Why the OECD Socioeconomic Gap Hasn’t Narrowed: A Change in Measurement, Not in Results, in PISA 2025

Marc Colomer|22/09/2026|4 min read

In collaboration with: Eudald Correig, Lucas Gortazar, Isaac Sayol, Lucía Cobreros

Why the OECD Socioeconomic Gap Hasn’t Narrowed: A Change in Measurement, Not in Results, in PISA 2025

The background: what the PISA 2025 report published

On September 8, the OECD published the results of the PISA 2025 edition. The results showed a sharp drop in performance among students across OECD countries, placing the OECD average at its lowest level of any PISA edition since 2000.

Volume I of PISA 2025 (OECD) notes that the decline was especially pronounced among students from affluent families, and that this caused the performance gap between socially advantaged and disadvantaged students to narrow between 2022 and 2025 in numerous countries —including Spain—. Specifically, the average drop for the highest socioeconomic quartile was 14 points in mathematics, 7.7 in science, and 20.7 in reading: far greater than for the other three quartiles, which fell by an average of 3.1 points in mathematics, 6.5 in reading, and even improved by 4.4 points in science.

The report described this as “a new phenomenon” of students from relatively affluent families facing academic difficulties, without addressing the causes that might explain this new pattern. This “new phenomenon” has had a major scientific and media impact worldwide and has fueled numerous hypotheses about its potential causes, including the idea of greater AI use in wealthier households.

Findings from the Innovamat Research Lab study

The working paper shows that the decline observed among more advantaged students is explained by a change in the tool the OECD uses to measure students’ socioeconomic status. However, when measured consistently over time, the socioeconomic gap has not narrowed, and in 2025 it reaches its highest level in the past decade.

The OECD introduced several changes to how it measures the Economic, Social and Cultural Status index (ESCS) of students in the 2025 edition. This index is calculated as the arithmetic mean of three components: parental education, parental occupation, and an index of a broad set of household possessions. Within the household possessions index, the 2025 questionnaire removed eleven items, including the eight questions about types of books at home — the ones most closely linked to learning. In their place, it substituted new items related to living conditions (guest room, security system, swimming pool, running water, or electricity), which are nearly universal across the OECD and less closely tied to academic performance.

Still, this change meant the 2025 index was no longer comparable to the 2022 one. The change in measurement meant that a notable share of students with strong PISA results and high cultural capital, but lower living conditions, no longer appeared in the top socioeconomic quartile in PISA 2025 (whereas they would have in previous editions). At the same time, students with weaker results and lower cultural capital, but higher living conditions, took their place in the highest socioeconomic group. As a result of these changes, the drop observed in 2025 using the socioeconomic index published in the report was larger than for the other groups. Although PISA 2025’s own Technical Report warned that the 2025 index is not comparable to that of previous cycles, Volume I compares the two statistics anyway.

Adapting these indices makes sense as societies and living conditions change, but making comparisons across years required methodological adjustments. It’s reasonable for the way we measure the socioeconomic conditions students grow up in to be updated as economies and societies change: that’s what has happened in every PISA edition, and in that sense, this is nothing new. This paper’s goal is in no way to evaluate the OECD’s socioeconomic status measures in this edition, but rather to reliably compare the evolution of performance across socioeconomic groups over time.

The question we answer is what happens to the drop in results by socioeconomic group when measured with an index that is genuinely comparable across editions. By reconstructing the ESCS using public microdata and a methodology comparable to the OECD’s, the study finds that: