Puerto Rico and the PISA Tests: An Educational Advancement Opportunity

Full analysis in English · Resumen en español al final del análisis.

By Héctor Navedo, Ed.D.

Updated October 27, 2026.

Puerto Rico’s participation in PISA 2015 offers a useful basis for educational planning, provided leaders connect the results to decisions the assessment can actually support. The strongest case for international assessment is not a position in a league table. It is the opportunity to examine students’ application of knowledge, investigate unequal learning opportunities, and test whether proposed improvements address needs documented locally. Participation alone does not improve instruction.

The Programme for International Student Assessment, coordinated by the Organisation for Economic Co-operation and Development (OECD), assesses reading, mathematics, and science literacy among 15-year-old students. Its emphasis on applying knowledge complements, rather than substitutes for, evidence about the curriculum taught in Puerto Rico. The OECD framework defines particular competencies; it is not a complete account of educational quality, cultural learning, or civic purpose (OECD, 2017a).

What the 2015 evidence establishes

Puerto Rico’s sample included public and private schools and was separate from the U.S. national sample; Puerto Rico’s results were not included in the reported U.S. average. Treating that average as Puerto Rico’s result would therefore be incorrect (Kastberg et al., 2016).

The National Center for Education Statistics (NCES) reports the following estimates.

Verified evidence and its limits

Science: 403; standard error 6.1.

Scope and primary source: Puerto Rico, 2015. NCES, n.d.-d, Table S1.

Interpretation limit: An estimated population mean, not an individual diagnosis.

Reading: 410; standard error 7.1.

Scope and primary source: Puerto Rico, 2015. NCES, n.d.-c, Table R1.

Interpretation limit: Not a percentage correct or a school grade.

Mathematics: 378; standard error 5.6.

Scope and primary source: Puerto Rico, 2015. NCES, n.d.-b, Table M1.

Interpretation limit: Not a direct measure of mastery of the local curriculum.

1,398 students; 47 schools; 100% eligible-school participation; 93.0% weighted student participation.

Scope and primary source: Achieved Puerto Rico sample. NCES, n.d.-a, Tables A-2–A-3.

Interpretation limit: Response rates do not describe coverage of all 15-year-olds.

Population coverage: 60.1%; overall student exclusion rate: 3.1%.

Scope and primary source: Different denominators. NCES, n.d.-a, Table A-3; OECD, 2017b, Table 11.2.

Interpretation limit: Neither the exclusion rate nor noncoverage is a dropout rate.

NCES reported Puerto Rico below the U.S. average in each domain. The corresponding U.S. means were 496 in science, 497 in reading, and 470 in mathematics. These statistically significant differences identify substantial performance concerns in the assessed population; they do not identify their causes (Kastberg et al., 2016). Comparing Puerto Rico’s three domain scores with one another is not a substitute for examining each domain’s framework and proficiency distribution.

The distribution also matters. NCES estimated that 55% of the population represented by Puerto Rico’s sample performed below Level 2 in science, 50% in reading, and 73% in mathematics. These rounded proficiency estimates indicate the scale of the instructional challenge; Level 2 is an assessment-specific benchmark, not a local passing grade (Kastberg et al., 2016).

Uncertainty belongs beside the scores. Using the published rounded means and standard errors, approximate 95% confidence intervals are 391–415 for science, 396–424 for reading, and 367–389 for mathematics. These are calculations of mean ± 1.96 standard errors, not additional official estimates. They express statistical uncertainty, not protection against coverage bias or every source of measurement error.

Coverage and comparability change the policy interpretation

PISA targets students enrolled in grade 7 or higher. Coverage Index 3 compares the weighted population represented by the non-excluded sample with the total population of 15-year-olds, including those outside school. For Puerto Rico, 60.1% coverage and 93.0% student participation answer different questions. The first concerns who is represented; the second concerns response among eligible sampled students. OECD also recorded that the 1,398 participants fell below the 1,500-student standard for an additional adjudicated paper-mode entity, while explicitly accepting the data into the final database (OECD, 2017b, pp. 204–209, 273).

The policy implication is to report learning and inclusion together. Noncoverage does not reveal the achievement or educational status of every unrepresented young person. Leaders should investigate enrollment, grade placement, exclusions, withdrawals, and population estimates before assigning a cause. In a future comparison, expanding participation could change the composition of the assessed population: a lower mean would not automatically mean that instruction deteriorated.

