Data Sources

Comprehensive guide to the PISA data underlying EduStrat (Educational Stratification in PISA)

EduStrat is designed to study the intergenerational transmission of educational achievement—specifically, how parental characteristics (education, occupational status, and household wealth, captured through the ESCS index) relate to children's academic performance at age 15 in mathematics, reading, and science across more than 100 countries worldwide.

1. The PISA Programme

1.1 Programme Overview

The Programme for International Student Assessment (PISA) is a triennial international survey coordinated by the Organisation for Economic Co-operation and Development (OECD) since 2000. PISA assesses the extent to which 15-year-old students near the end of compulsory education have acquired the knowledge and skills essential for full participation in modern societies.

1.2 PISA Assessment Cycles

PISA has been conducted in the following years, with each cycle focusing on one major domain while assessing all three:

Year	Major Domain	Countries/Economies	Students Assessed	Status
2000	Reading	43	~265,000	Available in App
2003	Mathematics	41	~276,000	Available in App
2006	Science	57	~398,000	Available in App
2009	Reading	65	~475,000	Available in App
2012	Mathematics	65	~510,000	Available in App
2015	Science	72	~540,000	Available in App
2018	Reading	79	~600,000	Available in App
2022	Mathematics	81	~690,000	Available in App

This application includes data from all PISA cycles from 2000-2022 (8 assessment cycles: 2000, 2003, 2006, 2009, 2012, 2015, 2018, 2022), covering 513 country-year combinations across 101+ unique countries/economies.

2. Assessment Framework

2.1 Mathematics Assessment

PISA mathematics assesses students' capacity to formulate, employ, and interpret mathematics in a variety of contexts. It includes reasoning mathematically and using mathematical concepts, procedures, facts, and tools to describe, explain, and predict phenomena.

2.2 Reading Assessment

PISA reading literacy assesses students' capacity to understand, use, evaluate, reflect on, and engage with texts in order to achieve goals, develop knowledge and potential, and participate in society.

2.3 Science Assessment

PISA science literacy assesses the ability to engage with science-related issues and with the ideas of science, as a reflective citizen. It includes understanding natural phenomena, designing scientific enquiry, and interpreting evidence.

3. Sampling Design

3.1 Target Population

PISA targets students who are between 15 years 3 months and 16 years 2 months at the time of assessment, regardless of their grade level. This age range was chosen because students at this age are approaching the end of compulsory schooling in most OECD countries.

3.2 Two-Stage Stratified Sampling

PISA employs a sophisticated two-stage stratified sampling design:

1. Stage 1 - School Sampling: Schools are sampled with probability proportional to size (PPS), where size is the number of eligible 15-year-old students. Typically, 150-200 schools per country are selected.
2. Stage 2 - Student Sampling: Within each selected school, approximately 35 students are randomly sampled from the complete list of eligible 15-year-old students.

3.3 Sampling Weights

PISA provides several types of sampling weights to ensure representative estimates:

• Student Weight (W_FSTUWT): Adjusts for unequal probability of selection and non-response. Use this for most analyses. In the learningtower package this variable is named stu_wgt.
• Senate Weight (W_FSENWT): Gives each country equal weight regardless of population size. Use for cross-country comparisons where you want each country to contribute equally.
• Replicate Weights (W_FSTR1-W_FSTR80): Used for balanced repeated replication (BRR) variance estimation.

4. The learningtower R Package

4.1 Package Overview

EduStrat uses data processed through the learningtower R package (Wang et al., 2024), which provides harmonized, analysis-ready PISA data in a consistent format.

4.2 Package Installation

The learningtower package is available on CRAN:

install.packages("learningtower")
library(learningtower)

4.3 Data Access via learningtower

The package provides easy access to PISA data:

# Load all student data for a specific year
data_2018 <- load_student(2018)

# Load data for specific countries
data_usa <- load_student(2018, countries = "USA")

# Access codebook
codebook <- load_codebook()

4.4 Package Citation

If you use data from this application, please cite the learningtower package:

5. Data Structure in This Application

5.1 JSON Chunk Format

EduStrat pre-generates 513 country-year specific JSON files (e.g., USA_2018.json) for efficient progressive loading. Each chunk contains:

{
  "country": "USA",
  "year": 2018,
  "n_students": 4838,
  "data_quality": {
    "missing_math": 0,
    "missing_reading": 0,
    "missing_science": 0,
    "missing_escs": 0,
    "complete_cases": 4838
  },
  "students": [
    {
      "year": 2018,
      "country": "USA",
      "school_id": "USA_2018_84000001",
      "student_id": "USA_2018_84000250",
      "mother_educ": "ISCED 3A",
      "father_educ": "ISCED 3A",
      "gender": "male",
      "computer": "yes",
      "internet": "yes",
      "math": 534.751,
      "science": 488.669,
      "stu_wgt": 646.6246,
      "desk": "yes",
      "room": "no",
      "television": "3+",
      "computer_n": "1",
      "car": "3+",
      "book": "11-25",
      "wealth": 0.7125,
      "escs": -0.4811,
      "reading": 590.33
    }
  ]
}

