Background:The Australian Tertiary Admissions Rank (ATAR), represented by a number between 0.00 and 99.95, is the most important metric used by Australian tertiary institutions when selecting students for entry to undergraduate degrees.
Objective: To quantify the statistical precision of an ATAR by creating a realistic database of scores for 10,000 students taking six VCE subjects in 100 different schools.
Methods: Approximately 500,000 test scores were processed into study scores, scaled study scores, aggregate scores and finally converted into an ATARfor each of the 10,000 students. The same dataset was processed twice: once using outcome scores that had been rounded to integer values, as is required by the VCAA (ATAR1), and again using students’ actual, non-rounded outcome scores (ATAR2). The rounding error (i.e. the difference between ATAR1 and ATAR2) was then analysed.
Results: In this simulation, we found that not only are the two decimal places on an ATAR statistically insignificant, but that ATAR results are only accurate for 100% of students when reported to the nearest multiple of two (i.e. to the nearest even number).
Limitations: Privacy laws prevented us from using real student data in this analysis.
Conclusion: This paper calls for ATARs to be reported only to the nearest even number, and for continued modernisation of the undergraduate admissions criteria used by Australian tertiary institutions to include a broader range of student attributes.