data custodian and the party requesting access. The Working Group found that agreements “often take months to complete, are sometimes never completed, and often the need for data access is long gone by the time the agreement is in place.” It concluded that “making the process of establishing a data-sharing agreement more efficient and timelier is critical for improving data-sharing across the federal government.”Proposed solutions: (a) Build a collection of agency templates and standard clauses that can be used to draft agreements. (b) Develop an agreement-building tool to draft agreements. (c) Implement agency-level repositories to maintain copies of existing agency agreements. (d) Implement a Chief Data Officer-developed process describing the standard steps and associated timeframes. As discussed in Section D of this Chapter, several of these suggestions are in the process of being addressed by the NSDS demonstration.
Improve data awareness.Problem: “There is insufficient visibility,” the Working Group found, “of the data that are available from each government agency and which of those data can be shared across the federal government.”Proposed solutions: (a) Reinforce the Data.gov website[82] as the government-wide metadata inventory that can be used to discover what datasets are available across the federal government. (b) Create and adopt a data classification mechanism that will help with the identification of data that can be shared and in addition can mark data with existing security classification schemes. (c) Draft a data-sharing infrastructure playbook to include lessons learned, success stories, and pitfalls in order to accelerate the data-sharing capabilities and culture for agencies that are at the early stages.^[83] (d) Create a recognition mechanism to incentivize sharing. This recognition does not have to be associated with material benefits, such as funding, but should at the very least award the organizations that are willing to and do share data.
Improve data trustworthiness.Problem: The Working Group concluded that “There is no standard application of methods for collecting data or evaluating data quality.”Proposed solutions: (a) Perform periodic data quality control reviews that use standardized data quality measures and assessment techniques that are easily understood and can be adopted across federal government agencies and conduct periodic data assessments to validate syntax and factual correctness. (b) Perform pre- and post-data quality reviews to ensure data standards and practices are followed. Agencies can promote data quality across the federal government by ensuring rigorous data collection practices before sharing the data with other government agencies.

In addition to these three high-level categories of recommendations, the Working Group identified two general proposals for agency CDOs that it believed would benefit data sharing: CDOs should (1) serve as the central source of information on data sharing and (2) establish data-sharing centers of excellence, providing standardized data quality measures and assessment techniques.

C. Overarching Themes Emerged from Surveys of Chief Data Officers

The Data Foundation has surveyed federal agency CDOs five times, starting in 2020.^[84] Several overarching themes emerged from our analysis of these surveys and shed further light on the difficulties facing interagency data sharing, which were similar to those encountered by the interagency GI Bill data-linkage project (the numbers in parentheses below indicate the dates of the annual reports identifying these themes):[85]

Lack of dedicated funding. CDO staffing remained an issue at some agencies, and a key priority for CDOs was funding to support their mission (2022). CDOs would benefit from more funding to support their mission (2022). Consistently, CDOs reported financial and budgetary issues as key challenges, and each year they cited a need for staff and technology (2024). Organizational leadership could help alleviate capacity challenges for CDOs by allocating budgets and staff support (2024).
Confusion over roles and responsibilities. There was a need for greater clarity about the Data Officers’ role in general and the different expectations of CDOs versus Chief Information Officers (CIOs) (2022). The growing number of data-and technology-related officials in agencies suggested the CDO position could benefit from clearly defining its responsibilities relative to other roles (2023). More than half of CDOs who reported to CIOs found their reporting structure difficult to navigate (2023). The placement of CDOs within agencies was inconsistent, with the majority reporting to CIOs or the agency head but with some reporting to the Chief Financial Officers (CFOs) or Chief Operating Officers.^[86] Previous surveys found the CIOs were the most common reporting position for CDOs (2023).^[87] In 2023, a majority of those CDOs reporting to CIOs noted that they found the reporting structure to be challenging or very challenging (2023). Though the roles are viewed as highly complementary, the reporting structure was viewed less favorably, and the perceived benefit was declining (2024). CDOs reported declining clarity in the responsibilities of the role (2024).
Insufficient support from agency leadership. Although organizational leadership could help alleviate capacity challenges for CDOs by allocating budgets and staff support, one-third of CDOs pointed to leadership resistance as among the top barriers to becoming a data-driven organization, a challenge that applied to the CDO community broadly (2024). The Federal CDO Council should develop templates and talking points to help CDOs communicate value to leadership, bolstering CDOs’ efforts to advocate for resources within their organizations and build stronger relationships with leaders across their organizations. The Council should prioritize a coordinated approach to communicating the value of the CDOs’ role for advancing CDOs’ organizational priorities (2024).
Insufficient White House and OMB guidance and support. CDOs reported they would benefit from increased support from OMB with frequent check-ins and follow-ups (2022). When asked about the need for additional guidance from OMB on privacy (a question new to the 2022 survey), several CDOs stated that there needed to be greater clarification in policies and directives and that there was a need for agencies to have privacy officers who were familiar with privacy policy and implications (2022). While the CDO Council can help with knowledge sharing and networking opportunities for the CDO community, 60 percent of CDOs stated a need for further guidance from the White House and OMB to implement key Evidence Act provisions (2024).
Inadequate ability to leverage data for decision-making. When asked to assess their organization’s data maturity (referring to the extent the organization leverages data for decision-making), the vast majority of CDOs indicated their organization was “a little” or “somewhat” (89%) mature – and only 11% indicated “very” or “completely” mature (2023). CDOs indicated the need for a more coordinated data-governance approach across the federal government, indicating that they faced challenges related to creating a federated versus a siloed approach to data sharing (2024).

The CDOs’ “ever growing call” for clarity on their roles and responsibilities was addressed by the January 15, 2025, OMB guidance on implementing the Evidence Act’s Title II Open Government Data provisions.[88] OMB’s guidance gives CDOs a leading role in making federal agency data “open by default.”^[89] According to a Data Foundation blog, this guidance is “a major shift from the pre-guidance status quo, where agencies lacked clarity on how to implement the Evidence Act’s open data goals.”[90] However, the Data Foundation’s assessment of OMB’s guidance highlighted several implementation concerns, including whether CDOs have the resources, capacity, and support from agency leadership to fully implement OMB’s directives. The Data Foundation concludes by emphasizing the importance of OMB’s leadership in providing “critical coordination and guidance to ensure CDOs are well equipped to carry out their statutory duties to manage federal data assets” and the need for the data policy community “to closely monitor progress and continue advocating for the resources and support needed to realize the full potential of open government data.”

On February 11, 2025, the Data Foundation issued a statement on the evolving federal data and evidence ecosystem in 2025.[91] The statement acknowledged the “significant changes in federal data, evaluation, and evidence policy implementation stemming from recent Executive Orders and administrative decisions” with potential impacts on federal CDOs, evaluation officers, statistical agencies, program evaluators, and research partners. Since February 2025, the Data Foundation has issued monthly “Evidence Capacity Pulse Reports.”[92]

D. Status of the National Secure Data Service Demonstration

Another perspective on interagency data-sharing comes from the NSDS, a 5-year demonstration project to inform a governmentwide effort on strengthening data linkage and data access infrastructure. This demonstration project was originally recommended by the bipartisan Commission on Evidence-Based Policymaking’s final report in 2017; fleshed out in detail in 2022 by the Advisory Committee on Data for Evidence Building, which was established pursuant to the Evidence Act; and finally created by the 2022 CHIPS and Science Act. Responsibility for conducting the 5-year Data Service demonstration project was assigned to the National Center for Science and Engineering Statistics at the National Science Foundation.^[93] Unlike other Evidence Act initiatives, the Data Service is not agency-centric but rather “a new entity entering the evidence ecosystem that provides coordination and capacity-building services, that is, the Data Service will build on the framework of the Evidence Act and supplement, not replace, the work of other evidence enablers and users.^[94]

Data Service initiatives. The demonstration project’s 2024 annual report[95] to Congress set forth several themes:

Data security is the top priority. The majority of the demonstration projects focused on data security. The report noted that data security is paramount for all federal agencies and that they have been slow to embrace new technologies for secure multiparty computation to securely link de-identified data, including new technologies tested by the demonstration. These projects, the report concluded, have shown that the use of privacy-preserving record linkage techniques are feasible and can provide a means to securely link de-identified data while balancing privacy and utility.^[96]
Multiple infrastructure tools were developed to simplify data sharing. Several demonstration projects involved developing new data-sharing infrastructure to address key challenges. First, the Data Service’s demonstration developed a model for a data concierge service to (1) coordinate between agencies assisting data users in answering general questions, (2) identify confidential data assets for users, (3) provide data-linkage support, and (4) help develop evidence-building proposals to apply for access to confidential data. Had it been up and running, this service could have assisted the interagency GI Bill data-linkage project. The eventual scope of the data concierge’s services depends on the adequacy of resources provided by Congress and on the willingness of federal agencies to participate.
Second, the demonstration developed a website to potentially serve as the front door to a Data Service once the demonstration is over in 2027. Third, it has conducted research to establish a prototype dashboard that includes information about projects that use federal data in order to help inform future data use. Finally, the demonstration developed two templates, one for sharing data that is covered under differing statutory mandates and another for data covered under similar statutory mandates.^[97]
Negotiating data-sharing agreements for demonstration projects was challenging. Based on two data security projects, the Data Service’s report noted that interagency data-sharing agreements were challenging because drafting them required significant resources and the time taken to negotiate them was disappointing. Similarly, the Evidence-Based Policymaking Commission’s 2017 report found that data-sharing Memoranda of Understanding between two or more agencies can take years to develop.^[98]
Statutory impediments hamper data sharing. The Data Service’s report cited “inconsistent legal authorities,” which the 2017 Evidence-Based Policymaking Commission’s assessment identified as a fundamental, underlying challenge:“The Commission also cited delays caused by inconsistent legal authorities. To mitigate that risk, they called on Congress and the President to review and amend statutes as appropriate to allow statistical use of data for evidence-building. Recommendations include amending statutes such as Title 13 to allow statistical uses of survey and administrative data for evidence-building within a secure CIPSEA [Confidential Information Protection and Statistical Efficiency Act] environment; repealing current and limiting future bans on the collection and use of data for evidence-building; and enacting statutory changes to ensure state-level data on earnings are available for statistical purposes.”^[99]

The Data Service’s 2024 annual report concluded that scaling-up the demonstration project will require appropriate funding, effective governance, stakeholder engagement, the existence of a robust infrastructure, and a shared services model that supports data access for evidence-building across the nation.

E. Findings of GAO Reports on Evidence Act Initiatives[100]

The Evidence Act requires GAO to monitor and report on the Act’s implementation.^[101] As of December 2024, GAO had released eight reports that highlighted the Act’s implementation successes and challenges. In addition, GAO periodically updates an “issue summary” page on the status of certain Evidence Act initiatives. Themes that emerged from GAO’s work are similar to those identified in the Data Foundation’s and Working Group’s reports and those shared by individuals interviewed for this report. Overall, GAO’s reports helped to separate the easy, early successes from the harder, longer-term challenges.[102] Our analysis identified the following themes:

Capacity issues may reflect resource constraints. In 2021 and 2024 reports and a 2024 issue summary, GAO identified capacity (staff and tools) as a challenge to improving an agency’s ability to pursue evidence-based policymaking.[103] Capacity challenges may well reflect the adequacy of resources in terms of the availability of sufficient and trained staff as well as the dispersion of evidence-based activities across agencies.^[104]
Long-term bureaucratic challenges remain despite short-term progress. GAO obliquely acknowledged the bureaucratic challenges encountered during implementation. For example, its December 2020 report[105] focused on the progress in establishing a data-governance framework at four agencies, including the appointment of CDOs and the creation of a CDO Council. The report noted a warning provided by data experts that GAO had interviewed: Effective implementation of data governance requires a culture change to achieve a shared understanding of the importance of using data as a strategic asset. Echoing GAO’s experts, several individuals interviewed for this report on the interagency project commented that the appointments of CDOs are an example of an easy, short-term success, but getting agencies to understand the importance of using data as a strategic asset is a much harder, long-term challenge. Finally, GAO’s report did not address inconsistent legal authorities that contribute to bureaucratic risk-averse behavior, which often stymies data-sharing projects.
OMB guidance shortcomings. Several GAO reports pointed out that the lack of timely and adequate OMB guidance is a roadblock to the Evidence Act’s effective implementation.[106] An October 2020 report noted that OMB missed a deadline for guidance on data inventories, which was not issued until January 2025, and had not submitted a required biennial report on agencies’ compliance with the Act. Similarly, its 2024 report noted the need for additional OMB guidance to improve agencies’ evidence-building capacities. In addition, GAO also provided examples of agency officials’ missing Evidence Act deadlines, such as conducting certain data assessments or determining staff data literacy skills.
Attributing progress to Evidence Act. GAO’s November 2021 report[107] included survey statistics on the modest improvement in agencies’ use of performance data to shape policymaking. However, isolating the impact of the Evidence Act is not straightforward because the movement to base policy on facts predates the Act, as do initiatives to centralize data use and create CDOs at several agencies. In effect, some federal agencies were not starting from scratch.

