Abstract. In large-scale educational assessments such as the Third International Mathematics and Sciences Study (TIMSS) or the Program for International Student Assessment (PISA), sizeable numbers of test administrators (TAs) are needed to conduct the assessment sessions in the participating schools. TA training sessions are run and administration manuals are compiled with the aim of ensuring standardized, comparable, assessment situations in all student groups. To date, however, there has been no empirical investigation of the effectiveness of these standardizing efforts. In the present article, we probe for systematic TA effects on mathematics achievement and sample attrition in a student achievement study. Multilevel analyses for cross-classified data using Markov Chain Monte Carlo (MCMC) procedures were performed to separate the variance that can be attributed to differences between schools from the variance associated with TAs. After controlling for school effects, only a very small, nonsignificant proportion of the variance in mathematics scores and response behavior was attributable to the TAs (< 1%). We discuss practical implications of these findings for the deployment of TAs in educational assessments.
Sample Attrition in the Canadian Survey of Labor and Income Dynamics
