Whether you're studying learning outcomes, analyzing differences in opportunity and outcomes across student groups, or investigating the effectiveness of digital learning tools, access to quality data determines the type of questions your research can answer. The good news: more education datasets are publicly available than ever before. The challenge: knowing where to look and understanding what each source can actually deliver.
Most sources also make a trade you should see coming. The more granular, recent, and linkable the data, the harder it is to get: aggregate indicators are a download away, while student-level records that follow learners over time sit behind months of agreements, if they can be had at all.
Depending on your needs, the best places to find education datasets include:
Resources for privacy-preserving research infrastructure:
SafeInsights
ADRF/Coleridge Initiative
Resources for national and federal data:
National Center for Education Statistics (NCES)
Data.gov
Resources for regional and institutional data:
State and institutional open data portals
Resources for international and cross-national comparisons:
UNESCO Institute for Statistics
World Bank Education Statistics
OECD Education Data
Resources for machine learning and data science projects:
Kaggle
Resources for curated research datasets:
ICPSR (Inter-university Consortium for Political and Social Research)
Stanford Educational Opportunity Project
Urban Institute Education Data Portal
Resources for learning analytics and interaction data:
DataShop / LearnSphere
So, what is the best source of education data? It depends on the question. For aggregate trends, the federal and international sources below are authoritative and free. If your question needs student-level, longitudinal data from the educational apps students actually use, SafeInsights was built for that case. Your analysis code goes to the data, keeping individual student records private while returning large-scale insights, and the months usually spent negotiating data use agreements are streamlined into one review process.
Resources for privacy-preserving research infrastructure
SafeInsights
SafeInsights takes a different approach to education research infrastructure, often called compute-to-data. Rather than downloading static datasets, researchers bring their questions directly to where data already lives: in the secure environments of edtech platforms, educational institutions, and state and local education agencies.
Key strengths:
Researchers never see row-level student data, which can simplify review by your Institutional Review Board (IRB)
Delivers both speed and scientific rigor for PhD students, established researchers, and institutional research teams
Your insights can inform edtech development and pedagogical best practices towards better outcomes for all students
Limitations:
Researchers work with code and aggregate outputs; hands-on exploration happens against sample or synthetic data
Every study needs approval from the organization that holds the data, in addition to IRB review
Available data depend on which data organizations have joined
Best for: Large-scale longitudinal studies on student learning that require analyzing student-level data without revealing personally identifiable information.
ADRF/Coleridge Initiative
The Administrative Data Research Facility (ADRF) (Coleridge Initiative, n.d.), built by the nonprofit Coleridge Initiative, enables government agencies to securely link and share confidential administrative data across state lines and agencies in a federally authorized cloud environment.
Key strengths:
Enables cross-state and cross-agency data linking
Provides secure access to longitudinal education and workforce data
Applied Data Analytics training builds agency capacity to conduct rigorous analyses
Facilitates multi-state collaborative research
Approved data stewards maintain control over access and usage of their data
Limitations:
Access limited to approved researchers working on approved projects
Requires data governance agreements and coordination across multiple agencies
Availability depends on which states and agencies have uploaded data to the platform
Focus primarily on education and workforce administrative data
Researchers work within the secure cloud environment rather than downloading data
Best for: Cross-state longitudinal studies tracking students and workers across education systems, workforce programs, and state boundaries using secure administrative data.
Resources for national and federal data
National Center for Education Statistics (NCES)
The National Center for Education Statistics serves as the primary federal source for education data in the United States, offering comprehensive datasets spanning early childhood through postsecondary education.
Key strengths:
Methodologically rigorous with nationally representative samples
Major surveys include the National Assessment of Educational Progress (NAEP), the Common Core of Data (CCD), and the Integrated Postsecondary Education Data System (IPEDS)
Longitudinal studies tracking cohorts over years
Freely accessible with excellent documentation
Widely cited and authoritative
Limitations:
Release lags of roughly one to two years after collection
Focus on aggregate patterns rather than granular learning behaviors
Not designed for within-year learning-process or platform-specific research
Best for: Macro-level trends, policy impact studies, and national comparisons requiring authoritative data.