F. Interviewees’ Views on the Implementation of Evidence Act Initiatives and on Interagency Data-Sharing Generally Ranged from Skepticism to Patience

Several overlapping themes emerged from discussions with individuals interviewed for this report on the status of the Evidence Act initiatives and interagency data-sharing generally, ranging from skepticism to the need for patience:

Skepticism about overcoming bureaucratic challenges. Bureaucratic delays were a major contributor to the interagency GI Bill data-linkage project’s long timeframes, according to interviewees. Although Evidence Act initiatives may eventually ameliorate some of these bureaucratic roadblocks, changing the bureaucratic culture to encourage data sharing will likely remain a longer-term goal.
Resources needed to implement the Evidence Act. Many of those interviewed believed that insufficient funding of the Evidence Act initiatives was a major barrier to progress. Numerous interviewees pointed out that agencies were not given additional funding to hire staff for the new positions the Act had created, suggesting that the initiatives were not a high congressional priority. Existing staff were simply given new titles; according to one interviewee, many staff members working on data sharing were wearing two hats, retaining their more immediate day-to-day administrative responsibilities and the more esoteric job of advancing an enterprise goal of using data to improve agency operations as a whole. The lack of dedicated funding also means that, within an agency, data-sharing projects rely on the generosity of departmental components to provide the time and resources to make them happen because they require not only staff but funding to execute. The availability of funding will also determine whether the concierge model proposed by the Data Service will be lean or robust.
Privacy-related obstacles have not been addressed. Interviewees noted the ambiguity of statutory privacy provisions. This ambiguity permits agency lawyers’ interpretations of statute to become a serious barrier to data sharing. One interviewee characterized asking Congress to make statutory changes as “a big lift” and another stated that it would be unrealistic to think that congressional committees would align statutory privacy restrictions. According to an interviewee, OMB’s strategy is to identify similarities in agency data-sharing statutes in the hope that would resolve statutory ambiguities, but that strategy remains unproven. Suggesting that little has changed at least at one agency, an interviewee raised new privacy concerns during a January 2025 interview about providing even aggregate data for a recent data-sharing initiative, even though it had been offered to the interagency GI Bill data-linkage project as an alternative to sharing records containing PII.
Concerns about adequacy of CDOs’ authority, responsibilities, and tenure. An overarching concern expressed by interviewees is whether CDOs have sufficient authority absent the support of agency leadership to influence the activities of federal bureaucracies that control administrative and statistical data. One interviewee stated, “Even if you have a great Data Officer, lower-level staff can just ignore you because they [CDOs] don’t have the authority to make data sharing actually happen.” Another interviewee stated that “CDO are offices, not actors. They can stop a lot of things from happening but can’t make things happen.” A different interviewee, however, expressed some optimism because “the CDO Council has put agencies in touch with one other, which should make a difference and make it easier for data-sharing projects like the interagency GI Bill data-linkage project.”^[108] In addition to concerns about sufficient authority, several interviewees also believed that there was a lack of clarity about the responsibilities of CDOs, whose roles overlap with those of CIOs and CFOs.^[109] According to a GAO report, however, some CDOs report to offices with planning and data management responsibilities.^[110] Similarly, interviewees remarked that CDOs face challenges in getting the attention of agency components and leadership, who are typically preoccupied with other pressing business.Tenure and turnover were other concerns raised by interviewees. Although the CDO surveys have some data on the incumbent’s federal experience, it provides an incomplete picture of turnover and vacancies. Several interviewees believed that many CDOs had not been on the job for very long and “weren’t sticking around,” and, as a result, there was “not a lot of continuity.” One explanation offered for the turnover was that the position of CDO was not an “empowered job with a lot of levers.” A different interviewee suggested that data should be collected on CDO turnover and the proportion who were “acting” CDOs. ED has had the CDO position change hands once, while at VA the third CDO stepped into this role in 2024 after serving in an acting capacity for almost 2 years.[111] DOD also is on its third CDO since July 2018.^[112] At a fourth agency, one interviewee commented that the CDO did not really want the job but saw it as a stepping stone. Lack of clarity about roles and authorities could serve to undermine the desirability of the position.
Focus to date has been on technological issues. Several interviewees stated that technological solutions were easier to address than bureaucratic challenges. Echoing their comments, a February 2021 presentation by ED’s then-CDO noted that significant attention had been focused on the “technological solutions” concerning the secure transmission and integration of data as opposed to “people and processes,” that is, the “bureaucratic processes” cited in the 2017 Commission report.[113] Such technological solutions do not address what he termed the “ad hoc and costly exercise” that continues to inhibit data linkage projects and to “regularly require leadership intervention to overcome.”
Need for more data-linkage success stories. Another theme from interviews was that both the Evidence Act and CDOs are focused on process rather than helping to facilitate actual data sharing. As one interviewee noted, CDOs are currently distracted with the “big picture”—working on establishing a framework or plan for data sharing—and not working on actual data sharing. A different interviewee’s comments expanded on this critique: “The Chief Data Officers are mostly data enterprise people who focus on infrastructure and the nuts and bolts—the ‘system and mechanics’—for sharing data. They don’t think that much about enhancing policymaking through data sharing; that is, rather than generating value, they are just producing a lot of reports on using evidence without actually doing so.” Reflecting this focus on process, another interviewee commented that the annual reports are transmitted only to congressional committees that focus on process not policy, so no one is asking CDOs to demonstrate results that support policymaking. This interviewee and others acknowledged, however, that some of the process initiatives are important, such as standardized data-sharing applications or a library of data-sharing examples, but believed it remains to be seen to what extent they will be used. On the other hand, when asked about the perceived focus of CDOs on process, one interviewee was able to provide concrete examples of actual data-sharing projects that had either been completed or proposed, which underscores the difficulty of generalizations.^[114]
Adequacy of overarching coordinating mechanisms. Some interviewees were concerned that, given the size of this new ecosystem, there are insufficient overarching coordinating mechanisms to ensure all of these new team members share common goals and are speaking to one another. As noted earlier, OMB was given the important role of coordinating evidence-based policymaking, but one interviewee commented that OMB is “missing in action” and that no one from OMB is driving a proactive agenda across agencies on data sharing. As discussed earlier, GAO found that OMB had missed a July 2019 deadline to issue guidance on implementing the Evidence Act’s Title II Open Government Data provisions, which includes a requirement for comprehensive data inventories. That guidance was not issued until January 2025, more than 5 years later. Moreover, the guidance does not address inconsistent and ambiguous legal authorities identified by the 2017 Commission report and others, which contributes to risk aversion with respect to data sharing. Rather, it requires agencies to evaluate the risk of disclosure versus public access to data, creating uncertainty about whether an agency can still conclude it does not want PII to leave its servers unless there is an explicit legal prohibition.In December 2024, OMB implemented a new Title III rule—the so called Trust Regulation—intended to address the decentralization of statistical agencies, which are often situated within parent agencies, such as the Census Bureau (Commerce), the Bureau of Labor Statistics (Labor), and the National Center for Education Statistics (ED’s Institute for Education Sciences).^[115] Striking the right balance between centralization and autonomy will be challenging as differing assessments of the Trust Regulation suggest.[116] One interviewee expressed a positive attitude about OMB’s Trust Regulation while acknowledging that establishing such standardization is a difficult transformation that will take some time.
Many interviewees acknowledged that the complexity and scope of the transformation envisioned by the Evidence Act demanded patience. The initiatives are still in their infancy and governmentwide adoption of evidence-based policymaking could take some time to mature. “A few small steps each year should be counted as progress,” according to one interviewee. Interviewees also agreed that it was difficult to assess the progress made to date by the dozens of agencies involved in the Act’s implementation. Finally, one interviewee noted that, compared to 2016 when the interagency GI Bill data-linkage project was initiated, there has been an improved awareness of the benefits of data sharing that should help to move similar projects along more quickly. Another interviewee emphasized that acknowledging the rewards as well as the risks of data sharing is a prerequisite to having a productive conversation about balancing these two objectives. Overall, interviewees expressed hope that the Evidence Act’s initiatives will eventually succeed. As one interviewee stated: “The Evidence Act is a good idea; that is, it is necessary but not sufficient. It is going to take a long time for the Act’s initiatives to mature, and overcoming the bureaucracy and the disincentives to data sharing will not be easy.” Another interviewee offered that it is important to keep “nudging, encouraging, and reminding Congress about what needs to be done to jump start data sharing.”

VIII. CONCLUSIONS

The interagency GI Bill data-linkage project produced the first-ever data and analysis of education and labor market outcomes from the nearly $100 billion spent on the Post-9/11 GI Bill in its first decade since 2009 enactment. By sharing data across agencies, the interagency project was able to draw on multiple sources of data covering an unprecedented dataset size of 2.7 million veterans. Through careful collaboration across agencies, the interagency team was able to bring in the Defense Department’s data on service members’ academic preparation at the time of enlistment (an important “control” for the assessment of later student outcomes) as well as IRS data on earnings (the gold standard for researchers on earnings). The interagency team produced five important reports detailing which veterans are using the Post-9/11 GI Bill and how their educational and labor market outcomes differ by demographic and military characteristics as well as by institutional characteristics, type of educational program, and field of study, and, additionally, which veterans are not using their Post-9/11 GI Bill benefits and how they are faring economically. The team found, for example, that veterans’ graduation rates are nearly double that of comparable financially independent students, and that an institution’s instructional spending levels had a high degree of correlation with veterans’ labor market success. The team was also able to conduct deep dives on specific population groups and regions of the country.

But the lengthy timeframe for the project highlighted existing bureaucratic obstacles to data sharing. Through interviews with individuals involved in the project, we learned that the principal contributors to the approximately 8 years of negotiations and delays were a combination of (1) bureaucratic culture, (2) statutory privacy concerns, and (3) unpredictable situational delays. Although placement of the interagency project in the evidence-building component of the Census Bureau was intended to preempt anticipated agency privacy concerns, it also created challenges. For example, the Census Bureau assigned a higher priority to its statutorily required decennial census, and its reorganization of its evidence-building unit led to a lower priority being assigned to such activities.

Through a review of reports regarding the implementation of the Evidence Act, we identified cultural elements that feed the bureaucratic responses and delays, including (1) stove-piping of agency data, (2) feelings of proprietary ownership of data by agency officials, and (3) lack of consistent support for interagency data-sharing and for the CDOs who could enable better use of data.

Despite universal agreement that the concept and benefits of evidence-based policymaking are worthwhile, it remains to be seen if or when the Evidence Act initiatives will succeed in overcoming the roadblocks to data-linkage projects. At a minimum, strong oversight is imperative. Changing the bureaucratic culture that each agency and component within an agency owns their data calls for the same persistence and perseverance exhibited by the interagency GI Bill project. Concerns about the adequacy of CDOs’ authority, responsibilities, and tenure must be addressed. Securing adequate resources and addressing statutory ambiguities will require congressional action. Congress must also enact the Transparency Act to create a student unit record system that addresses both the longstanding shortcoming in ED’s data on student outcomes and ED’s data ownership bias.

Finally, as a February 2025 statement by the Data Foundation noted, current efforts to reduce the size of the federal workforce also may have a negative effect on interagency data sharing.

APPENDIX A

RESEARCH QUESTIONS AND KEY FINDINGS FROM
FIVE REPORTS OF THE INTERAGENCY GI BILL DATA-LINKAGE PROJECT

A First Look at Post-9/11 GI Bill-Eligible Enlisted Veterans’ Outcomes^a

Who uses Post-9/11 GI Bill, what are their postsecondary outcomes, and what are their labor market outcomes?

▪ About half (54%) of Post-9/11 GI Bill-Eligible Enlisted Veterans used their Post-9/11 GI Bill benefits, about half (47%) of Post-9/11 GI Bill users completed a degree within six years, and Post-9/11 GI Bill users with an associate or bachelor’s degree earned around $50,000 ($44,100 and $55,700, respectively).

▪ Use of benefits, degree completion, and earnings increase with academic preparedness as measured by aptitude test scores prior to enlistment.

▪ Female veterans use benefits and complete degrees at higher rates than male veterans, but their earnings lagged behind.

▪ Veterans from historically underrepresented groups were more likely to enroll but less likely to complete.

▪ Unmarried veterans with dependents were less likely to complete a degree; married veterans were more likely to complete a degree and to earn more.

▪ Veterans with disability ratings were more likely to use benefits, but groups with the highest disability ratings were less likely to complete degrees and earned less.

▪ Veterans in rural areas were less likely to use Post-9/11 GI Bill benefits and were less likely to complete a degree, and they earned less.

Source: Extracts from Radford, A.W, P. Bailey, A. Bloomfield, B. H. Webster, Jr., & H. C. Park (2024). A first look at post-9/11 GI Bill-eligible enlisted veterans’ outcomes. Washington, D.C.: American Institutes for Research, U.S, Census Bureau, and National Center for Veterans Analysis and Statistics at the U.S. Department of Veterans Affairs, February. The report is available at https://vetsedsuccess.org/wp-content/uploads/2024/02/First_Look_PGIB_Outcomes_2-20-24.pdf.

^aThe decision to focus on veterans who were enlisted and on active duty as of their last recorded pay plan was driven by the reality that a prerequisite for most officers is to have a bachelor’s degree prior to joining the military. In contrast, individuals at the enlisted rank constitute the vast majority of servicemembers; most enlist without a postsecondary degree and are thus more likely to benefit from the Post-9/11 GI Bill. Data on dependents’ use of benefits was not yet available from VA. The First Look report noted that examination of both officers’ and dependents’ use of benefits could be future projects.

How Do Veterans’ Outcomes Differ Based on the Type of Education They Received? And How Are Veterans Who Have Not Used Their Education Benefits Faring?

These two research questions examined how veteran characteristics such as sex, race/ethnicity, rurality, aptitude test scores, and military rank are associated with benefits use and earnings.

▪ Slightly more than half of veterans (51%) did not personally use the Post-9/11 GI Bill (i.e., were “Nonparticipants”), but this varied by veteran characteristics. Compared to veterans at large, veterans who settled in rural and micropolitan areas and veterans who left the service in the two lowest military ranks and in the three highest military ranks were at least 5 percentage points more likely to be Nonparticipants. In contrast, female, American Indian/Alaska Native, Black, and Hispanic veterans, as well as veterans who left the service with a midlevel military rank, were at least 5 percentage points less likely to be Nonparticipants than veterans at large.

▪ On average, veterans who did not personally use Post-9/11 GI Bill benefits were earning $44,800 in 2018. Female Nonparticipants and Nonparticipants in the three lowest military ranks were earning at least $10,000 less than Nonparticipants overall, and American Indian/Alaska Native and Black Nonparticipants were earning at least $5,000 less. Meanwhile, veterans in the two highest ranks who did not use the Post-9/11 GI Bill were earning at least $10,000 more than the average Nonparticipant, and veterans in the three ranks below them were earning at least $5,000 more.

▪ A relatively small proportion of veterans used the Post-9/11 GI Bill for a nondegree program. About 5% of veterans used the Post-9/11 GI Bill for a nondegree program at a non-IPEDS^a provider and 6% did so at an IPEDS institution. Diving deeper into education provider type, the most common provider for nondegree seekers was a for-profit non-IPEDS provider, where 4% of veterans enrolled. The two next most common providers were two-year public and two-year for-profit IPEDS colleges at 2% each. Regardless of IPEDS status or the control or type of provider, there were few differences in participation rates by sex, rurality, race/ethnicity, or military rank.

▪ Veterans in nondegree programs earned an average of $40,400 at non-IPEDS providers and an average of $37,100 at IPEDS institutions. At both non-IPEDS and IPEDS providers, veterans who used the Post-9/11 GI Bill for nondegree programs at for-profit providers had substantially lower earnings than those who did so at public providers.

▪ More degree-seeking veterans used Post-9/11 GI Bill benefits at IPEDS than non-IPEDS providers. Specifically, 43% of veterans used the Post-9/11 GI Bill for a degree program at an IPEDS provider, whereas only 6% used the Post-9/11 GI Bill for a degree program at a non-IPEDS provider. Compared to the average veteran, female, Black, and Hispanic veterans pursued degrees at IPEDS institutions at higher rates, whereas veterans who settled in rural and micropolitan areas did so at lower rates. As for the type of provider where veterans pursued degree programs, enrollment was most common at two-year public IPEDS colleges (18%) followed by four-year public (14%), four-year for-profit (11%) and four-year nonprofit (9%) IPEDS colleges.

