Predictors of individual performance changes related to item positions in PISA assessments

Background: Item position effects have been a common concern in large-scale assessments as changing the order of items in booklets may have an undesired effect on test performance. If every test taker would be affected by the effect in the very same way, comparisons between groups of individuals would still be valid. However, research has shown that in addition to a general fixed effect of item positions, the extent of the effect varies considerably across individuals. These individual differences are referred to as persistence. Test takers with a high level of persistence are able to keep up their performance better throughout the test administration, whereas those with a lower level of persistence show a larger decline in their test performance. Methods: The present study applied a multilevel extended item response theory (IRT ) framework and used the data from the PISA 2006 science, 2009 reading, and 2012 mathematics assessments. The first objective of this study is to provide a systematic investigation of item position effects across the three PISA domains, partially replicating the previous studies on PISA 2006 and 2009. Second, this study aims to gain a better understanding of the nature of individual differences in position effects by relating them to student characteristics. Gender, socio-economic status, language spoken at home, and three motivational scales (enjoyment of doing the subject being assessed, effort thermometer, perseverance) were used as person covariates for persistence. Results: This study replicated and extended the results found in previous studies. An overall negative item cluster position effect and significant individual differences in this effect were found in all the countries in the three PISA domains. Furthermore, the most frequently observed effect of person covariates on persistence is gender, with girls keeping up their performance better than boys. Other predictors showed little or inconsistent effects on persistence. Conclusions: Our study demonstrated inter-individual differences as well as group differences in item position effects, which may threaten the comparability between persons and groups. The consequences and implications of item position effects and persistence for the interpretation of PISA results are discussed. (DIPF/Orig.