Putting exams to the test: Do entrance exams predict academic achievement?

4open study suggests entrance exam scores may not predict future success at university

Soon, students around the world will begin poring over textbooks for their university entrance exams. Many hope to reach that slim, tail end of the results curve that shows the very best performers. Others may worry about ending up at the other end.

But in Brazil, being at the ‘good’ end of the curve doesn’t necessarily mean staying there after you’re accepted, according to a new study published in the open access journal 4open.

Mathematicians at the University of Campinas in Brazil investigated the link between students' university entrance exam results and their later undergraduate performance.

“The general idea here in Brazilian universities is that you can predict some abilities, especially in the ‘exact sciences’ like maths or physics,” says lead author Professor Verónica González-López.

Calculating success

The university uses entrance exams to place students in their undergraduate courses. For those wishing to enter the statistics course, the most relevant tests are Mathematics and Physics (known before 2015 as Mathematics and Natural Sciences).

The assumption is that students who perform well in these two exams will also perform well in the university’s statistics course – a linear way of calculating success that Prof González-López and her colleagues decided to test.

“You can do calculations in an ‘automatic’ way, but you need something else. Some part is missing,” she explains, pointing to the lack of evaluating a student’s logic and use of techniques.

Copulas point the way

To see if there was a better method, the authors used a different mathematical model – called a copula – to look at how students with very high and very low scores in the entrance exam performed in the following undergraduate Calculus I exam.

Copulas give mathematicians better flexibility. They can indicate whether normal linear models should be used, but they also factor in other nuanced calculations when finding relationships between the entrance exam results and academic performance.

The study showed that the copula model never indicated that the linear model should be used. In other words, the model used by the university is not complex enough to best predict future performance.

Even though the linear model’s ability to predict performance has gradually increased over the years, this has happened at a very poor rate. Prof González-López believes this may be due to an entrance exam that is not tuned with the ‘preliminary notions’ of calculus or with different methods of teaching at school and at university.

Better evaluation means better education

Prof González-López also thinks those who perform poorly in the entrance exam could still do well at university. “I really believe that you can find potentially good students in these low grades. Maybe they just need to learn something,” she says. She hopes future work will find out what factors can help those with low to medium scores in the entrance exams earn higher marks in their university exams.

Interest in education results also led Prof González-López and her colleagues to study the relation between students below the baseline of educational development per school and a school’s annual failure rate.

Such evaluations are needed for a good educational system, says Prof González-López. “The Brazilian system is very young and vigorous, and undergoing consolidation – though this hasn't happened yet. Qualified studies such as ours are very important to help provide a solid system for the student.”

This "Key Summaries" article is brought to you in collaboration with SciencePOD.