▪ Veterans who pursued a degree at an IPEDS institution earned $44,700, whereas veterans who did so at a non-IPEDS provider earned $39,300. By veteran characteristics, for every veteran group examined, the earnings of those who pursued a degree at an IPEDS institution were higher than the earnings of those who did so at a non-IPEDS provider. Drilling down to look at specific types of providers, both within non-IPEDS and IPEDS providers, earnings for veterans at for-profit providers were lower than for veterans at public providers, and a regression analysis for IPEDS providers found that this remained true at the two-year level, after accounting for other veteran characteristics. Two-year for-profit IPEDS institutions also cost the Post-9/11 GI Bill program more in terms of average payment per veteran than two-year public IPEDS institutions did.

▪ Overall, veterans were more likely to pursue degrees at institutions with lower levels of instructional spending—and the earnings of these veterans’ were lower than the earnings of veterans’ who pursued degrees at institutions that spent more on their instruction. Veteran’s degree completion rate increased sharply as their institution’s instructional spending quintile increased. Approximately 40% of “Clearinghouse veterans” (shorthand for institutions that report to the National Student Clearinghouse) who attended an institution in the lowest quintile for instructional spending completed their education compared to 71% of those in the highest quintile—a 31% percentage-point difference. In short, there was a significant increase in veterans’ completion rates as their institutions’ instructional spending increased. These results held true during a regression analysis. Earnings also rose with each quintile of instructional spending. IPEDS degree seekers had better earnings the higher the quintile with veterans in the fifth quintile standing out as having the highest earnings. Specifically, veterans who first pursued a degree program at an institution in the lowest instructional spending quintile earned an average of $41,600, or $16,700 less than the average earnings of those who first pursued a degree program at an institution in the highest instructional spending quintile. Even accounting for other veteran characteristics in a regression analysis, the gap in earnings between veterans at institutions in the lowest and highest spending quintiles stood at $9,100. However, only 1% of veterans pursued a degree program at an IPEDS institution in the highest quintile of instructional spending, whereas 17% of veterans’ pursued a degree program at an IPEDS institution in the lowest quintile of instructional spending.

▪ About 13% of veterans’ pursued a degree program at a college that fell in the highest quintile in terms of distance education prevalence, whereas only 2% of veterans’ attended an institution that fell into each of the lowest three quintiles. Female and Black veterans were more likely than veterans at large to attend an institution in the highest quintile. As for earnings, in contrast to the clear gap observed by institutions’ instructional spending, the gap by institutions’ prevalence of distance education was small. The minimal variation in earnings by distance education prevalence that did exist followed a “U-shaped” pattern, with veterans who attended an institution in the highest distance education quintile having the highest earnings, followed closely behind by veterans attending an institution in either of the two lowest distance education quintiles.

▪ Six-year degree completion rates for veterans were roughly double those of independent students. Six-year completion rates can be determined by looking at institutions that report to the National Student Clearinghouse (“Clearinghouse Veterans”). About 47% of veterans’ who attended institutions reporting to the Clearinghouse completed a degree within six years—roughly double that of independent students—but completion gaps varied widely by sector (24 percentage points), by instructional spending (31 percentage points), and, to a lesser extent, by distance education prevalence (10 percentage points). The Post-9/11 GI Bill program spent roughly the same amount of money per veteran attending four-year for-profit and four-year public colleges, even though veterans at four-year for-profit colleges had a completion rate 15 percentage points lower than those at four-year public colleges after accounting for other factors. Those in the highest quintile for instructional spending had the highest earnings and completion rates, but only 1% of veterans attended institutions in this quintile.

Source: Extracts from Radford, A.W., P. Bailey, A. Bloomfield, B.H. Webster Jr., & H. C. Park (2024). Post-9/11 GI Bill Benefits: How Do Veterans’ Outcomes Differ Based on the Type of Education They Received? And How Are Veterans Who Have Not Used Their Education Benefits Faring? Washington, D.C.: American Institutes for Research, U.S. Census Bureau, and National Center for Veterans Analysis & Statistics at the U.S. Department of Veterans Affairs, July. The report is available at https://vetsedsuccess.org/wp-content/uploads/2024/10/pgib-outcomes-by-use-enrollment-characteristics.pdf.

^aIPEDS stands for the Integrated Postsecondary Education Data System. All schools participating in federal student aid must report data to IPEDS.

Note: The interagency GI Bill data-linkage project analyzed veteran outcomes at nondegree institutions that participated in federal student aid and that reported to ED’s IPEDS with data from VA on students who also attended nondegree institutions that do not participate in federal student aid.

Post-9/11 GI Bill-Eligible Enlisted Veterans’ Enrollment and Outcomes at Public Flagship Institutions, with a Focus on the Great Lakes Region

(1) What percentage of veterans’ using their Post-9/11 GI Bill benefits use them at flagship universities? (2) What percentage of veterans’ completed bachelor’s degrees? (3) What were veterans’ earnings after bachelor’s degree completion? (4) What were the outcome differences by race/ethnicity and flagship status?

▪ Both nationally and in the Great Lakes region, 3% or less of Post-9/11 GI Bill-eligible enlisted veterans first enrolled at public flagship universities. For context, analysis of IPEDS enrollment data indicates that 6% of all undergraduate students enrolled at flagship universities in the 2020‐21 academic year.

▪ The study found that veterans who first enrolled at flagship institutions completed a bachelor’s degree at a higher rate on average than veterans who first enrolled at other institutions, but they completed degrees at a slightly lower rate than that of all students who attended flagships.

▪ The results indicate that veterans used their Post-9/11 GI Bill benefits to attend flagship universities, but at a rate lower than that seen in the general population (2% vs. 6%). Attendance at flagship universities was positively associated with the likelihood of completing a bachelor’s degree in six years (60% vs. 45%), and this relationship held true in a regression analysis controlling for a host of variables likely to influence completion, such as academic preparedness, military experiences, and demographic characteristics. Similarly, earning a bachelor’s degree at a flagship university was associated with higher wages (3% higher on average) and this relationship remained significant even after controlling for other variables.

▪ The same relationships found for flagship universities overall were present for flagships within the Great Lakes region (Illinois, Indiana, Michigan, Minnesota, Ohio, and Wisconsin) in comparison to other four‐year institutions in that region, with some minor variations. Specifically, the veterans’ earnings gap between flagship and non-flagship institutions in the Great Lakes region was larger than the national earnings gap, and Great Lakes region veterans completed non‐flagship four‐year institutions at higher rates than their peers in other regions of the country, while the completion rate at flagships was the same for veterans in the Great Lakes and nationally.

▪ In our deeper dive into the national data, our analysis found variation by veterans’ race/ethnicity in how flagship status was related to enrollment and completion, but not earnings. In particular, Black veterans were less likely to enroll at a flagship university than veterans at large (less than 1% vs. 2%). However, Black veterans who attended a flagship university were nearly 10 percentage points more likely to complete in six years than those who attended a non-flagship, after controlling for other veteran characteristics. This significant difference in completion rates is important to policy conversations about how to increase college enrollment and attainment for Black Americans and merits additional study. American Indian/Alaska Native veterans were more likely to attend flagships than veterans at large (4% vs. 2%) but their bachelor’s degree completion rate did not significantly differ by flagship attendance after controlling for other variables. Understanding why the apparent flagship advantage in completion did not extend to American Indian/Alaska Native veterans the way it did for Black veterans and veterans of other racial and ethnic backgrounds merits additional study.

▪ It is important to remember that there are many other potential variables that could affect outcomes that the study was not able to control for in its analyses. For example, parental income has been found in previous research to be related to student outcomes in postsecondary education and the percentage of Pell Grant recipients at flagship universities is lower than the percentage of Pell Grant recipients at other public universities, but the study was not able to include parental income. It is also possible the flagship universities, which tend to be better resourced than other four‐year public institutions, are offering services that increase completion rates for veterans who enroll, such as more hands-on academic counseling or early interventions for struggling students. In addition, completion of a bachelor’s degree from a flagship may affect wages because employers perceive flagships as producing more well‐prepared graduates. Finally, the study was unable to measure intangible factors like personal motivation.

Source: Extracts from Bloomfield, A., A.W. Radford, P. Bailey. B. H. Webster Jr., & H. C. Park (2024). Post-9/11 GI Bill-eligible enlisted veterans’ enrollment outcomes at public flagship institutions, with a focus on the Great Lakes region. Washington, D.C.: American Institutes for Research, U.S. Census Bureau, and National Center for Veterans Analysis & Statistics at the U.S. Department of Veterans Affairs, July. The report is available at https://vetsedsuccess.org/wp-content/uploads/2024/10/pgib-outcomes-public-flagship-great-lakes.pdf.

Which Veterans Are Forgoing Their Post-9/11 Gi Bill Benefits?

This report identifies the proportion of Post-9/11 GI Bill-eligible enlisted veterans who forgo those benefits and examines the demographic and military characteristics associated with benefit nonuse. The sample included veterans who met specific criteria related to Post-9/11 GI Bill eligibility, separation date, age, rank, and educational background.

§ Overall, 38% of veterans did not personally use or transfer their benefits (i.e., were Nonusers). Yet this percentage varies by academic preparation at time of enlistment, age, race/ethnicity, sex, family responsibilities, rurality, disability rating, rank, and military branch. Overall, nonuse was highest for veterans who separated at age 55 to 65 (82%). Nonuse was lowest (in other words, use of the Post-9/11 GI Bill was highest for those who left the military with a midlevel rank of E-4 or had a VA disability rating of 10% to 20% (27%).§ Academic preparedness. Results revealed that veterans in lower Armed Forces Qualification Test quintiles were more likely to be Nonusers (i.e., were less likely to use their Post-9/11 GI Bill benefits) than those in higher quintiles. The AFQT measures arithmetic reasoning, mathematical knowledge, paragraph comprehension, and word knowledge of incoming service members, and thus can provide a snapshot of veterans’ academic preparedness at the time they enlisted.

§ Age. For the most part, the older that veterans were when they separated from the military, the more likely they were to be Post-9/11 GI Bil Nonusers.

§ Race/Ethnicity. We found that White, Asian, and non-Hispanic veterans were more likely to be Nonusers than veterans from racial and ethnic groups that have been historically underrepresented in postsecondary education. In other words, American Indian/Alaska Native, Black, and Hispanic veterans were more likely to use Post-9/11 GI Bill benefits.

§ Sex. Male veterans were 8 percentage points more likely than female veterans to be Nonusers. This gap is consistent with national patterns that show men in America generally enroll in postsecondary education at lower rates than women.

§ Family responsibilities. Married veterans with dependents were 9 percentage points more likely to be Nonusers than married veterans without dependents, single veterans without dependents, and single veterans with dependents. In short, married veterans with dependents were the least likely to use Post 9/11 GI Bill benefits.

§ Rurality. The less urban the community in which veterans reside, the more likely veterans are to be Nonusers, with rural veterans the least likely to use their Post-9/11 GI Bill benefits.

§ Disability rating. Veterans with no disability rating and those with a disability rating of 100% were more likely to be Nonusers (i.e., less likely to use Post-9/11 GI Bill benefits) than those with disability ratings between 0% and 90%.

§ Rank. In considering these results, it is important to note that military rank is both an indication of the length of service and a sign of success in the military. Gaps in nonuse by rank were large, stretching as much as 49 percentage points, with veterans who separated from the military at the lowest ranks (E1-E2) and the highest ranks (E6-E9) much more likely to forgo Post-9/11 GI Bill benefits than those in the middle ranks (E3-E5).

§ Service. Air Force veterans were most likely to forgo Post-9/11 GI Bill benefits (45%), followed by veterans from the Army (41%) and Coast Guard (39%). Those in the Navy and Marine Corps were least likely to be Nonusers at 32% and 30%, respectively. Gaps between Air Force veterans and veterans from other branches generally shrank after we accounted for other characteristics, although the gap did not shrink between Air Force and Coast Guard veterans.

Source: Extracts from Radford, A.W., K.M. Mayer, A. Bloomfield, P. Bailey, B.H. Webster, Jr., and H.C. Park (2025). Which Veterans Are Forgoing Their Post-9/11 GI Bill Benefits? Washington, D.C.: American Institutes for Research, U.S. Census Bureau, and National Center for Veterans Analysis & Statistics at the U.S. Department of Veterans Affairs, February. The report is available at https://www.air.org/sites/default/files/2025-02/which-veterans-are-forgoing-post-9-11-gi-bill-benefits.pdf.

A Deeper Look at Post‐9/11 GI Bill Outcomes for American Indian/Alaska Native, Black, and Hispanic Veterans

This report takes a deeper look at outcomes for American Indian/Alaska Native, Black, and Hispanic enlisted veterans who were eligible for the Post-9/11 GI Bill (henceforth referred to as “veterans” for brevity). More specifically, it looks at these groups’ use of the Post-9/11 GI Bill, degree attainment, and subsequent earnings.

§ Sex. Compared to veterans at large of the same sex, American Indian/Alaska Native, Black, and Hispanic veterans consistently were more likely to use Post-9/11 GI Bill benefits but less likely to complete a degree (with one exception). However, within these racial/ethnic groups, females consistently were more likely than males to use the Post-9/11 GI Bill and complete a degree. Compared to veterans at large with the same degree attainment and sex, American Indian/Alaska Native and Black veterans consistently earned less but Hispanic veterans earned more in several instances. Finally, within these racial/ethnic groups, females consistently earned less than males when they attained an associate degree and when they attained a bachelor’s degree.

§ AFQT quintile. Compared to veterans at large in the same AFQT quintiles, (1) American Indian/Alaska Native, Black, and Hispanic veterans were consistently as likely or more likely to use Post-9/11 GI Bill benefits, and (2) American Indian/Alaska Native veterans were consistently less likely to complete a degree but Black veterans in three AFQT quintiles were more likely to complete, and Hispanic veterans’ were consistently as likely or more likely to complete across all five AFQT quintiles. Compared to veterans at large with the same degree attainment and the same AFQT quintile, American Indian/Alaska Native and Black veterans consistently earned less but Hispanic veterans often earned more. Finally, among American Indian/Alaska Native, Black, and Hispanic veterans, use of Post-9/11 GI Bill benefits, degree completion, and earnings increased the higher veterans’ AFQT quintile—with a couple of exceptions.

