Negative marking in UMAT

You will never know whether or not there is negative marking in UMAT. This is because ACER does not release information on how your ‘markings’ on the answer sheet are converted into UMAT scaled scores. This information is kept ‘confidential’.

Even with negative marking, you will not be disadvantaged by random guessing. For example, in the Australian Mathematics Competition negative marking was used in the past – you got minus one fourth for choosing the wrong option out of the five options. This means if there are 100 questions and you guess randomly, you will get 20 right and 80 wrong, so 20-80/4=0. When negative marking is used, it is weighted so that students are not disadvantaged by random guessing. So random guessing is a zero sum game. Most high stakes tests use ‘negative marking’ to avoid the ‘lucky monkey’ problem.

While ACER says ‘no marks are deducted for an incorrect answer’ and ‘all questions carry equal marks’, unless the ‘confidential’ information on how your answers to UMAT questions are converted into scaled scores, you can never be certain. ACER also says ‘all questions carry equal marks,’ but how is this possible when all candidates do not get the same questions? It is known that there are some differences in the questions given to students sitting the UMAT in the same year. Further, ACER says you can’t prepare for UMAT. Evidence is to the contrary.

Most high stakes tests take several factors into account (one of them being negative marking) when calculating results. However, since they do not release the information on the factors they take into account, we can only draw conclusions based on available scientific evidence – on Rasch models and so on. ACER may or may not use negative marking in obtaining the raw score (which is irrelevant), but almost certainly uses some variation of it (eg, the percentage of questions right out of those attempted) in arriving at the scaled score (which is what is relevant).

The best approach for students is to try to eliminate at least one (preferably two) of the options and then guess; rather than randomly guess.