High-Stakes Exams: Search for the Optimal Test Structure

Due to the ambiguous attitude towards assignments with the multiple-choice items and the tendency to completely exclude them from the structure of the Russian Unified State Exam (USE) in all subjects without reliance on empirical data, within this work we: - studied the world experience in the design of high-stakes graduation, and/or entrance exams; - compared items of different forms and types with each other, determining their advantages and disadvantages in comparison with each other; - empirically determined the optimal and most effective structure of the test in terms of measuring test subjects‘ teta. The hypothesis of our study was that a combination of items of various forms is able to provide the optimal structure of high-stakes exams. We proved that in the practical chapter by comparing the initial test of biology with a similar structure of the USE, when it still included three parts, with tests simulated on its basis having different structures. We compared reliability indicators and average measurement errors, consent statistics, indicators of the amount of information provided by the test as a result of testing subjects, and others. Initial test turned out to be most effective and optimal from all points of view. It combines items of various forms and types makes it possible to most profoundly measure subjects’ level of preparedness with the minimum measurement error and the greatest amount of information. Items of different forms and types complement each other, smoothing out some of their disadvantages. Also, the availability of items of this type enhance the objectivity and reliability of the measurement, allowing you to cover a large area of checked knowledge of a subject, without significantly increasing the time of execution of the entire test. We can conclude that it is worth taking into account the advantages of multiple-choice items and do not exclude the possibility of using them in testing.