§ Family responsibilities. (1) Compared to veterans at large with the same Family responsibilities, American Indian/Alaska Native, Black, and Hispanic veterans were consistently more likely to use Post-9/11 GI Bill benefits but frequently less likely to complete a degree. (2) Among American Indian/Alaska Native, Black, and Hispanic veterans, unmarried veterans were more likely than married veterans to use Post-9/11GI Bill benefits, but married veterans were more likely than unmarried veterans to complete a degree. (3) Compared to veterans at large with the same degree attainment and the same family responsibilities, American Indian/Alaska Native and Black veterans consistently had lower average earnings, whereas Hispanic veterans often had higher average earnings. (4) Finally, Among American Indian/Alaska Native, Black, and Hispanic veterans who attained either an associate degree or a bachelor’s degree, those who were single with dependents earned the least, and those who were married with dependents earned the most.

§ Rurality. Compared to veterans at large in the same rurality categories, (1) American Indian/Alaska Native, Black and Hispanic veterans consistently were more likely to use Post-9/11 GI Bill benefits, and (2) American Indian/Alaska Native and Black veterans were consistently less likely to complete a degree, but Hispanic veterans from rural and micropolitan areas were more likely to do so. Compared to veterans at large with the same degree, micropolitan areas were more likely to do so. (3) Compared to veterans at large with the same degree attainment and rurality category, American Indian/Alaska Native and Black veterans consistently earned less, whereas Hispanic veterans earned more (with one exception). (4) Finally, among American Indian/Alaska Native, Black, and Hispanic veterans who attained either an associate degree or a bachelor’s degree, those in metropolitan areas earned the most (with one exception).

Source: Excerpts from Radford, A.W., A. Bloomfield, P. Bailey, K.M. Mayer, B.H. Webster, Jr., and H.C. Park (2025). A Deeper Look at Post‐9/11 GI Bill Outcomes for American Indian/Alaska Native, Black, and Hispanic Veterans. Washington, D.C.: American Institutes for Research, U.S. Census Bureau, and National Center for Veterans Analysis & Statistics at the U.S. Department of Veterans Affairs, February. The report is available at https://www.air.org/sites/default/files/2025-02/deeper-look-at-post-9-11-gi-bill-outcomes.pdf

APPENDIX B

INDIVIDUALS CONTACTED FOR THIS REPORT

In researching this report, we contacted 30 individuals who (a) helped to conceptualize the project; (b) negotiated the data sharing agreements, analyzed the linked data, and wrote the project’s reports; (c) are recognized experts on data sharing; and/or (d) are involved in implementing the Foundations for Evidence-Based Policymaking Act.^[117] We interviewed 27 individuals and three additional individuals responded by email to discrete questions, which provided context on several issues.

We developed standardized protocols that differed across interviewees depending on the extent of their involvement in the project and their area of expertise, such as Evidence Act implementation. The interviews were conducted between November 2024 and February 2025. To encourage candid responses, this report does not attribute insights obtained during the interviews to specific individuals. The names and expertise of interviewees are listed below, except for one individual who asked not to be identified.

Name	Why interviewed
Paul Arnsberger	Program manager at IRS dealing with data access and disclosure issues. Served as IRS point of contact for the interagency GI Bill data-linkage project.
Shannon Arvizu	Senior Advisor to Commerce Department’s Chief Data Officer; Public Services fellow at Georgetown University, working with state-level Chief Data Officers; and former staff member of the Chief Data Officers’ Council’s Data-Sharing Working group.
Ashley Austin	Program manager, Census Bureau, Center for Administrative Records Research and Applications, specializing in interagency agreements. Interagency GI Bill data-linkage project staff member through February 2019.
Paul Bailey	Principal economist at American Institutes for Research, joining in 2012. Interagency GI Bill data-linkage team member since late 2016 or early 2017. Contributor to all five interagency project reports as a special-sworn-status temporary employee of the Census Bureau during the project. PhD in economics from U-Maryland; Master’s degree in statistics from University of Chicago.
Scott Boggess	Economist at Census Bureau, interagency data-linkage project staff member directing negotiations for data sharing agreements from 2016 to 2023.
Jack Buckley	Senior Vice President for Research, American Institutes for Research, December 2016 through January 2019, supervising researchers assigned to the interagency GI Bill data-linkage project for approximately 1 year as a special-sworn-status temporary employee of the Census Bureau during the project.
Patrick Campbell	Policy Analyst and Deputy Assistant Director, Office of Servicemember Affairs, Consumer Financial Protection Bureau, 2015-2019, detailed to VA’s Education Service to supervise the crosswalk of data between ED and VA for the GI Bill Comparison Tool.
Melissa Chiu	Program manager for evaluation and outreach, Census Bureau, Center for Administrative Records Research and Applications, during the early phase of interagency GI Bill data-linkage project and then Acting Executive Director, Data Governance and Analytics, Office of Enterprise Integration, Department of Veterans Affairs. Currently, Director, the Committee on National Statistics at the National Academies of Sciences, Engineering, and Medicine.
Rajeev Darolia	Senior Advisor and Chief Economist, Department of Education, beginning in 2023 and January 2024, respectively.
Patrick Dworakowski	Assistant Director, Oversight and Accountability Division, Education Service, Veterans Benefits Administration, Department of Veterans Affairs.
Joe Garcia	Director, Education Service, Veterans Benefits Administration, Department of Veterans Affairs.
Michael S. Kofoed	Professor of Economics, U.S. Military Academy, 2014-2023 and author of an October 2020 Brookings report on veterans’ outcomes using Army Armed Forces Qualification Test scores.
Stephanie Logan	Deputy Director, Education Service, Veterans Benefits Administration, Department of Veterans Affairs.
Erika McEntarfer	Labor Economist, Center for Economic Studies, Census Bureau, 2010-2024, who worked on data-sharing negotiations with the Clearinghouse to post earnings data for institutions of higher learning. No agreement was reached.
Angella McGinnis	Data-sharing point of contact at DOD’s Defense Manpower Data Center.
Heather Madray	Formerly with Census Bureau, Center for Enterprise Dissemination. Currently Supervisory Program Director, National Center for Science and Engineering Statistics at the National Science Foundation and part of a team leading the National Secure Data Service demonstration project.
Jordon Matsudaira	Deputy Under Secretary, 2021-2024, and Chief Economist, 2022-2024, Department of Education.
Amy O’Hara	Director, Census Bureau’s Center for Administrative Records Research and Applications, evidence-building division, 2004-2017. Currently, research professor at Georgetown University.
Hyo Park	PhD, Senior Statistician at the National Center for Veterans Analysis and Statistics, which is a component of the Office of Enterprise Integration, Data Governance and Analytics, Department of Veterans Affairs; former statistician at Census Bureau (1998-2009); and contributing author to all five Interagency GI Bill data-linkage project reports as a special-sworn-status temporary employee of the Census Bureau.
Alexandria Walton Radford	Senior director, American Institutes for Research, Center for Applied Research in Postsecondary Education. Joined the interagency GI Bill data-linkage project in November 2020 as lead analyst. Contributor to all five interagency project reports as a special-sworn-status temporary employee of the Census Bureau during the project. PhD and Master’s in Sociology from Princeton University.
Ross Santy	Chief Data Officer, Department of Education, since 2023, and Associate Commissioner, Administrative Data Division, National Center for Education Statistics (Integrated Postsecondary Education Data System), 2013-2023.
Doug Shapiro	Vice President, Research, and Executive Director, Research Center, National Student Clearinghouse.
Mark Schneider	Vice President, American Institutes for Research, 2016-2018, leading the interagency GI Bill data-linkage project as a special-sworn-status temporary employee of the Census Bureau during the project through early 2018, when he left to head the Institute for Education Statistics at the Department of Education.
Bill Skimmyhorn	Professor of Economics and Long Term Research Coordinator, Office of Economic & Manpower Analysis, U.S. Military Academy, West Point, 2012-2018, and co-author of a 2021 report on veterans’ outcomes using Army Armed Forces Qualification Test scores.
Kathy Stack	Director, Office of Management and Budget’s evidence team, 2013-2015, and Vice President, Arnold Ventures, for several years starting in 2015.
Eddie Thomas	Director, Data Governance and Analytics, Office of Enterprise Integration, and Statistical Official at Department of Veterans Affairs. Member of the interagency project’s team until his promotion to Director.
Kate Tromble	Vice President, Federal Policy, Data Quality Campaign.
Bruce Webster	Statistician at Census Bureau, interagency GI Bill data-linkage project staff member since 2020 and contributing author to all five interagency project reports.
Carrie Wofford	President of Veterans Education Success who helped to conceptualize the interagency GI Bill study.

Source: Author’s records of interviews. All individuals were interviewed via Google Meets.

APPENDIX C

NOTEWORTHY INTERVIEWEE COMMENTS ON INTERAGENCY GI BILL
DATA-LINKAGE PROJECT’S CHALLENGES IN
DEALING WITH AGENCY BUREAUCRATS

Bureaucratic culture: There was a lot of bureaucratic “red tape.” The process is so bureaucratic, with so many rules that apply across different agencies. It’s so complicated. It felt like we were always going back and forth, and every meeting took ages to schedule, and by the time a meeting took place, the people had forgotten stuff, staff members had changed, and you were going back to square one. The agreements were created one after another–de novo. Each agency was different and there was no learning from one agreement to the next.

Whom to ask: Not knowing who to ask was an obstacle to data sharing. Finding the right people at an agency to navigate the data sharing process was a big concern and challenge.

Importance of mindset: The success of data sharing project depends on the personality of the individuals you are dealing with, and even if that person changes, the project will proceed to a successful conclusion if that person is replaced by someone of a like mindset. On the other hand, mindset leads people to accept restrictive interpretations of ambiguous statutory language even when those interpretations are arguable.

Sufficiency of high-level support: If someone in leadership wants it to happen and devotes time and resources to it, data will be shared. Without such sustained support, data sharing projects will continue to run into brick walls.

Communications roadblocks: The interagency project’s researchers talk to administrative types at each agency—they never actually speak with the person who uses the agency’s data. There was usually a central contact at an agency who wasn’t the data user. Data providers don’t ask helpful follow up questions, such as “have you thought about using his data element?” They tend to view being open about the data available as exposing them to risk and lean towards providing as little as possible.

Confusion over data elements: There was a lot of confusion during the negotiations about the specific data elements requested, what level of detail might be included, and their sensitivity levels. This issue added to the difficulty of negotiating agreements because they were constantly having to spell out long lists of specific data elements that each side wanted to include or exclude from data sharing. The descriptors often used shorthand—age or DOB—without articulating what data were actually included in each data element.

One way door: Everyone thinks the data should come to them and that they should never send any data out.

Differences in legal opinions: The long delays in establishing data sharing agreements were mainly caused by differences in legal opinions from respective general counsel offices for minor issues.

Prioritization: The low priority attached to data sharing and bureaucratic obstruction are the principal data sharing challenges. No one ever prioritized the interagency project. It’s a question of priorities within each agency and who if anyone is asking agencies to make it happen. On obtaining earnings data to enforce the Gainful Employment rule, the White House essentially told the lawyers to make it happen. The issue was that providing the data to the interagency project was not in any agency’s performance evaluation and therefore may not have been their top priority. Prioritization and the bureaucracy remain obstacles during negotiations.

Interagency project’s timeframe not unusual: That the interagency project took this long to be successful may not be unusual but it’s not unusual for a set of reasons that don’t need to exist—for example, that some agencies are reluctant to share their data and need to be cajoled into doing so. The agencies really didn’t have good reasons not to participate.

Checking all the boxes: It is not in anyone’s job description to share data, but what is in job descriptions is not to share data unless all the boxes are checked from program offices down to legal counsel offices—it’s no one’s job to make data sharing happen. If data sharing involves multiple agencies, there is always at least one actor who doesn’t have an incentive to agree.

Consequences and incentives: There are no consequences for delays or bad data and no incentive to do better. Data sharing needs to be viewed as a useful product that informs the end user, not just as another project requesting agency data. Historically, federal agencies tend to be risk-averse when it comes to data sharing because they focus on program administration rather than the creation of value related to the analysis of data. Hence, there are limited incentives for agencies to engage in data sharing where their research objectives are limited or secondary. Data providers get very little out of these projects—it’s lots of work for them and no gain.

Source: These comments on data-sharing challenges from notes taken during 27 interviews have been lightly edited for clarity and are grouped together under overlapping topics.

APPENDIX D

CHRONOLOGY OF SELECTED MILESTONES OF INTERAGENCY GI BILL DATA-LINKAGE PROJECT AND EVIDENCE ACT INITIATIVES, 2016-2025

This chronology consists of milestones that occurred in a specific month as well as many that spanned a longer time period. In the case of the latter, several different individual milestones have been merged into a single entry that summarizes related activities that occurred over a year or more. Such merged milestones are found at the end of yearly sections in the table.