Data.gov
Data.gov aggregates thousands of datasets about education from federal, state, and local government agencies into a single searchable portal.
Key strengths:
Broad coverage of the U.S. education system
Includes school report cards, district finances, college scorecards, and more
Good for geographically specific studies
Free and publicly accessible
Limitations:
Inconsistent data quality, formats, and documentation across sources
Variable update frequencies
Requires significant time evaluating and reconciling datasets
Lacks standardization for systematic large-scale analysis
Best for: Exploratory research, local policy studies, or projects requiring specific geographic data.
Data specificity for particular geographic contexts and state policy questions
More frequent updates than national datasets
Often more granular detail than federal sources, including student-level longitudinal tracking
Integration across education sectors (PreK-12, postsecondary, workforce) within states
Dedicated research centers in some states, such as the Texas Education Research Center, the Michigan Education Data Center, Washington’s Education Research and Data Center, and the North Carolina Education Research Data Center
Publicly available dashboards and reports for educators, policymakers, and families
All 50 states, the District of Columbia, and five territories have received at least one federal SLDS grant since 2005 (National Center for Education Statistics, n.d.)
Limitations:
Dramatic variation in quality, format, and documentation across state portals
Significant data harmonization challenges for multi-state research
Access to student-level data typically requires formal applications and approvals
Privacy protections appropriately limit individual record access to authorized researchers
Not all states have fully functioning systems integrating all education sectors
Cross-state data linkages remain challenging without platforms like ADRF
State practitioners themselves note that SLDS data will not answer many of their most pressing research questions, especially about instruction and learning processes (Conaway et al., 2015)
Best for: State-specific policy studies, district-level interventions, institutional effectiveness research, and longitudinal tracking within a single state's education system.
Resources for international and cross-national comparisons
UNESCO Institute for Statistics
UNESCO Institute for Statistics (UIS) is the official source of internationally comparable education data for more than 200 countries and territories and the custodian agency for the education indicators of Sustainable Development Goal 4 (SDG 4).
Tracks enrollment, literacy, expenditure, and learning outcomes globally
Historical time series for longitudinal international research
Monitors progress toward SDG 4
Limitations:
Data collection varies by country
Updates depend on national statistical capacity
Focus on system-level indicators, not student-level data
Gaps exist where countries lack collection infrastructure
Best for: Cross-national comparisons, global education trends, and international development research.
World Bank Education Statistics
World Bank Education Statistics (EdStats) (World Bank, n.d.) provides comprehensive data on education systems worldwide, emphasizing developing countries and education's role in economic development. The EdStats portal is being revamped; the full set of roughly 2,500 indicators remains available through the World Bank DataBank.
Key strengths:
Policy-relevant indicators including enrollment, completion, expenditure
Integration with broader development indicators
Companion datasets on Learning Poverty and Harmonized Learning Outcomes that put national assessments on a common scale
Case studies supplement quantitative data
Strong focus on education finance and economic outcomes
Limitations:
Timeliness and consistency challenges across countries
Coverage gaps in nations with limited statistical capacity
Emphasizes macro-level system performance over learning processes
Some portal indicators are not current during the revamp; check DataBank for the latest release
Best for: Research at the intersection of education and development economics, policy analysis in developing contexts.
PISA assesses 15-year-olds in reading, math, and science on a three-year cycle
Rigorous methodology with excellent documentation
Also includes data on teacher quality, expenditure, and adult skills (Programme for the International Assessment of Adult Competencies, PIAAC)
Widely treated as the reference point for cross-national comparisons
Limitations:
Coverage skews toward higher-income systems
Three-year assessment cycles create temporal gaps
Emphasis on standardized test performance, not broader outcomes
Best for: International comparisons of education systems, studying policy effectiveness across countries.
Resources for machine learning and data science projects
Kaggle
For education datasets, Kaggle collections can be a powerful source. Kaggle attracts data science researchers and machine learning practitioners with its curated library of datasets, competitions, and shared notebooks.
