Are differentially functioning mathematics items reason of low achievement of Turkish students in PISA 2015?

Abstract

In PISA 2015 the average mathematics score of Turkey decreased dramatically. One of the reasons could be the psychometric properties of mathematics items of PISA 2015. Therefore, it is necessary to evaluate PISA mathematics items for language DIF. In the study, three different DIF detection methods were used: logistic regression (LR), Mantel-Haenszel (MH) and structural equation modeling (SEM). Eleven items were found to have DIF when Turkish and English speaking students were compared. The effect sizes of mathematics performance differences between Turkish and English speaking students before and after excluding DIF items did not change which indicated that DIF items did not cause Turkish students to perform lower than expected. All the DIF items were open response format in which answers were rated by experts and computers. The DIF items favoring Turkish students were mainly related to the basic cognitive process.