Date	Milestone
2016
Spring	Congress created the Commission on Evidence-Based Policymaking to develop a roadmap for using administrative and statistical data collected by agencies through data sharing to improve federal programs.
Summer and fall	Discussions to conceptualize the interagency GI Bill project commenced with agency representatives in mid-2016.
September	Letters supporting data sharing solicited and received from agency offices that administer federal student aid and GI Bill benefits at ED and VA, respectively.
December	DOD’s Defense Manpower Data Center (Manpower Center) suggested changes to a draft data-sharing agreement, which were rejected by the interagency project’s team, delaying an agreement until August 2018.
2017
February	Interagency GI Bill data-linkage project accepted as pilot for the Census Bureau’s initiative on evidence-based research, further emphasizing the interagency project’s use of the Census Bureau’s longstanding protections to safeguard privacy.
Spring	Data-sharing agreements continued to be pursued at DOD, ED, VA, IRS, and the Clearinghouse. ED legal team reviewed data-sharing agreement language and concluded that the Census Bureau might have to request an exemption from an ED statutory privacy requirement.
Spring	Access to data was complicated by not always knowing who owned the various datasets, which were not only siloed between agencies but also within agencies. For example, access to VA data was complicated by the fact that separate VA components are responsible for and must sign off on data sharing, including its Data Analytics group, Housing Administration, Education Service, Compensation Services, and Veteran Readiness and Employment Services.
August	American Institutes for Research signs contract to have staff work at the Census Bureau as special-sworn-status employees to analyze linked agency datasets.
September	Commission on Evidence-Based policymaking released a report with 22 recommendations that paralleled the data-sharing challenges encountered by the interagency GI Bill data-linkage project.
2017-2018	Given difficulties encountered in negotiating data-sharing agreements with agencies, the interagency project’s team explored workarounds—such as other agreements between the Census Bureau and agencies that could include the interagency GI Bill project as an add-on.
2017-2019	Interagency project’s team refined research objective to focus on enlisted^a veterans using the Post-9/11 Bill, forgoing analysis of other extant GI Bill programs, such as the Montgomery GI Bill, and of officers using the benefit, as well as spouses and children to whom benefits were transferred.
2018
Early 2018	The interagency team explored alternative data sources such as state unemployment data.
April	DOD’s Manpower Center and Census Bureau signed data sharing agreement.
June	Data-sharing agreement with VA’s Data Analytics group signed.
July	Agreement signed to embed the interagency project’s research team at the Census Bureau with special-sworn status and other Census Bureau privacy safeguards.
July	Veterans Benefits Administration’s Education Service component, the manager of the GI Bill, signed agreement to provide data that would flow through the Clearinghouse, which would match the VA data to college-supplied information on the outcomes of veterans using the Post-9/11 GI Bill.
2018	Negotiations were still ongoing with ED.
2018 to 2019	The interagency project’s timeline for releasing reports was delayed. In retrospect, the interagency project’s timeline was aspirational. Realistic projections for reports could not be established until the data began to arrive in early 2021.
2019
January	Evidence Act of 2018 signed into law requiring establishment of new entities charged with promoting evidence-based policymaking as well as new data-sharing expectations and processes. If fully implemented, these initiatives could reduce bureaucratic challenges and timeframes for research similar to the interagency GI Bill project. The Evidence Act also created the Advisory Committee on Data and Evidence Building to make recommendations about the establishment of a National Secure Data System.
2019	Fundamental to the interagency project was the exchange of data between VA and the Clearinghouse, where it could be matched to school enrollment and outcome data and shared with the interagency project’s team for analysis. Those negotiations continued throughout 2019.
2019	Two ancillary memoranda of understanding between the Census Bureau and VA’s Data Analytics group signed to authorize transfer of PII and to authorize acquisition of U.S. Veterans Trends and Statistics data.
2019	Ongoing negotiations with ED.
2020
January	Clearinghouse signed data-sharing agreement with the Veterans Benefits Administration’s Education Service component. VA data began arriving at Clearinghouse in batches in late spring 2020.
September	Interagency project signed data-sharing agreement with Clearinghouse allowing access to PII data.
December	Data Foundation releases first annual surveys of CDOs, a new position created by the Evidence Act of 2019. Its most recent survey covered 2024.
2020-2022	COVID results in interagency project delays throughout 2020 and into 2021 as access to data was slowed by key staff being infected, complications in moving to remote work, and delays in both meetings and negotiations.
2021
January	Interagency project’s team began receiving Clearinghouse outcome data that had been matched to records provided by VA, which required the Clearinghouse to devote significant time to cleaning the agency’s “messy” data. According to some interviewees, data cleaning dwarfed the amount of time required for analysis.
February to March	Census Bureau submits and IRS approves a Predominant Purpose Statement that allows the provision of data to improve Census Bureau surveys. IRS provided 2018, 2019, and 2020 earnings data. While 2018 and 2019 data were used in the interagency project’s reports, a decision was made not to use the 2020 data, which were likely not representative because of COVID.
March	Advisory Committee on Data for Evidence Building publishes first of two reports on the establishment of a NSDS demonstration.
May	ED proposed returning aggregate student loan debt data to the interagency project’s team if VA or the Census Bureau would provide PII data on GI Bill beneficiaries. By statute, the Census Bureau cannot transmit PII data outside the Bureau, and negotiations were essentially at a standstill. No outreach by VA occurred because aggregate data would not allow the team to match with individual-level records obtained from other federal agencies.
Summer	Interagency project’s team began linking data from participating agencies and discovered that data from DOD’s Manpower Center was not what they believed they had requested, which resulted in the drafting of an amended agreement in October 2021.
2022
February	Amendment to data-sharing agreement with DOD’s Manpower Center signed.
April	Revised dataset arrived from DOD’s Manpower Center and data cleaning commenced.
August	The Advisory Committee on Data and Evidence Building issued its second report providing a roadmap for the creation of a NSDS demonstration project to coordinate data sharing across federal agencies.
October	Draft interagency GI Bill data-linkage project’s reports sent to DOD, VA, and Census Bureau for review and approval, with anticipated release in November or December 2022.
2022	Concerned about delays, interagency project’s team decided to forgo a refresh of the Clearinghouse data in 2026.
2023
August	ED and Census Bureau signed an agreement mandated by Congress to share encrypted PII data. This agreement had been under negotiation for more than 10 years. Obtaining this agreement was a higher priority at the Census Bureau that took precedence over the interagency GI Bill data-linkage project.
Late 2023 or early 2024	VA initiates negotiations with ED and several other agencies to collect aggregate data on completion, earnings, homelessness, and other issues to help veterans choose an institution at which to use their educational benefits.
2023	Reports stuck in a longer agency review and clearance process than was anticipated.
2024
February	Call between ED Chief Economist, VA Undersecretary for Benefits, and Director of Education Service on obtaining aggregate data on federal student aid for the GI Bill Comparison Tool.
February	First interagency project’s report released, and two additional reports undergoing review and clearance.^b
February	OMB congratulates interagency GI Bill data-linkage project on first report and requests copies of data-linkage agreements to use as models for future projects.
July	Second and third interagency project’s reports are released.^b
2024	Several additional reports submitted for review and clearance are delayed for 6 months because they were not an agency priority.
2025
January	Elizabeth Dole Act of 2024 requiring outcome data to be reported on VA’s GI Bill Comparison Tool signed into law.
February	Final two reports are released by the interagency project’s team.^b

Source: Interviews, data-sharing agreements, email traffic, and notes.

^aThe decision to focus on veterans who were enlisted and on active duty as of their last recorded pay plan was driven by the reality that a prerequisite for most officers is to have a bachelor’s degree prior to joining the military. In contrast, individuals at the enlisted rank constitute the vast majority of servicemembers; most enlist without a postsecondary degree and are thus more likely to benefit from the Post-9/11 GI Bill. Data on dependents’ use of benefits was not yet available from VA. The First Look report notes that examination of both officers’ and dependents’ use of benefits could be a future project.

^bSummaries and links to these reports are included Appendix A.

APPENDIX E

TIMELINE FOR DATASETS ANALYZED
BY INTERAGENCY GI BILL DATA-LINKAGE PROJECT

Agency	Data description	Timeline
VA data sent to Clearinghouse
VA’s Veterans Benefits Administration	Enlisted veterans who used the Post-9/11 GI Bill	April 2020^a
Other VA components	Demographic data on Post-9/11 eligible veterans who did not use their benefits, disability ratings, etc.	April 2020
Data receipt by interagency project’s team
VA and Clearinghouse	Data matched to produce outcomes for enlisted veterans who used the Post-9/11 GI Bill	Late January 2021
IRS	W-2 income from tax year 2018-2020 and marital and dependents status, region, and zip code as of year of first separation	July 2021
DOD’s Defense Manpower Data Center	Academic preparedness scores for veterans upon enlistment	April 2022^b

Source: Author’s analysis of contemporaneous interagency project team’s notes, emails, and interviews.

^aTransmission of data to Clearinghouse began in batches of 300,000.

^bData from DOD’s Defense Manpower Data Center was initially received by the interagency project’s team in October 2019 but could not be examined until other agency datasets had also been received. In 2021, it was determined that the dataset was incomplete, and an amendment to the dataset was submitted in October 2021 and finally executed in February 2022.

APPENDIX F

BUREAUCRATIC CHALLENGES TO DATA SHARING IDENTIFIED
BY THE 2017 COMMISSION ON EVIDENCE-BASED POLICYMAKING

▪ “The Commission heard repeatedly about the difficulties that cumbersome and onerous procedures, often the result of idiosyncratic processes that vary across government, cause for members of the evidence-building community seeking to securely access confidential data. For researchers and evaluators external to government, no standard for applying for data access currently exists across government agencies, making it necessary to navigate different and varied processes for each agency.”

▪ “The non-standardized and lengthy processes by which access is negotiated and memorialized via an agreement were among the most common data access barriers cited by stakeholders. Formal data access agreements (e.g., Memoranda of Understanding or MOUs) between two or more agencies can take years to develop. Delays associated with negotiating MOUs are compounded by the challenges described earlier in this chapter when legal authorities are inconsistent. A lack of clear legal authority can result in extended reviews and negotiations by lawyers within multiple offices or departments prior to granting access. An agency’s ability to effectuate MOUs can also be impacted by capacity constraints.”

▪ “To access confidential data for the development of statistics, evaluation, and policy research, members of the evidence-building community today, both inside and outside the Federal government, must navigate a complex array of processes, protocols, and approaches. They must negotiate legal documents and bureaucratic processes that increase in volume and complexity when using data from multiple policy domains, jurisdictions, or agencies. Often, such processes consider the value proposition of data use only in the context of the mission of the originating agency, irrespective of its broader value. For example, generally Title 26 of the U.S. Code limits the use of tax data to those projects that would improve “tax administration.” The application of this narrow standard to research on human services or transportation, for example, may undervalue the available public good. These kinds of barriers limit the effective, efficient, and transparent use of existing data.”

▪ “These challenges are compounded when a researcher seeks to access multiple agencies’ data. In this situation, the researcher must apply to and be approved for access by all of the agencies. The result can create scenarios in which one agency provides an approval and another does not, necessitating ongoing negotiations between the agencies and the researcher.”

▪ “Moreover, inefficiencies in data access processes for evidence-building create administrative expenses and researcher burdens that can impede Federally funded research. The costs associated with excessive administrative burdens were acknowledged by the Congress and the President in a somewhat different context with the enactment of the American Innovation and Competitiveness Act in 2017, which found that “administrative costs faced by researchers may be reducing the return on investment of Federally funded research and development” and that “it is a matter of critical importance to United States competitiveness that administrative costs of Federally funded research be streamlined so that a higher portion of Federal funding is applied to direct research.”

Source: 2017 Commission report, pp. 24, 35, 36, 37 available at https://www2.census.gov/adrm/fesac/2017-12-15/Abraham-CEP-final-report.pdf.

APPENDIX G

STATUTORY AMBIGUITY AND LACK OF CONSISTENT SUPPORT
FOR EVIDENCE-BUILDING IDENTIFIED BY 2017 REPORT OF THE
COMMISSION ON EVIDENCE-BASED POLICYMAKING

Ambiguity contributes to risk aversion: “Except for that provided under CIPSEA [Confidential Information Protection and Statistical Efficiency Act], the authority to share data for evidence building is rarely explicit. In cases where authorizing laws are ambiguous, agency interpretations ultimately govern access to and use of data. In some cases, multiple agencies interpret the same law differently. This can cause confusion and limit the efficient use of existing data for evidence building. The complex web of statutes, regulations, and implementing guidance—or absence thereof—drives risk aversion in agencies, causes frustrations for the evidence-building community, and limits the value of data for statistical activities. In effect, the existing legal environment limits the government’s ability to steward data responsibly as a valuable resource for the American people and for policymakers.”

Lack of consistent statutory support for evidence building. “The Commission proposes to amend the Privacy Act and CIPSEA.…In some cases, the purposes for which administrative data may be used are defined narrowly, preventing their use for evidence building. The Commission proposes a review of such statutes to ensure that limitations that preclude the use of administrative data for evidence building are applied only when the Congress and the President deem the limitations still to be necessary. In some cases, existing laws specifically prohibit the collection or analysis of information to support evidence building. Again, the Commission calls for a reconsideration of such bans and restraint in the enactment of future bans.”

Identification of specific statutory requirements changes needed. “The Commission also cited delays caused by inconsistent legal authorities. To mitigate that risk, they called on Congress and the President to review and amend statutes as appropriate to allow statistical use of data for evidence building. Recommendations include amending statutes such as Title 13 to allow statistical uses of survey and administrative data for evidence-building within a secure CIPSEA environment;^brepealing current and limiting future bans on the collection and use of data for evidence-building; and enacting statutory changes to ensure state-level data on earnings are available for statistical purposes.”

Source: The 2017 Commission report, pages 11, 18, and 29 available at https://www2.census.gov/adrm/fesac/2017-12-15/Abraham-CEP-final-report.pdf.

^aThe Privacy Act focuses on individual privacy across all government records, while the Confidential Information Protection and Statistical Efficiency Act is focused on safeguarding sensitive statistical data collected under a confidentiality pledge. The Privacy Act provides an exemption for material required by statute that is maintained and used solely as statistical records.

^bThis recommendation was intended to resolve the disagreement between Census Bureau and IRS lawyers over whether Title 13 authorizes data sharing for the purpose of evidence-building.

APPENDIX H

KEY PROVISIONS OF THE FOUNDATIONS OF EVIDENCE-BASED POLICYMAKING ACT OF 2018

TITLE I–FEDERAL EVIDENCE-BUILDING ACTIVITIES

(Sec. 101) The bill requires agencies to submit annually to the Office of Management and Budget (OMB) and Congress a systematic plan for identifying and addressing policy questions. The plan must include, among other things

questions for developing evidence to support policymaking;
data the agency intends to collect, use, or acquire to facilitate the use of evidence in policymaking;
methods and analytical approaches that may be used to develop evidence to support policymaking; and
challenges to developing evidence to support policymaking, including any statutory or other restrictions to accessing relevant data.

Each agency shall designate a senior employee as Evaluation Officer to coordinate evidence-building activities and an official with statistical expertise to advise on statistical policy, techniques, and procedures.

The OMB shall establish an Advisory Committee on Data for Evidence Building. Agency strategic plans shall contain an assessment of the coverage, quality, methods, effectiveness, and independence of the agency’s statistics, evaluation, research, and analysis efforts. The OMB shall issue guidance for program evaluation and identify best practices for evaluation. Agencies must implement such guidance. The Office of Personnel Management shall, for program evaluation within an agency, identify key skills and competencies, establish or update an occupational series, and establish a new career path.

TITLE II–OPEN GOVERNMENT DATA ACT

Open, Public, Electronic, and Necessary Government Data Act or the OPEN Government Data Act

(Sec. 202) This bill requires public government data assets to be published as machine-readable data. The General Services Administration must maintain an online federal data catalogue to provide a single point of entry for the public to access agency data. Each agency shall develop and maintain a comprehensive data inventory and designate a Chief Data Officer. The bill establishes in the OMB a Chief Data Officer Council for establishing government-wide best practices for the use, protection, dissemination, and generation of data and for promoting data sharing agreements among agencies. The OMB shall biennially report on agency performance and compliance with this bill.

