The validity of placement testing in freshman mathematics at the University of Tennessee, Knoxville

Methods of determining the best entry level mathematics course for college students vary greatly. Most commonly, test scores determine placement. Either a test designed specifically for placement or the score on some achievement test, such as the ACT or SAT test, is used. Historically, results have been less than satisfying , characterized by low correlations between predictor variables and student success rates.

This research focuses on finding a collection of predictor variables which can be used to determine the proper placement of college students into their entry level mathematics course. Several sections of various freshman level mathematics courses at the University of Tennessee, Knoxville (UTK), were selected at random for participation in the study. Students in those classes were given two placement examinations published by the Mathematical Association of America during the first week of classes in the fall quarter of 1987. Other pertinent data, including the high school from which each subject graduated, the number of credits in high school mathematics, mathematics score on the ACT test, and high school grade point average, were collected from the Office of Student Data Analysis at UTK.

We found low correlations between predictors and measures of student success. Prediction equations were obtained which accounted for as little as 20% and as much as 50% of the variance in grade. In addition, we found that a certain amount of sex bias appears to exist in the placement tests used; this bias results in low placement for females although they outperform males in the same mathematics courses.

We recommend the use of several predictor variables in determining placement rather than relying on one placement test score. Finally, we feel that the use of different cut-off scores to determine placement for males and females should be investigated.