Administration matters as well. Puerto Rico used paper-based trend items, with instruments adapted from a Spanish source; the U.S. main assessment was computer-based. Puerto Rico did not receive science subscale scores. These differences require attention to the specific constructs available, rather than either dismissing all comparisons or assuming identical coverage of every skill (Kastberg et al., 2017, pp. 2, 39, 56).

Use the results to choose inquiries, not predetermined reforms

A system-level benchmark can justify investigating opportunities to learn: whether students encounter demanding texts, explain mathematical reasoning, or interpret scientific evidence. It cannot establish that a particular curriculum, teacher practice, governance arrangement, or funding decision caused a score. Associations between background characteristics and achievement can generate hypotheses; they do not supply an intervention’s causal effect.

For doctoral research, secondary analysis requires the survey weights, replicate-weight variance procedures, and all ten plausible values, which represent proficiency uncertainty. Ordinary analyses that treat the observations as a simple random sample can misstate precision. PISA is designed for population estimates, not reliable individual student scores (Kastberg et al., 2017, pp. 64–65). A single 2015 observation also cannot establish a current trend or evaluate a reform introduced afterward.

For district leaders, the next step should be triangulation with current curriculum-aligned assessments, student work, attendance, course access, and enrollment records. Those sources must retain their own populations and scales: PISA points should not be converted into local grades or mechanically merged with another assessment. The question is whether independent evidence points toward the same actionable need.

A proposed decision matrix for educational leaders

The following is a planning proposal, not a validated instrument or an OECD/NCES protocol. Its purpose is to make authority, resources, and decision rules explicit.

From evidence to a bounded decision

Investigate a learning need.

Owner and resources: Curriculum leader; paid teacher planning time and assessment expertise.

Evidence and stopping rule: Check current student work and curricular alignment. Do not purchase a program on PISA scores alone.

Audit inclusion and data quality.

Owner and resources: Data director; secure enrollment records and a sampling specialist.

Evidence and stopping rule: Reconcile age, grade, sector, exclusions, and missing records. Pause conclusions when denominator discrepancies remain unresolved.

Pilot an instructional change.

Owner and resources: Instructional director and evaluator; coaching, implementation time, and evaluation budget.

Evidence and stopping rule: Specify outcomes, baseline, comparison strategy, review date, and meaningful improvement beforehand. Stop or redesign for harm, infeasibility, or failure to meet the agreed criterion.

Consider another PISA cycle.

Owner and resources: Education authority; costed sampling, accessibility, administration, analysis, and dissemination plan.

Evidence and stopping rule: Identify the decision that new evidence would change. Defer commitment without adequate coverage planning, funding, or analytical capacity.

Address the strongest objections before committing resources

Cost and instructional time are legitimate objections. A participation budget should include staff release time, accommodations, quality assurance, analysis, and follow-through, not merely test administration. Leaders should compare that investment with alternatives capable of answering the same policy question. If current local evidence already answers it, another assessment may add little.

Curricular and cultural comparability are also substantive concerns. International competencies can broaden discussion without becoming Puerto Rico’s entire curriculum. A defensible approach is to review released tasks against local learning goals, preserve culturally important content that PISA does not measure, and avoid test-preparation policies that narrow instruction. Technical standardization supports comparison; it does not make educational purposes identical.

Puerto Rico can use PISA productively by connecting a credible historical benchmark to current, locally grounded inquiry. Success should be judged by better decisions, more inclusive evidence, and demonstrable improvements in learning opportunities, not by participation itself or by a ranking detached from its population and uncertainty.

Resumen en español

Puerto Rico participó en PISA 2015 con una muestra separada de la muestra nacional estadounidense. Sus promedios fueron 403 en ciencias, 410 en lectura y 378 en matemáticas. La cobertura poblacional de 60.1% no equivale a la participación estudiantil ponderada de 93.0%, ni permite calcular directamente la deserción escolar. Los resultados orientan preguntas sobre aprendizaje e inclusión, pero no demuestran qué política causó el desempeño. La propuesta es contrastarlos con evidencia local reciente, asignar responsables y recursos, y evaluar cualquier intervención con criterios establecidos de antemano. La matriz presentada es una herramienta propuesta de planificación, no un instrumento validado. Fuentes: NCES y OECD, citadas arriba.

References