TITLE III–CONFIDENTIAL INFORMATION PROTECTION AND STATISTICAL EFFICIENCY

Confidential Information Protection and Statistical Efficiency Act of 2018

(Sec. 302) The bill codifies and revises the Confidential Information Protection and Statistical Efficiency Act of 2002. Each statistical agency or unit shall (1) produce and disseminate relevant and timely statistical information; (2) conduct credible, accurate, and objective statistical activities; and (3) protect the trust of information providers by ensuring the confidentiality and exclusive statistical use of their responses.

(Sec. 303) An agency shall make data assets available to any statistical agency or unit for purposes of developing evidence. This shall not apply to any data asset that is subject to a statute that prohibits the sharing or intended use of such asset. Each statistical agency or unit shall expand access to data assets acquired or accessed to develop evidence while protecting such assets from inappropriate access and use.

The OMB shall (1) make public all standards and policies established under this bill; (2) ensure that statistical agencies and units have the ability to make information public on the federal data catalogue; and (3) establish a process through which agencies, the Congressional Budget Office, state, local, and tribal governments, researchers, and other individuals may apply to access the data assets accessed or acquired by a statistical agency or unit for purposes of developing evidence through a standard application process.

Source: Congressional Research Service summary of the Evidence Act. The CRS summary and text of the Evidence Act are available at https://www.congress.gov/bill/115th-congress/house-bill/4174.

APPENDIX I

COMPARISON OF ANALOGOUS CHALLENGES: DATA-SHARING
WORKING GROUP AND INTERAGENCY GI BILL DATA-LINKAGE PROJECT

Data Sharing Working Group challenges

Interagency project’s challenges

Slow process to develop agreements

▪ Process from start to finish was incredibly slow. (2)

▪ Agreements were prepared and executed one at a time. (2)

▪ Everything was ad hoc and written from scratch each time. (15)

▪ Memorandums of understanding and agreement requirements varied by agency, which further complicated the issue of what rights and responsibilities agencies had to protect/audit the data. (15)

▪ The interagency project was conceptualized in 2016-2017, and the first data sharing agreements were signed in 2018. However, the final such agreement was not signed until 2021.

▪ The process of negotiating data-sharing agreements was ad hoc. There was no standardized process within or across agencies, and there were no established forms for data sharing.

▪ Getting the layers of bureaucracy at an agency on the same page, including senior leadership, administrative staff, technical staff, and lawyers was challenging. It was also challenging to be able to talk to the right person.

▪ Meetings with agency officials from whom data were sought were often difficult to schedule and by the time a meeting took place it was difficult to pick up the thread of the negotiations.

▪ The interagency project’s long timeframe also resulted in the expiration of several memorandums of agreement.

▪ To speed up the process, the interagency project’s team tried to piggy-back its data requests with ongoing Census Bureau work.

Decentralized process within agencies

▪ There was no agency-wide approach; each program handled these issues differently. (15) Need to move to agency-wide data governance (17)

▪ There was stove-piping of authorities, and with multiple authorities at play, no centralized authority existed. (15)

▪ Lack of centralization meant there was no single office approving data sharing requests. There were reasons for this, but it was a challenge. (15)

▪ Multiple agreements were required between the interagency project’s team, VA, and the National Student Clearinghouse because separate VA entities owned the data.

▪ Decentralization of data ownership was particularly evident at VA, where multiple components had to sign off on sharing their data. One VA component declined to provide a dataset sought by the interagency project’s team, forcing it to find an alternative source for the data.

No mechanism to end stalemates

▪ Need for high-ranking “champions” to facilitate communication and change management, involve the right parties in discussions, bring on knowledgeable talent, and generally provide the focus and momentum needed to complete data sharing projects. (10)

▪ Cultural obstacles (openness to data sharing) often impeded process of finalizing agreements. (2)

▪ Agencies were reluctant to share their data, citing “what’s in it for me?” (2)

▪ Senior-level officials at VA and DOD intervened to help get negotiations back on track when they stalled, but the interagency project’s team still had to deal with the same lower-level bureaucrats who did not assign the same priority to the interagency project as their superiors.

▪ Each agency had a culture concerning data sharing. Some agencies were open to linking data while others were not.

▪ The interagency project needed to demonstrate that agencies, particularly VA, would benefit from the data sharing.

▪ Lower-level agency staff to whom the negotiations were delegated did not always keep their superiors apprised of the status of negotiations.

Risk aversion concerning PII

▪ Agencies’ aversion to risk, especially concerning the interpretation of statute (Privacy Act) supporting data sharing led to a historical posture of inaction. (2)

▪ Statistical officials within agencies have experience sharing restricted data and working through data agreements and may have methodologies to assist their Data Officers. (7)

▪ Many agencies emphasized the need to work with and through their Offices of General Counsel to develop and iterate data-sharing frameworks that served their unique circumstances. (12)

▪ There were creative ways to share data while still adhering to legal or regulatory restrictions, such as the use of Joint Statistic Projects. (15)

▪ Data-sharing challenges: privacy, technology/technical requirements of sharing, ownership and data sharing agreements, discovering what could be shared (inventory/catalog), and clear policies and authority. (15)

▪ The Family Educational Rights and Privacy Act and Privacy Act lawyers at ED could not agree on overcoming PII concerns despite a precedent in which ED shared PII data with IRS for objectives similar to the interagency project’s goal of reporting veteran outcomes.

▪ A stumbling block in the VA and the Clearinghouse negotiations was whether the Clearinghouse could send matched outcome data containing PII directly to the Census Bureau or whether VA should transmit it. After months of debate, the idea of a Clearinghouse/Census Bureau memorandum of understanding was abandoned, and VA sent the matched data to the interagency project’s team.

▪ Data sharing at some agencies involved establishing Joint Statistical Projects, but one agency would share data only through a Predominant Purpose Statement. A few agreements were simply memoranda of understanding.

▪ Statutory restrictions do not allow some agencies to share data for evidence-building.

Lack of data-quality standards

▪ There was no standard application of methods for collecting or evaluating data quality. (5)

▪ There were quality issues with trusting the data as well as inconsistent internal agency approaches to design, implementation, and maintenance of data. (17)

▪ VA data were messy, and considerable time was devoted to cleaning the data by both the interagency project’s team and the National Student Clearinghouse.

▪ Data is primarily used by agencies to administer benefit programs, not for research.

No inventory of data that can be shared

▪ There was insufficient visibility of available data from each agency and what could be shared. (4)

▪ Lack of knowledge of what data were available for sharing within government. (15)

▪ Lack of centralized/federated data repository or inventory. (15)

▪ Inability to easily find what data were available (17)

▪ Many people requesting data did not know the exact data they wanted or were requesting. (17)

▪ One dataset had to be retransmitted in early 2022 because of a communication problem concerning the data needed for the interagency project.

▪ Interagency project’s researchers were not always able to speak with the technical staff at an agency who could explain the dataset or provide accurate information on what data were available.

▪ Decentralization of data ownership at VA complicated negotiations.

Challenges not identified in the Working Group report

▪ Competing organizational priorities led to delays. For example, the decennial census was the top priority at the Census Bureau.

▪ A reorganization at the Census Bureau lowered the priority attached to evidence-based pilot studies intended to demonstrate the value of data sharing for policymaking.

▪ Throughout the 8-year duration of the interagency GI Bill data-linkage project, there was staff turnover both at agencies and on its own team. Agency turnover could be particularly problematic, the equivalent of having to begin the negotiations all over again.

▪ The emergence of COVID in March 2020 initially shut down the Census Bureau’s research data center hosting the interagency project’s team. Accessing the center remotely was challenging initially, infections interrupted the work of interagency project team members, and the virus/remote work made in-person meetings impossible. Moreover, COVID resulted in the interagency project’s timelines being extended.

▪ COVID also had a tertiary effect on what IRS earning data to use. The interagency project’s team decided to report only 2018 and 2019 earnings data because the 2020 earnings might have been skewed by the pandemic and not been representative of veterans’ experiences absent the virus.

Source: Author’s analysis of Working Group report (available at https://resources.data.gov/assets/documents/2021_DSWG_Recommendations_and_Findings_508.pdf) and challenges encountered by the interagency GI Bill data-linkage project.

Note: The numbers in parenthesis correspond to the pages of the Working Group’s 2022 report that identified each challenge. The specific challenges identified in this table may overlap across the classification categories. For example, a decentralized process complicates data sharing and slows down the negotiations.

APPENDIX J

SYNOPSIS OF COMMENTS FROM INTERVIEWEES
REGARDING A DRAFT OF THIS REPORT

We gave individuals interviewed for this report the opportunity to comment on the final draft. Although most interviewees were asked to read the executive summary and conclusions, they were not asked to review the entire draft report because of concern that its length might discourage the provision of feedback. Rather, we identified specific sections of the report that focused on the expertise that each interviewee brought to this lessons-learned assessment of the interagency GI Bill data-linkages project.

Overall, we received feedback on the draft report from 15 interviewees. Eleven other individuals never responded to several requests for comments or indicated they lacked the time to provide feedback. Finally, four individuals had left their positions by the time we asked for their comments. We incorporated the technical suggestions and corrected any inaccuracies that were identified by those who provided comments; their observations that were more general in nature are summarized below.

General. As he worked his way through the draft report, one interviewee said that his disappointment in how long it took to obtain the data and complete the interagency project’s reports was replaced by amazement that anything came out at all. Two interviewees who were members of the interagency project team provided an overview of the report. One individual said that “It [the draft report] offers a thorough overview of the GI Bill data-linkage project and the Evidence Act. The report effectively addresses key elements—challenges, outcomes, and future implications—while emphasizing the importance of persistence in navigating bureaucratic and legal obstacles.”[118] A second team member said that report was “thorough and detailed—nice work.” Another individual who was a consultant to the project team, noted that it was “a chronicling of what is happening behind the scenes, that is, the little seen sausage-making process.” Finally, an Evidence Act expert commented that “the report looks great.”

Privacy Act. One interviewee believed that the Privacy Act was the main impediment to ED’s participation in the interagency project. “I think,” he commented, “that ED needs to completely realign how it engages with the Privacy Act and I think there is probably an opportunity to modernize the Privacy Act itself.”

Statistical agency roles. Several individuals commented on the data-sharing responsibility for evidence building of federal statistical agencies. One interviewee suggested that Congress could combine the dozen plus statistical agencies into a single agency similar to “Statistics Canada” in order to enhance cooperation and data sharing. He also believes that OMB’s January 2025 trust regulation could lead to conflicts between the statistical units and their parent agencies. A different interviewee emphasized that the responsibilities of federal statistical agencies spelled out in the Evidence Act may give their Statistical Officials more authority than CDOs. Another interviewee, however, pointed out that a GAO report released on September 24, 2025, highlighted data-sharing challenges faced by federal statistical agencies. In particular, a panel of experts concluded that addressing the challenge of interagency coordination required both modernizing legislation and establishing shared data infrastructure.

Importance of White House and OMB involvement in “culture change” to support data sharing. One interviewee commented that reading the draft report “validated for me the importance of very senior, sustained support from the White House and OMB if there’s ever going to be ‘culture change’ that facilitates data-sharing. In the work I’ve been doing with states, it’s extraordinarily clear that states whose governors (often with the legislature too) issued Executive Orders and set up empowered CDOs have made tremendous strides (IN, OH, KY, AR) that has been sustained when new governors come in. The states that haven’t had a signal from the governor/legislature face the same challenges you’ve described in the report.”

Evidence Act implementation. According to one interviewee, most of the National Secure Data Service demonstration work appears to be at a standstill. For example, this individual said that the creation of a dashboard to access publicly available education and workforce data for each state was halted in early 2025 because of funding and staffing cuts at the National Science Foundation where the NSDS is located. A different interviewee commented that, based on conversations with past and current agency officials, the implementation of the Evidence Act has been a disaster” and some of these individuals are already thinking about how to fix it going forward.

Veterans Education Success is a nonprofit organization whose mission is to work on a bipartisan basis to advance higher education success for veterans, service members, and military families and to protect the integrity and promise of the GI Bill® and other federal postsecondary education programs. The organization offers free help, advice, and college and career counseling to veterans using the GI Bill and helps them participate in their democracy by engaging with policymakers. Veterans Education Success also provides non-partisan policy expertise to federal and state policymakers and conducts non-partisan research on issues of concern to student veterans.

[1]A DOD webpage explaining the test is available at https://www.asvabprogram.com/media-center-article/Educators/What-is-an-AFQT-Score.

[2]“Data linkage” refers to the process of combining data from different datasets regarding the same entity or person. Because such projects entail accessing data from multiple agencies, they are more complex than those seeking data from a single agency. Throughout this report, we use the term data sharing in a broader sense that encompasses data linkages.

[3]The Foundations for Evidence-Based Policymaking Act (P.L. 115-435), available at https://www.congress.gov/115/statute/STATUTE-132/STATUTE-132-Pg5529.pdf. It was signed into law in January 2019.

[4]See Congressional Research Service, The Post-9/11 GI Bill: A Primer, at https://www.congress.gov/crs-product/R42755 for more detailed information on the most recent version of the GI Bill, which was first enacted in 1944 and has seen numerous iterations since then.

[5]See https://vetsedsuccess.org/wp-content/uploads/2019/10/VES_Fact-Sheet_GI-Bill_Payments.pdf.

[6]For example, see https://vetsedsuccess.org/wp-content/uploads/2020/04/Vets_Ed_Success_SAS_Global_Forum_Report_2020.pdf.

[7]In 2016, Congress enacted the Jeff Miller and Richard Blumenthal Veterans Health Care and Benefits Improvement Act (P.L. 114-315), available at https://www.congress.gov/114/plaws/publ315/PLAW-114publ315.htm. Section 3326 gives VA the authority to require GI Bill-participating schools to report the academic progress of beneficiaries starting in 2018. Some, but not all, schools had been voluntarily reporting graduation rates. It is unclear if the data schools must now report will capture veterans who have exhausted their GI Bill benefits because the law applies only to individuals who received a benefit payment in that year.

[8]Establishing Principles of Excellence for Educational Institutions Serving Servicemembers, Veterans, Spouses, and Other Family Members. The executive order is available at https://obamawhitehouse.archives.gov/the-press-office/2012/04/27/executive-order-establishing-principles-excellence-educational-instituti

[9]See https://nces.ed.gov/statprog/outcomemeasures/.

[10]The advisory group is known as the Technical Review Panel. It meets periodically to solicit expert input on changes to data submitted by schools for inclusion in the Integrated Postsecondary Education Data System (IPEDS) The review panel report is available at https://edsurveys.rti.org/IPEDS_TRP_DOCS/prod/documents/Report%20and%20Suggestions%20from%20TRP36_final.pdf. In commenting on a draft of this report, one interviewee noted that the Technical Review Panel is advisory to ED’s IPEDS contractor and may not be the best forum for moving forward on the reporting of veteran outcomes.

