Relationship between mathematics achievement and home possessions: Cluster analysis across 73 countries using PISA 2022

• In Event: AERA Roundtable Session Friday 11:40 am Four Seasons Ballroom 2-3
  In Roundtable Session: Exploring Student SES and OTL From the Perspectives of Large-Scale Education Studies (Table 2)

Abstract

Understanding educational disparities requires valid, reliable, and comparable data on students’ SES because it explains differences in academic outcomes and contextualizes assessment results (Broer et al., 2019; Lee, 2019). This study focuses on PISA’s home possession items, which is one of the components used to measure SES, along with parents’ education and occupation (Avvisati, 2020). The aim of this study is to identify key patterns across countries in the relationship between home possessions and academic performance, which can help improve future measures of the home possession indicators across various countries.

Data from the 2022 PISA student questionnaire (OECD, n.d.), covering 73 countries (35 OECD members and 38 partners), were used for the study. Each country was weighted to be nationally representative. Sample sizes for each group are in Table 1. Logistic regressions were conducted to predict whether a student achieved Math level 2 (considered as the baseline level of proficiency that students need to participate fully in society; OECD, 2023) or above using each of the 17 home possession items (5 dichotomous, 12 polytomous). The IDB Analyzer was used to take into account all 10 plausible values for Math. The home possession items and the response categories are in Table 2. The lowest category for each item was the reference category, so the coefficient for each subsequent category in the logistic regression refers to the increase in the log odds of a student achieving Math level 2 and above, compared to a student with the lowest number of the item. Subsequently, a cluster analysis was conducted with the logistic regression coefficients using the K-means clustering method in SPSS. This grouped countries with similar logistic regression coefficients across the items.

Six clusters were identified, as presented in Figure 1. The breakdown of the countries in each cluster based on the income level (Eric, Young, & Eapen, 2024) and region are presented in Tables 5 and 6, respectively. For each cluster, the average of the logistic regression coefficients for the dichotomous items are presented in Table 3 and Figure 2, while the results for the polytomous items are presented in Table 4 and Figure 3. In Figure 2 and Figure 3, it can be noted that Cluster 4 has higher coefficients than the other clusters, representing a strong relationship between home possessions and math achievement. In Figure 3, three patterns can be observed in the average of the logistic regression coefficients for the polytomous items: (1) Increasing with the number of items; (2) Increasing until the second-to-last category, then dropping; (3) Mostly negative coefficients across the categories.
This study provides insights into how the home possessions scale can be improved in future cycles of PISA. For example, items for which the logistic regression coefficients drops for the last category can be modified to have less categories, as the last category may include too few students to properly capture the relationship between their wealth and mathematics achievement.

Authors

• Doo Rhee Lee, The George Washington University

• Selene Sunmin Lee, McGraw Hill

• Anita M. Sands, Educational Testing Service

• Jonas P Bertling, Educational Testing Service