Key strengths:
Curated datasets prepared for machine learning applications
Active community discussions and baseline models
Clear documentation and community-contributed analyses
Often small convenience samples, not representative populations
Many datasets lack rigorous methodology documentation
Unclear provenance or consent procedures for user-uploaded datasets
Rarely support longitudinal analysis
For publishable work, go to the original source for documentation, versioning, and terms of use
Best for: Prototyping predictive models, teaching data science, or algorithm development; verify the provenance before using any dataset for publishable research.
Resources for curated research datasets
ICPSR (Inter-university Consortium for Political and Social Research)
Comprehensive archive with rigorous data curation and preservation standards
Specialized thematic collections
Includes major longitudinal studies and federally funded education research
Excellent documentation with codebooks, technical reports, and variable-level metadata
Data formatted for statistical software (SPSS, SAS, Stata, R)
Summer training program in quantitative methods to support effective data use
Publicly accessible datasets alongside restricted-use data for approved researchers
International data available, though most datasets focus on U.S. studies
Limitations:
Requires institutional membership for full access (though many datasets are publicly available)
Primarily archival data rather than current information
Focus on completed studies rather than ongoing data collection
Variable quality and coverage across different time periods and topics
Most useful for secondary data analysis rather than primary data collection
Best for: Secondary analysis of existing education research studies, historical trend analysis, and researchers at member institutions seeking curated, well-documented datasets for quantitative analysis.
Comprehensive district and county-level achievement data standardized across states
Interactive Opportunity Explorer tool for creating custom visualizations and reports
Achievement gap measures by race/ethnicity, gender, and socioeconomic status
Longitudinal tracking of educational opportunity trends over time
Data on school and neighborhood segregation patterns
Pandemic-era achievement change and recovery tracking through the Education Recovery Scorecard, a joint effort with Harvard’s Center for Education Policy Research
Free public access to data files in multiple formats (Stata, CSV)
Extensive documentation and research publications using the archive
Limitations:
Limited to grades 3-8 standardized test data (not PreK-2 or high school)
Focus on achievement outcomes rather than instructional practices or inputs
Aggregated to districts, counties, and, in the main archive, schools; no individual student records
Test score focus may not capture full spectrum of educational quality
Updates lag by 1-2 years as state test data are compiled and standardized
Best for: Research on achievement gaps, differences in educational opportunity, district-level performance comparisons, and studies examining how community factors influence academic outcomes across the United States.
Urban Institute Education Data Portal
The Urban Institute's Education Data Portal harmonizes data from major national sources on schools, districts, and colleges into a unified, accessible platform with a powerful API.
Key strengths:
Aggregates multiple federal datasets into a single searchable platform
Harmonized variables and documentation across years and sources
API access through the web browser, R (Tyagi et al., 2025), and Stata packages
Covers K-12 schools, districts, and postsecondary institutions from the 1980s to present
Interactive Education Data Explorer for custom data downloads
Free and openly licensed under the Open Data Commons Attribution License
Includes demographic, financial, enrollment, and outcome measures
Regular updates as new federal data releases become available
Limitations:
Limited to data already collected by federal agencies (not new data collection)
Primarily institutional-level data rather than student-level records
API and technical documentation require programming knowledge for full utilization
Does not include state-specific assessments or non-federal data sources
Coverage varies by topic and time period depending on source dataset
Best for: Researchers, policymakers, and analysts needing programmatic access to harmonized federal education data across schools, districts, and colleges for trend analysis and cross-dataset comparisons.
Resources for learning analytics and interaction data
DataShop / LearnSphere
DataShop (based at Carnegie Mellon University; Koedinger et al., 2010) and LearnSphere (Stamper et al., 2024) provide what DataShop describes as the world's largest open repository of fine-grained educational data from interactive learning systems.