[11]Retention, graduation, average student loan debt, and federal student loan “repayment” (percentage of students not making progress on paying off their loans) are still available for some schools on the full dataset, a massive excel file with data on almost 18,000 approved schools with up to 105 columns of statistics. The full dataset, primarily a resource for researchers, is accessed through a link at the bottom of the school search page where beneficiaries can find summary information on any participating institution.

[12]Personal communication with VA staff on January 24, 2020.

[13]Miller, Ben (2016). Building a Student-Level Data System. Washington, D.C.: Center for American Progress, May. available at https://www.ihep.org/wp-content/uploads/2023/07/2016-Building-a-Student-Level-Data-System.pdf. This report was part of a larger series, Envisioning the National Postsecondary Infrastructure for the 21^st Century.

[14]A student-level system would be less time-consuming for schools participating in federal student aid. Rather than filling out labor-intensive surveys on aggregate outcomes, schools could simply upload the individual-level data they already maintain. See McCann, Clare and Amy Laitinen (2014). College Blackout: How the Higher Education Lobby Fought to Keep Students in the Dark. Washington, D.C.: New America, March, available at https://na-production.s3.amazonaws.com/documents/college-blackout.pdf.

[15]Information on the Texas and Virginia systems are available at https://nces.ed.gov/pubs97/97992.pdf. See https://education.mn.gov/MDE/dse/prev/locres/data/ for information on Minnesota’s system.

[16]The most recent iteration of the Transparency Act is available at https://www.congress.gov/bill/118th-congress/house-bill/2957/text. Information about its 2023 reintroduction is available at https://www.ihep.org/press/ihep-celebrates-the-reintroduction-of-the-bipartisan-college-transparency-act/.

[17]The 2017 bill is available at https://www.congress.gov/bill/115th-congress/house-bill/2434/text. The Transparency Act passed the by the House of Representatives in 2022. See https://www.insidehighered.com/news/government/2024/02/12/potential-breakthrough-federal-student-data-system-us.

[18]The summary is available at https://www.help.senate.gov/imo/media/doc/cta_one_pager.pdf.

[19]The blog is available at https://www.thirdway.org/blog/now-is-the-time-to-pass-the-college-transparency-act.

[20]In a 2022 statement commenting on the Transparency Act, Representative Foxx stated that “The more information the federal government has, the more they can control and that is exactly what this amendment is about—more control. From registries to lists, databases to files, bureaucrats would have unfettered access to pry into the lives of Americans.” Her statement is available at https://www.insidehighered.com/news/2022/02/07/house-passes-college-transparency-act.

[21]See § 113(1) (D) of the Act is available at https://www.congress.gov/bill/118th-congress/house-bill/6951.

[22]A 2017 report notes that the model used on the Free Application for Federal Student Aid, which asks for students or parents to consent to having IRS automatically provide tax data, could be applied to its suite of surveys. See p. 13 of Soldner, Matthew (2017). Leveraging Administrative Data to Strengthen NCES Postsecondary Sample Surveys. Washington, D.C.: National Postsecondary Education Cooperative, August, available at https://nces.ed.gov/npec/pdf/NPEC-S_Leveraging_Administrative_Data.pdf.

[23]Changes at ED and NCES during 2025 may result in changes to NCES surveys going forward.

[24]See p. 3 of a 2016 NCES report on veterans use of GI Bill benefits after enactment of the Post-9/11 GI Bill available at https://nces.ed.gov/pubs2016/2016435.pdf.

[25]In 2009, ED instituted a skip pattern in the online version of its Free Application for Federal Student Aid, which is used by the overwhelming majority of applicants. If online applicants are financially independent from their parents (as determined by criteria such as marital status and age), they never see a question about veteran status. ED’s goal was to reduce the number of questions applicants are required to answer.

[26]Bergeron, David (2016). Leveraging What We Already Know: Linking Federal Data Systems. Washington, D.C.: Center for American Progress, May. See https://www.americanprogress.org/article/linking-existing-federal-data-systems-to-expand-knowledge-of-higher-education/

[27]The Clearinghouse is a nonprofit organization established in 1993 by the higher education community as “the source for trusted, authenticated, and secure education data insights.” Information on the Clearinghouse is available at https://www.studentclearinghouse.org/about/. Participating institutions voluntarily provide enrollment and outcome data on students, which can be matched across institutions to capture a student’s engagement with higher education. It is the only source of information on students who do not receive federal student aid. Its coverage of institutions overall was about 97 percent in the fall of 2019 when the interagency project was negotiating for data on veteran outcomes. A Clearinghouse excel file of enrollment coverage through 2022 is available at https://nscresearchcenter.org/workingwithourdata/.

[28]The Armed Forces Qualification Test measures arithmetic reasoning, mathematical knowledge, paragraph comprehension, and word knowledge of incoming service members, and thus provided the interagency team with a snapshot of veterans’ academic preparedness at the time they enlisted.

[29]In general, these reports suggested that the linking of administrative data was possible and demonstrated the value of such linkages in assessing veterans’ outcomes. The researchers already had access to Army aptitude scores (but not those of the other military services) because they worked for an Army analytic organization based at the U.S. Military Academy at West Point. Pre-existing agreements with VA also provided access to data on veterans using GI Bill benefits. One of the researchers commented that obtaining Clearinghouse outcome data matched to VA’s beneficiary data took a year or two and that the IRS data were the most challenging to obtain. According to one of the researchers, the IRS provided de-identified individual level W-2 data. Because of the need to negotiate data-sharing agreements with the Clearinghouse and IRS, that report took several years to complete. It is available at https://www.nber.org/system/files/working_papers/w29024/w29024.pdf. In contrast, the other report was written and published in a few months because updated data were already in-house, including a more recent sample of veterans who had used the Post-9/11 GI Bill. That report is available at https://www.brookings.edu/wp-content/uploads/2020/10/ES-10.13.20-Kofoed-2.pdf.

[30]Cate, C.A., J.S. Lyon, J. Schmeling, and B.Y. Bogue (2017). National Veteran Education Tracker: A Report on the Academic Success of Student Veterans Using the Post-9/11 GI Bill. Student Veterans of America, Washington, D.C. The report is available at https://studentveterans.org/research/nvest. An earlier 2014 project by Student Veterans of America had already shown that VA data could be matched to Clearinghouse postsecondary outcome data. See the organizations “Millions Records Project Report” at this link: https://studentveterans.org/research/million-records-project/.

[31]Letters supporting the interagency project were received from VA’s Deputy Undersecretary for the Office of Economic Opportunity, to whom the manager of the GI Bill programs reported, and from the Chief Operating Officer at Federal Student Aid at ED.

[32]Additional details on Special Sworn Status are available at https://www.census.gov/topics/research/guidance/restricted-use-microdata/standard-application-process.html.

[33]A complete description of this system is available at https://www.census.gov/content/dam/Census/library/working-papers/2014/adrm/carra-wp-2014-01.pdf.

[34]The matched person record is assigned a unique person identifier called the protected identification key, which is an anonymous identifier. Once assigned, this key serves as a person linkage key across all files that have been processed using the Person Identification Validation System. The key also serves as a person unduplication key within files.

[35]The process is described at https://www2.census.gov/library/publications/decennial/2020/2020-census-disclosure-avoidance-handbook.pdf. A history of the disclosure review board is available at https://www.census.gov/library/working-papers/2019/adrm/CED-WP-2019-003.html.

[36]The publication of several additional reports encountered bureaucratic delays during review and clearance and were not published until February 2025.

[37]The 2017 report of Commission on Evidence-Based Policymaking, which came to a similar conclusion about the ad hoc and nonstandardized processes for negotiating data sharing projects, are discussed in Chapter VII of this report and extracts from its report are included in Appendices F and G. As noted in a Section D of this chapter and Chapter VII, the Evidence Act required federal statistical agencies to create a Standard Application Process for entities seeking access to statistical data.

[38]As one interviewee pointed out, the involvement of lawyers in negotiations is an important precondition for data sharing because it ensures that legitimate privacy concerns are to be debated and resolved.

[39]The Census Bureau and research staff assigned to the interagency project also experienced turnover.

[40]The 8^th Edition of Principles and Practices for a Federal Statistical Agency published by the National Academies of Sciences, Engineering, and Medicine provides a historical perspective on the privacy-related laws and their impact on evidence-building. It is available at https://nap.nationalacademies.org/read/27934/chapter/1. In particular, see Principle 3 (trust among the public and data subjects) and Practice 8 (respect for data subjects and data holders and protection of their data). Direct links to Principle 3 and Practice 8 are available at https://nap.nationalacademies.org/read/27934/chapter/5#34 and https://nap.nationalacademies.org/read/27934/chapter/6#71, respectively.

[41]The 2017 Commission’s report assessed the “complex web of statutes and regulations” that govern privacy. Its findings called for the review and amendment of statutes and regulations to address what it termed the ambiguous and inconsistent statutory support for evidence-building.

[42]Statutes governing the Census Bureau (Title 13) and IRS (Title 26) requirements for data sharing are addressed in Section D of this Chapter.

[43]A 2023 Congressional Research Service report (The Privacy Act of 1974: Overview and Issues for Congress, R74863, December 7, 2023) enumerates both the exemptions (Table A-2.10) and exceptions (Table A-3.12) and is available at https://www.congress.gov/crs-product/R47863.

[44]The Department’s overview of the Privacy Act is available at https://www.justice.gov/opcl/overview-privacy-act-1974-2020-edition/disclosures-third-parties.

[45]An ED description of exemptions is available at https://studentprivacy.ed.gov/faq/what-records-are-exempted-ferpa. Examples of exempted records include records of a law enforcement unit of an educational agency or institution that were created for law enforcement purposes; records relating to an employee of an educational agency or institution that relate exclusively to their capacity as an employee (and not as a student); and health records made and maintained by a health professional solely for treatment purposes.

[46]An ED summary of exceptions is available at https://studentprivacy.ed.gov/sites/default/files/resource_document/file/FERPA%20Exceptions_HANDOUT_horizontal_0_0.pdf. Examples of FERPA exceptions to the consent rule include officials of other schools where the student is seeking to enroll; state and local educational authorities, or authorized representatives for audit or evaluation of federal/state-supported education programs; and in response to a judicial order or lawfully issued subpoena.

[47]Soldner, Matthew (2017). Leveraging Administrative Data to Strengthen NCES Postsecondary Sample Surveys. Washington, D.C.: National Postsecondary Education Cooperative, August, available at https://nces.ed.gov/npec/pdf/NPEC-S_Leveraging_Administrative_Data.pdf. Soldner notes that IRS can share data only when it is directly related to the administration of specific programs. See p. 19.

[48]A more detailed description of data cleaning is available at https://www.tableau.com/learn/articles/what-is-data-cleaning#definition.

[49]As noted in Section D of this chapter, VA’s Data Governance and Analytics Group did not own (manage) or create the data sets it provided to the Clearinghouse on demographics, disability ratings, or eligible veterans who had not used their benefits.

[50]Appendix E provides a timeline for datasets analyzed by the interagency project’s team, which began arriving piecemeal in 2021. As discussed in Section D of this Chapter, data from DOD’s Defense Manpower Data Center was first received by the interagency project in October 2019 but was not the data requested. The correct data were retransmitted in April 2022.

[51]As noted earlier, all analytical reports using PII undergo a review by the Census Bureau’s Disclosure Review Board before being released to the public to ensure that the data cannot be used to reveal individual identities.

[52]Ultimately, as discussed later in this report, the Data Service was positioned within the National Science Foundation.

[53]A November 2016 presentation noted that OMB had directed the Census Bureau to use appropriated funds to “administer the Evidence-Based Policy Commission, including pilots to support evidence-building and program evaluation.” The presentation is available at https://bipartisanpolicy.org/download/?file=/wp-content/uploads/2019/03/HUD-Data-and-Capacity-ORegan.pdf.

[54]Data owned by other VA components feed into the annual Veterans Benefits Report, a compendium of veterans’ entitlement programs, which provide aggregate data on those programs.

[55]The long-term relationship of VA’s data analytics group with the Census Bureau made sharing the VA component’s data with the Census Bureau easier. Data Analytics updates the Census Bureau with U.S. Veterans Trends and Statistics at least once a year and when specific data elements are needed.

[56]See https://www.va.gov/opa/docs/remediation-required/oei/VA_Data_Strategy.pdf for a description of VA’s data enterprise strategy and the role of the council.

[57]According to VA, however, the Education Service data on veterans who used the Post-9/11 GI Bill was sent to the Clearinghouse by the Data Analytics group.

[58]Identifying educational institutions was also a stumbling block for a separate project linking Clearinghouse data with Census Bureau data from its Longitudinal Employer-Household microdata set to produce postgraduate student earnings by institution, major, and degree. Instead, the Census Bureau obtained data directly from states and university systems, making the data public on the Census Bureau’s Post-Secondary Outcomes Explorer. The results of this project are available at https://lehd.ces.census.gov/applications/pseo/?type=earnings&compare=postgrad&specificity=2&state=08&institution=08&degreelevel=05&gradcohort=0000-3&filter=50&program=52,45.

[59]Although one of the interagency project’s reports focused on outcomes at state flagship universities, the data were reported in the aggregate, not by institution, an approach that does not require the Clearinghouse’s written approval.

[60]In commenting on a draft of this report, one interviewee added context by explaining that “IRS data were already at the Census Bureau for statistical purposes; they didn’t need to share a file to do the project, they only had to review and approve a proposal to use what was there. (It’s a different level of burden than VA or the Defense Manpower Data Center where they had to prep extracts and do the paperwork).”

[61]Title 13 is available at https://www.govinfo.gov/content/pkg/USCODE-2007-title13/pdf/USCODE-2007-title13.pdf. A Census Bureau publication describing the legal authority and policies for data linkage is available at https://www.census.gov/about/adrm/linkage/about/authority.html.

[62]For example, the data-sharing agreements with two VA components were Joint Statistical Projects, which were backed up with two ancillary memoranda of understanding.

[63]The 2017 Commission on Evidence-Based Policymaking recommended that the Confidential Information Protection and Statistical Efficiency Act be amended to clarify that Title 13 authorizes data sharing for the purpose of evidence-building. While the Evidence Act of 2019 reauthorized and amended the Confidential Information Act, it did not address the Census Bureau’s authority under Title 13.

[64]Title 26 is available at https://www.ecfr.gov/current/title-26/chapter-I/subchapter-F/part-301/subpart-ECFR1b5d05d4bfe19f9/subject-group-ECFR2bb42ef5f1a3a92/section-301.6103(j)(1)-1.

