Scholarlink

Background <p><i>International large-scale assessments</i> collect achievement and background data to evaluate the effectiveness of teaching and learning processes, thereby informing education policy. The Western Balkan countries have recently begun participating in these studies to monitor their progress toward European Union benchmarks. A fundamental challenge in the context of international large-scale assessments is ensuring that scales function equivalently across diverse participating countries so that they measure the same construct. Given the region’s conflicts in the 1990s, which disrupted schooling and destroyed household resources, the comparability of background indicators such as the books-at-home item (ASBG04) remains underexamined. Proxies for socioeconomic status may not function equivalently across structurally diverse, post-conflict contexts, yet background measures are often assumed to be invariant without empirical verification.</p> Methods <p>We tested cross-country measurement invariance of the TIMSS 2019 <i>Home Resources for Learning</i> scale, focusing specifically on the books-at-home item (ASBG04). Using TIMSS 2019 Grade 4 data from 30 countries (24 EU countries, 6 Western Balkan countries; <i>N</i> = 130,714 students), we estimated item parameters using concurrent multiple-group IRT with partial credit models (MG-IRT) in SAS PROC IRT. Threshold parameters for the books-at-home item were constrained under three model specifications: <i>Model 1</i> constrained the thresholds of the books-at-home item to be equal across all 30 countries. In <i>Model 2</i>, the books-at-home item thresholds were held equal for the EU group and were freely estimated for each Western Balkan country. In <i>Model 3</i>, the books-at-home item thresholds were freely estimated for each country. The other four items of the <i>Home Resources for Learning</i> scale were constrained to be invariant across countries in all models, serving as anchors. We compared the books-at-home item’s thresholds across the three model specifications.</p> Results <p><i>Model 1</i>, which enforced equal thresholds, showed a poor fit to the data. Allowing country-specific thresholds in <i>Models 2</i> and <i>3</i> substantially improved model fit, indicating <i>differential item functioning</i> for the books-at-home item across countries. Findings from <i>Models 2</i> and <i>3</i> revealed marked variations in item difficulty. The books-at-home item was therefore not strictly invariant in this sample.</p> Conclusions <p>Treating the books-at-home item as invariant risks masking meaningful cross-country differences. Routine invariance checks of background scales are recommended before comparing socioeconomic gaps across countries. Ensuring at least partial invariance is essential for fair monitoring of educational outcomes and for targeting resources in education systems recovering from conflict.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Books as educational resource? Assessing cross-country measurement invariance of the TIMSS home resources for learning scale in post-conflict societies

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月19日

查看全文
  • Donieta Jusufi,
  • Güler Yavuz Temel,
  • Knut Schwippert

摘要

Background

International large-scale assessments collect achievement and background data to evaluate the effectiveness of teaching and learning processes, thereby informing education policy. The Western Balkan countries have recently begun participating in these studies to monitor their progress toward European Union benchmarks. A fundamental challenge in the context of international large-scale assessments is ensuring that scales function equivalently across diverse participating countries so that they measure the same construct. Given the region’s conflicts in the 1990s, which disrupted schooling and destroyed household resources, the comparability of background indicators such as the books-at-home item (ASBG04) remains underexamined. Proxies for socioeconomic status may not function equivalently across structurally diverse, post-conflict contexts, yet background measures are often assumed to be invariant without empirical verification.

Methods

We tested cross-country measurement invariance of the TIMSS 2019 Home Resources for Learning scale, focusing specifically on the books-at-home item (ASBG04). Using TIMSS 2019 Grade 4 data from 30 countries (24 EU countries, 6 Western Balkan countries; N = 130,714 students), we estimated item parameters using concurrent multiple-group IRT with partial credit models (MG-IRT) in SAS PROC IRT. Threshold parameters for the books-at-home item were constrained under three model specifications: Model 1 constrained the thresholds of the books-at-home item to be equal across all 30 countries. In Model 2, the books-at-home item thresholds were held equal for the EU group and were freely estimated for each Western Balkan country. In Model 3, the books-at-home item thresholds were freely estimated for each country. The other four items of the Home Resources for Learning scale were constrained to be invariant across countries in all models, serving as anchors. We compared the books-at-home item’s thresholds across the three model specifications.

Results

Model 1, which enforced equal thresholds, showed a poor fit to the data. Allowing country-specific thresholds in Models 2 and 3 substantially improved model fit, indicating differential item functioning for the books-at-home item across countries. Findings from Models 2 and 3 revealed marked variations in item difficulty. The books-at-home item was therefore not strictly invariant in this sample.

Conclusions

Treating the books-at-home item as invariant risks masking meaningful cross-country differences. Routine invariance checks of background scales are recommended before comparing socioeconomic gaps across countries. Ensuring at least partial invariance is essential for fair monitoring of educational outcomes and for targeting resources in education systems recovering from conflict.