Key strengths:
Largest open repository of transactional educational data (student-software interactions)
Fine-grained, longitudinal data tracking individual student performance and learning curves
Built-in analysis and visualization tools accessible through a web-based interface
Data from interactive tutoring systems, massive open online courses (MOOCs), educational games, and collaborative learning platforms
Tigris workflow authoring tool allows custom analyses without additional programming
Integrates data from DataShop, MOOCdb, DiscourseDB, and DataStage repositories
Supports research reproducibility by sharing data, analytic methods, and links to publications
Limitations:
Focus on data from computerized learning systems rather than traditional classroom instruction
Most datasets from heavily interactive systems (tutoring systems, games), not passive learning contexts
Primarily captures student-software interactions rather than student-student or student-teacher interactions
Some datasets are restricted and require permission from the dataset owner
Requires familiarity with learning analytics and educational data mining methods
Data format and analysis tools have a learning curve for new users
Coverage varies widely across subject domains and grade levels
Best for: Learning science researchers studying student interactions with educational technology, fine-grained analysis of learning processes, knowledge component optimization, and researchers developing or validating learning analytics methods.
How to choose the right dataset
Choosing the right education dataset starts with your research questions, not the data itself. The most common mistake researchers make is selecting a dataset based on availability or convenience rather than alignment with their actual research needs. Instead, consider:
Research scope and scale: Does your study require a nationally representative sample, specific geographic populations, or access to students in different contexts? Small convenience samples may work for exploratory studies but limit generalizability.
Temporal requirements: Are you investigating a single moment in time (cross-sectional) or tracking changes over weeks, months, or years (longitudinal)? Most static datasets capture snapshots, making longitudinal research challenging.
Data granularity: Do you need aggregate institutional statistics, classroom-level patterns, or individual student behaviors? System-level data answers different questions than student-level data.
Platform and context specificity: Are you studying learning within a particular edtech platform, across multiple platforms, or in traditional classroom settings? Many datasets are limited to single platforms or offline contexts.
Access timeline: How quickly do you need to begin analysis? Negotiating access to restricted data routinely takes many months per source (Mello et al., 2020), while some public repositories offer immediate downloads.
Privacy and ethical requirements: Does your IRB require specific privacy protections? Will you need to combine data across sources that have different consent frameworks?
Documentation and support: Is the dataset well-documented with clear methodology, variable definitions, and technical support? Poor documentation leads to analytical errors and irreproducible research.
Cost considerations: Beyond direct access fees, consider the hidden costs: staff time for data cleaning, legal reviews for data use agreements, and opportunity costs of delayed research timelines.
Researchers increasingly recognize that infrastructure purpose-built for education research can eliminate many traditional tradeoffs between scale, privacy, access speed, and data quality. Rather than compromising on sample size because privacy negotiations failed, or limiting your questions to what static datasets can answer, modern research infrastructure can enable you to design studies around your questions first and access appropriate data second.
Choosing an education dataset — FAQ
Where can I find datasets for research?
Education datasets are available through multiple sources. Government agencies like NCES and Data.gov provide free access to national and state-level data, while international organizations like UNESCO and OECD offer cross-national comparisons. For large-scale, longitudinal studies on student learning across contexts and over time, research infrastructure like SafeInsights enables studies that traditional repositories cannot support.
What are the best websites for education research resources?
For research standards and evidence reviews, start with the Institute of Education Sciences website, which publishes research standards and What Works Clearinghouse reviews, and the NCES website, which provides detailed technical documentation for federal datasets. For researchers navigating privacy-preserving research methods and accessing data across educational platforms and data sources, SafeInsights provides both the infrastructure and support to conduct studies that traditional repositories cannot facilitate.
What is the largest education dataset available?
The largest education datasets vary by definition — NCES maintains comprehensive U.S. national data, while UNESCO tracks education indicators for nearly every country worldwide. However, size alone doesn't determine research value. SafeInsights allows researchers to study student populations across platforms serving millions of U.S. students through privacy-preserving infrastructure, delivering a scale that extends beyond any single static dataset.
Are education datasets free to access?
Many education datasets from government sources like NCES, Data.gov, UNESCO, and OECD are freely accessible to researchers. However, "free" datasets often come with hidden costs: limited scope, delayed updates, small sample sizes, and months spent on data access negotiations. SafeInsights trades those hidden costs for a single review process; current cost and eligibility details are on the researchers page.