[65]It is worth emphasizing that the Transparency Act, first introduced in 2017 but never enacted, would establish a student unit record system that captures “military or veteran benefit status,” obviating the need for VA to identify veterans for ED.

[66]ED provided micro-level data to IRS on students, including their Social Security Numbers, the award year, and school code, and IRS linked that data to earning data that was returned to ED as medians (i.e., no PII). IRS used privacy techniques to render the medians noisy (perturbed), making it more difficult to identify individuals. If a data-sharing agreement had been signed with the interagency GI Bill project, ED would have similarly provided micro-level data to the interagency project, which, like IRS, would have produced aggregate data that protected students’ confidentiality.

[67]During the conceptualization of the interagency project, the team articulated a goal of creating “infrastructure” to facilitate future data-linkage projects. In general, interviewees had overlapping but not identical definitions of infrastructure, an issue that is also discussed in Chapter VII, Section D, of this report. For example, two interviewees defined infrastructure as tools that include websites and templates as well as centralized secure platforms not owned by any agencies that agree to share data. A different interviewee defined infrastructure not as a tool that facilitates data sharing but as a roadmap so that someone could step in and replicate a project’s success.

[68]ED and VA signed a data-sharing agreement in 2016 to provide aggregate data on veterans’ student loan debt on the GI Bill Comparison Tool. An interviewee said that the agreement lapsed because VA never supplied the necessary PII to ED to effectuate the agreement.

[69]The agencies contacted included ED (federal student aid), the Clearinghouse (outcomes for a sample of veterans), and the Departments of Housing and Urban Development (housing and homelessness), Labor (career and training services), and Treasury (earnings). However, the interviewee indicated that the precedents established by the interagency project did not help to jumpstart negotiations; rather, the process was “like starting all over again,” which was attributed to a lack of familiarity with that interagency project because of staff turnover. At some agencies, VA reportedly knocked on the wrong door and its negotiators were redirected to a different office within the agency.

[70]According to one interviewee who is a data sharing expert, reliance on aggregate data as an outcome from data sharing is an example of problem solving and rethinking what data is actually needed because there is no issue with releasing aggregate data.

[71]The Elizabeth Dole Act is available at https://www.congress.gov/118/bills/s141/BILLS-118s141es.pdf.

[72]Testimonies of Mark Schneider, American Institutes for Research, and Carrie Wofford, Veterans Education Success, before a Hearing of the Congressional Bipartisan Commission on Evidence-Based Policymaking on October 21, 2016, available at: https://www.youtube.com/watch?t=4514&v=Uxu4Fj4qT5E&feature=youtu.be. A transcript of Carrie Wofford’s testimony is available at https://vetsedsuccess.org/wp-content/uploads/2019/01/gi-bill-data-commission-evidence-based-policy-making-ves-testimony.pdf.

[73]The Federal Data Strategy’s 2020 Action Plan identifies 20 agency actions that must be implemented, which attests to the scope and complexity of the Evidence Act’s initiatives. The plan is available at https://strategy.data.gov/action-plan/.

[74]The Data Foundation is a Washington, D.C.-based nonprofit organization. For more than a decade, the Data Foundation has been bringing attention to key issues facing the data community. More information on the Data Foundation is available at https://datafoundation.org/about.

[75]The Evidence-Based Policymaking Commission Act of 2016 (P.L. 114-140) is available at https://www.congress.gov/bill/114th-congress/house-bill/1831.

[76]The Commission’s report is available at https://www2.census.gov/adrm/fesac/2017-12-15/Abraham-CEP-final-report.pdf.

[77]In addition, the National Academies of Sciences, Engineering, and Medicine published three peer-reviewed reports written by an expert panel on the establishment of a 21^st century national data infrastructure. Its first report, released in 2023 also noted that “… many laws and regulations do prohibit federal statistical agencies from using existing data for statistical purposes” and concluded that “Legal and regulatory changes are necessary to achieve the full promise of a 21st century national data infrastructure.” (Conclusion 3-3) National Academies of Sciences, Engineering, and Medicine, Toward a 21st Century National Data Infrastructure: Mobilizing Information for the Common Good. Washington, DC: The National Academies Press, 2023. The report is available at https://doi.org/10.17226/26688.

[78]The Evidence Act (P.L. 115-435) is available at https://www.congress.gov/bill/115th-congress/house-bill/4174.

[79]The Advisory Committee published reports in 2021 and 2022 laying out a roadmap for the creation of a Data Service; both reports are available at https://www.bea.gov/evidence. Section 10375 of the 2022 CHIPS and Science Act (P.L. 117-167) created the Data Service as a 5-year demonstration project and is available at https://www.congress.gov/bill/117th-congress/house-bill/4346.

[80]The Working Group report is available at https://resources.data.gov/assets/documents/2021_DSWG_Recommendations_and_Findings_508.pdf.

[81]The Working Group’s charter, which was approved 2 months later is available at https://www.cdo.gov/assets/documents/data-sharing-wg.pdf.

[82]The website can be accessed at https://data.gov

[83]The Chief Data Officer’s Council published a playbook in 2021 which it characterized as an iterative document that explores and defines the evolution of the federal Data Officer’s role. The playbook focused on the Evidence Act activities of Data Officers at several federal agencies. It is available at https://resources.data.gov/assets/documents/CDO_Playbook_2021.pdf. In January 2024. The Council also published a review of the Council’s accomplishments over the past several year, which is available at https://www.cdo.gov/year-in-review-2023/.

[84]The 2021to 2024 Data Foundation surveys are available at https://www.cdo.gov/year-in-review-2023/ and the 2020 survey is available at https://universoabierto.org/wp-content/uploads/2020/08/50764-effetcive_data_governance_a_survey_of_federal_chief_data_offcer_fina_2020.pdf. It should be noted that each survey represents a snapshot of the CDOs’ roles at a point in time and that the survey response rates appear to be low. Differences in responses from year to year may be attributable to variation in who responded to the survey. Moreover, the evolution of the CDO’s role is difficult to capture because of disparities in the maturation of the role across dozens of federal agencies.

[85]In commenting on a draft of this report, one interviewee stated that he found the comments by CDOs “far too predictable and self-serving,” without updating their expectations about future efforts for data sharing.

[86]The Evidence Act directs agency CFOs to designate a CDO, Evaluation Officer, and Statistical Official, potentially giving them the authority to require these new positions to report to the Financial Officer. The CFO is held accountable for ensuring that these designated roles are properly filled and that the agency is effectively utilizing data to inform policy decisions.

[87]The surveys do not contain any detailed statistics on the placement of CDOs’ at federal agencies. Similarly, a 2022 GAO report examined CDOs’ placement at a nongeneralizable sample at six large federal agencies (available at https://www.gao.gov/assets/gao-23-105514.pdf). The Evidence Act does not specify the placement of CDOs and, according to GAO, OMB guidance only specifies that this position should be high enough to regularly engage with other agency leadership.

[88]The OMB guidance is available at https://bidenwhitehouse.archives.gov/wp-content/uploads/2025/01/M-25-05-Phase-2-Implementation-of-the-Foundations-for-Evidence-Based-Policymaking-Act-of-2018-Open-Government-Data-Access-and-Management-Guidance.pdf.

[89]The presumption, even when there is uncertainty over the results of an assessment that balances the risks of releasing data versus the public interest in access, is that the default should be to release the data. However, if the data cannot be disclosed in full, agencies must explore alternative means of disclosure to appropriate users.

[90]The blog is available at https://datafoundation.org/news/blogs/560/560-Open-by-Default-OMBs-New-Data-Management-Guidance-.

[91]The statement by the President and CEO of the Data Foundation is available at https://datafoundation.org/news/press-releases-and-statements/571/571-Data-Foundation-Statement-on-the-Evolving-Federal-Data-and-Evidence-Ecosystem-in-.

[92]The reports are available at https://datafoundation.org/news/evidence-capacity/635/635-Evidence-Capacity-Pulse-Report-June-30-.

[93]The Data Service was inspired by the pilots launched in 2016 and 2017 by the Center for Administrative Records Research and Applications at the Census Bureau, and Census Bureau employees were sent to help start the Data Service demonstration.

[94]See opening letter to OMB in the 2022 final report of the Advisory Committee on Data and Evidence Building available at https://www.bea.gov/sites/default/files/2022-10/acdeb-year-2-report.pdf.

[95]The September 2024 report is available at https://ncses.nsf.gov/initiatives/national-secure-data-service-demo/two-year-congressional-report. A December 2024 data service press release available at https://www.nsf.gov/news/national-secure-data-service-demonstration-project-builds announced 10 new data service demonstration projects, primarily with universities and nonprofit organizations.

[96]As noted earlier, the Census Bureau also anonymizes PII data before providing it to researchers.

[97]These templates are intended for federal agency data-linkage projects. In contrast, the standard application process developed by federal statistical agencies is for individuals or state and local government that want to conduct research using federal databases. This application portal was launched in February 2023 and must be used by such entities to apply for access to federal statistical databases see https://www.census.gov/newsroom/press-releases/2023/standard-application-process.html). According to one interviewee, researchers seeking access to IRS data must now use this portal.

[98]The Commission’s report is available at https://www2.census.gov/adrm/fesac/2017-12-15/Abraham-CEP-final-report.pdf.

[99]The recommendation on amending the Confidential Information Act was intended to resolve the disagreement between the Census Bureau and IRS lawyers over whether Title 13 authorizes data sharing for the purpose of evidence-building.

[100]In commenting on a draft of this report, one interviewee noted that a new GAO report regarding the federal statistical system highlights challenges and opportunities, including public trust, data access and support, alternative data sources, and interagency coordination. Suggestions for strengthening interagency coordination included modernizing legislation and establishing shared data infrastructure. The report released on September 24, 2025, is available at https://files.gao.gov/reports/GAO-25-107124/index.html?_gl=1*djolem*_ga*MTgzOTM0Mjk1Ny4xNzUxMzk5NTU5*_ga_V393SNS3SR*czE3NTg5ODM0NzAkbzQkZzEkdDE3NTg5ODUxODUkajYwJGwwJGgw.

[101]The Evidence Act includes provisions requiring GAO to review different aspects of its implementation at several points in time and whether certain required activities improved the use of evidence and program evaluation in the federal government. In addition, GAO is required to report on findings and trends in agencies’ capacity assessments and, if appropriate, recommend actions to further improve agency capacity.

[102]In general, GAO’s body of work examined specific Evidence Act milestones and processes using surveys, nongeneralizable samples, record reviews, and interviews to capture progress and identify challenges.

[103]GAO-21-536 and GAO-24-106982 are available at https://www.gao.gov/assets/d21536.pdf and https://www.gao.gov/assets/gao-24-106982.pdf, respectively. The issue summary is available at https://www.gao.gov/improving-federal-programs-through-data-and-evidence.

[104]An example of this dispersion is the reference in one report to the role of the Interagency Evaluation Officer Council in helping to address this problem. An additional coordinating body, the Interagency Council on Evaluation Policy, predated the Evidence Act. According to GAO, OMB maintains that this older Interagency Council is a working group under the Evaluation Officer Council.

[105]GAO-21-152. This report is available at https://www.cdo.gov/assets/documents/GAO_Report_on_Data_Governance.pdf.

[106]GAO-21-536 and GAO-24-106982 are available at https://www.gao.gov/assets/d21536.pdf and https://www.gao.gov/assets/gao-24-106982.pdf, respectively.

[107]GAO-22-103910. The report is available at https://www.gao.gov/assets/gao-22-103910.pdf.

[108]This interviewee also noted that the Data Officers Council helps to promote the idea of stewardship versus agencies owning the data because Data Officers at different agencies have the opportunity to talk to each other and to make a value judgment about the benefits of data sharing.

[109]The Data Foundation’s blog on OMB’s January 2025 Evidence Act guidance stated that it provides clarity on Data Officers’ roles and responsibilities but questioned whether Data Officers have adequate resources, capacity, and support from agency leadership to fully implement the guidance. The blog is available at https://datafoundation.org/news/blogs/560/560-Open-by-Default-OMBs-New-Data-Management-Guidance-.

[110]See Appendix I, p. 21 of GAO’s report at https://www.gao.gov/assets/gao-23-105514.pdf.

[111]See https://www.govconwire.com/article/lisa-rosenmerkel-sworn-in-as-va-chief-data-officer.

[112]The placement of the DOD Data Officer and the office’s responsibilities have changed over time. The National Defense Authorization Act of 2020 specified that the Data Officer would report to the department’s Information Officer (see https://media.defense.gov/2022/Feb/02/2002931807/-1/-1/1/MEMORANDUM-ON-THE-INITIAL-OPERATING-CAPABILITY-OF-THE-CHIEF-DIGITAL-AND-ARTIFICIAL-INTELLIGENCE-OFFICER.PDF). DOD merged several data functions in 2021 and announced that the new Chief Digital and Artificial Intelligence Officer would report to the Secretary of Defense and Deputy Secretary of Defense through DOD’s Information Officer. The merger included four offices: the Data Officer, artificial intelligence, digital services, and advanced analytics. As of November 2024, the CDAO is now the principal staff assistant and advisor to the Secretary of Defense (see https://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodd/510589p.PDF?ver=Ikhn-60VR-GpxO78wiYQZA%3D%3D).

[113]The presentation is available at https://www.bea.gov/sites/default/files/2021-02/OCDO-ACDEBpresentation-FEB2021.pdf.

[114] While there are success stories about the implementation of Evidence Act initiatives, they tend to be discussed in individual CDO reports, making them more difficult to track given the number of new positions created by the legislation. Moreover, it is challenging to determine whether examples of progress are evidence of sustained progress or isolated examples.

[115]The Trust Regulation is available at https://www.federalregister.gov/documents/2024/10/11/2024-23536/fundamental-responsibilities-of-recognized-statistical-agencies-and-units#:~:text=To%20promote%20public%20trust%20in,statistical%20use%20of%20their%20responses. OMB’s regulations for the two other Title III requirements—presumption of access and facilitating access—are still being drafted. According to one interviewee, OMB wanted to publish the Trust Regulation before releasing the other two regulations.

[116]See https://www.aei.org/education/dont-trust-the-office-of-management-and-budgets-new-trust-regulation/ and https://nces.ed.gov/whatsnew/commissioner/remarks2024/11_01_2024.asp.

[117]Despite a guarantee of anonymity, our ability to gain insights on the implementation status of the Evidence Act initiatives was limited by the reluctance of some experts and agency officials and experts to be interviewed.

[118]This individual found the report illuminating because he was not part of the interagency project over its more than 8-year duration.

Lessons Learned from GI Bill Interagency Data Project.FINAL2

Lessons Learned from the Interagency GI Bill Data-Linkage Project

Share This

Veterans Education Success