EDUCATION

New Scale Gives Student Beliefs Clearer Scores

ChinaSun Nov 22 2026

Old tools for measuring students' political literacy often felt fuzzy. They did not always separate ideas clearly. They also did not always compare groups fairly. A new study tried to fix that. It built and tested a five-part scale for Chinese university students in the New Era. The parts are political identification, patriotic feeling, moral growth, rule-of-law awareness, and cultural confidence.

To make the scale, researchers used several steps. They coded ideas from a framework. They asked experts twice through a Delphi process. They checked content validity. They ran a pilot test. These steps produced 47 possible items. Then they collected calibration data from 3,716 undergraduates. The students came from 28 universities in 12 provinces. The sample used stratified cluster sampling.

The researchers used multidimensional item response theory to test the model. They applied a compensatory multidimensional graded response model. They estimated it with marginal maximum likelihood. They also used Metropolis-Hastings Robbins-Monro estimation as a same-sample check. They compared different dimension layouts. They examined reliability, measurement invariance, and differential item functioning. For the 47-item oblique model, the fit values were RMSEA = 0.043, CFI = 0.957, and TLI = 0.951.

After four items were removed, the 43-item model improved. Its fit values were RMSEA = 0.039, CFI = 0.964, and TLI = 0.958. The marginal reliability estimates ran from 0.82 to 0.91. Some bifactor models fit slightly better in approximate terms. But their indices did not support a broad general factor. The scale showed scalar invariance across gender, grade, and disciplinary clusters. Eight items showed differential item functioning across urban-rural or party-membership comparisons. That led to proposed wording changes or caution in interpretation.

The scale gives scores for each dimension. It also offers a base for more measurement work. Still, the study has limits. Model selection used the same sample. External validity evidence is incomplete. Clustering was not adjusted in the multidimensional item response theory and item-bias analyses. More independent validation is required. The findings do not show the scale is ready for an operational computerized adaptive test.

actions