Start conducting research that transforms education with SafeInsights
Traditional education datasets offer valuable snapshots. Some of the most important questions about learning, though, cannot be answered from snapshots: they need student-level data, followed over time, inside the apps where learning happens. SafeInsights was built for those questions.
Our privacy-preserving infrastructure enables longitudinal analyses that would not be feasible through conventional data access. Study how students engage with adaptive learning technologies, compare instructional approaches across different platforms, or investigate long-term outcomes from educational interventions — all while maintaining the highest privacy standards. Researchers using SafeInsights spend their time conducting studies, not negotiating access agreements.
Start your research with infrastructure purpose-built for the questions that matter most.
References
Choi, Y., Lee, Y., Shin, D., Cho, J., Park, S., Lee, S., Baek, J., Bae, C., Kim, B., & Heo, J. (2020). EdNet: A large-scale hierarchical dataset in education. In I. I. Bittencourt, M. Cukurova, K. Muldner, R. Luckin, & E. Millán (Eds.), Artificial intelligence in education: 21st international conference, AIED 2020, Ifrane, Morocco, July 6–10, 2020, proceedings, part II (pp. 69–73). Springer. https://doi.org/10.1007/978-3-030-52240-7_13
Conaway, C., Keesler, V., & Schwartz, N. (2015). What research do state education agencies really need? The promise and limitations of state longitudinal data systems. Educational Evaluation and Policy Analysis, 37(1 Suppl.), 16S–28S. https://doi.org/10.3102/0162373715576073
Koedinger, K. R., Baker, R. S. J. d., Cunningham, K., Skogsholm, A., Leber, B., & Stamper, J. (2010). A data repository for the EDM community: The PSLC DataShop. In C. Romero, S. Ventura, M. Pechenizkiy, & R. S. J. d. Baker (Eds.), Handbook of educational data mining (pp. 43–56). CRC Press.
Kuzilek, J., Hlosta, M., & Zdrahal, Z. (2017). Open University Learning Analytics dataset. Scientific Data, 4, Article 170171. https://doi.org/10.1038/sdata.2017.171
Mello, M. M., Triantis, G., Stanton, R., Blumenkranz, E., & Studdert, D. M. (2020). Waiting for data: Barriers to executing data use agreements. Science, 367(6474), 150–152. https://doi.org/10.1126/science.aaz7028
National Center for Education Statistics. (n.d.). Statewide Longitudinal Data Systems Grant Program: Grantee states. U.S. Department of Education, Institute of Education Sciences. https://nces.ed.gov/programs/slds/stateinfo.asp
OECD. (2023). PISA 2022 results (Volume I): The state of learning and equity in education. OECD Publishing. https://doi.org/10.1787/53f23881-en
Reardon, S. F., Fahle, E. M., Ho, A. D., Shear, B. R., Saliba, J., Min, J., Shim, J., & Kalogrides, D. (2025). Stanford Education Data Archive (Version SEDA 2024) [Data set]. Stanford University. https://purl.stanford.edu/pt329xg7054
Stamper, J., Moore, S., Rosé, C. P., Pavlik, P. I., Jr., & Koedinger, K. R. (2024). LearnSphere: A learning data and analytics cyberinfrastructure. Journal of Educational Data Mining, 16(1), 141–163. https://doi.org/10.5281/zenodo.11109638
Tyagi, E., Ueyama, K., & Martin, A. (2025). educationdata: Retrieve records from the Urban Institute's Education Data Portal API (Version 0.1.5) [Computer software]. CRAN. https://CRAN.R-project.org/package=educationdata
Zhang, J., Baker, R. S., Basu Mallick, D., Heffernan, C., & Heffernan, N. (2026). A case study reexamining the cold-start problem in knowledge tracing models and implications for SafeInsights, an education research infrastructure (arXiv:2606.11004). arXiv. https://doi.org/10.48550/arXiv.2606.11004
Join leading researchers and Data Partners advancing research that improves learning outcomes — with privacy protection, minimal operational burden, and full control of data.
