If you're doing a large-scale study, exam scores are basically the only way to get quantitative data.

And, exams aren't that bad! A well-designed exam can't be passed by merely recalling information, because it will give you novel problems that require reasoning with the material on a deeper level.

Also, explicit test prep—where you basically teach strategies for cheating the test—universally sucks, but presumably that's not what the study is measuring.