5.2 File Organization

• Location: data/country-year/
• Naming Convention: {COUNTRY}_{YEAR}.json (e.g., USA_2018.json, DEU_2022.json)
• Total Files: 513 files
• File Size Range: 0.1–15 MB per file
• Total Data Size: ~1.25 GB

5.3 Metadata File

The application also includes a metadata.json file that catalogs all available data:

{
  "countries": ["ALB", "ARG", "AUS", ...],
  "years": [2000, 2003, 2006, 2009, 2012, 2015, 2018, 2022],
  "variables": {
    "math": "Mathematics achievement score",
    "reading": "Reading achievement score",
    "science": "Science achievement score",
    "escs": "PISA index of economic, social and cultural status",
    ...
  }
}

6. Variable Codebook

All variable names in the JSON data files are taken directly from the learningtower R package (Wang et al., 2024), with the sole exception that read is renamed to reading during chunk generation. The following variables are used by EduStrat in its analyses. The JSON data files also contain additional learningtower fields (wealth, computer, internet, desk, room, book, television, car, computer_n) that are not currently used by the application.

6.1 Outcome Variables

Note: PISA uses plausible values to account for measurement error. EduStrat uses the first plausible value (PV1) for each domain. Full analyses should consider all plausible values (5 in earlier cycles, 10 from PISA 2015 onward) and combine results using Rubin's rules.

6.2 Predictor Variables

Note: EduStrat derives a parent_edu variable at runtime by taking the higher of mother_educ and father_educ (converted to years of schooling via ISCED mapping). This serves as the alternative predictor option in the application.

6.3 Demographic Variables

6.4 Sampling Weight

6.5 Identification Variables

7. Official OECD Data Access

7.1 OECD PISA Data Portal

The official source for all PISA data is the OECD PISA Data Portal:

7.2 Cycle-Specific Data

• PISA 2022: https://www.oecd.org/pisa/data/2022database/
• PISA 2018: https://www.oecd.org/pisa/data/2018database/
• PISA 2015: https://www.oecd.org/pisa/data/2015database/
• PISA 2012: https://www.oecd.org/pisa/data/2012database/

7.3 Technical Documentation

• PISA 2018 Technical Report: Technical Report
• PISA Data Analysis Manual: OECD (2009). PISA data analysis manual: SPSS (2nd ed.). https://doi.org/10.1787/9789264056275-en

8. Data Quality and Limitations

8.1 Strengths

• Large, representative samples from each participating country
• Standardized assessment framework enabling cross-country comparisons
• Rich contextual questionnaires on student background
• High-quality sampling design with rigorous quality control
• Transparent documentation and public data availability

8.2 Limitations

• Cross-sectional design: Data from each cycle are collected at a single point in time, limiting causal inference
• Coverage exclusions: Some countries exclude specific populations (e.g., remote regions, special education students)
• Response rates: Vary by country; some countries have lower participation rates
• Translation effects: Assessments are translated into 40+ languages, potentially affecting comparability
• Cultural context: Test items may have different cultural relevance across countries

8.3 Missing Data

PISA data contain missing values due to:

• Item non-response (student skipped a question)
• Not reached (student ran out of time)
• Not administered (question not in student's test booklet)
• Questionnaire non-response

This application uses complete-case analysis by default. Advanced users should consider multiple imputation methods for handling missing data.

9. Countries Included in This Application

9.1 OECD Countries

The following 38 OECD countries are available (availability varies by year):

9.2 Partner Countries/Economies

An additional 60+ partner countries and economies are available, including:

* Note: Some countries participate through specific regions or provinces rather than nationally

10. References and Further Reading

10.1 Primary Sources

1. OECD. (2024). PISA 2022 Technical Report. OECD Publishing. https://doi.org/10.1787/01820d6d-en
2. OECD. (2009). PISA data analysis manual: SPSS (2nd ed.). OECD Publishing. https://doi.org/10.1787/9789264056275-en
3. Wang, K., Yacobellis, P., Siregar, E., Romanes, S., Fitter, K., Dalla Riva, G. V., Cook, D., Tierney, N., Dingorkar, P., Sai Subramanian, S., & Chen, G. (2024). learningtower: OECD PISA datasets from 2000–2022 in an easy-to-use format (R package version 1.1.0). https://doi.org/10.32614/CRAN.package.learningtower

10.2 Methodological References

4. Avvisati, F. & Wuyts, C. (2024). The measurement of socio-economic status in PISA. OECD Education Working Papers, No. 321. OECD Publishing. https://doi.org/10.1787/0c5b793c-en
5. Wu, M. (2005). The role of plausible values in large-scale surveys. Studies in Educational Evaluation, 31(2–3), 114–128. https://doi.org/10.1016/j.stueduc.2005.05.005

10.3 Documentation in This Application

• Methodology Documentation - Statistical methods and formulas
• How to Cite - Citation guide for academic use

← Back to Application