Testing Kids and Predicting Future Success
by John E. Bonfadini, Ed.D.,
Contributing ColumnistProfessor,
George Mason University

Standardized tests are often used to predict how students will perform in the future. Colleges use the Scholastic Aptitude Test (SAT) to predict which graduating high school students will succeed in the higher education environment. Few parents and educators realize what the true effectiveness of standardized testing is in determining the success of students in college. This article is an attempt to briefly explain the relationship among standardized tests and other factors used in the selection process.

Colleges could use one measure to predict college success, but common sense says that additional factors should be considered. Most colleges use a combination of the SAT and the student’s grade point average (GPA). The statistical procedure called a correlation is done between the students’ high school grades or their SAT scores and their freshman college grades. The results are expressed in a coefficient, which can range from a minus 1 to a positive 1. The scale looks like this (-1…..0…..+1). The higher the coefficient, the stronger the correlation; direction can be either positive or negative.

The correlation between college grades and SAT scores is approximately r = .42. This value indicates a moderate positive relationship. In other words, students who have higher SAT scores tend to do acceptable work in college. To determine how much of college success can be predicted by SAT scores, a statistical procedure called regression is used. The regression procedure calls for squaring the correlation coefficient value (.42 x .42 = .18 or 18 %). The SAT explains 18% of the variance in college freshman grades. The relationship looks something like this:

To further try to predict college success, colleges will also consider a student’s grade point average (GPA). The correlation between GPA and freshman success is r = .48. The squared value is approximately .24 or 24%. Twenty-four percent of a student’s college success can be attributed to the GPA. The relationship looks like this:

Which of the two pictures most accurately represents what happens when we try to use both the SAT and the GPA to predict college success?

The one on the right is the correct answer. The diagram on the left assumes that we can use all the predictability of the SAT as well as all the predictability of the high school GPA. Some of what contributes to a student’s GPA also contributes to the SAT score. There is a redundancy, and we lose some predicting power of the second variable, which is the SAT. The following graph illustrates this concept:

High school GPA is the strongest predictor so we assign it the overlap area. Penn State University did a study to determine what effect adding a third predictor, class rank, to GPA and SAT would accomplish. The study found that class rank only explained an additional 1% of the unexplained variance. In other words, knowing the class ranking contributed little in predicting success in college.

Studies have found that substituting teacher recommendations for SAT is just as effective for predicting college freshman success. Colleges justify using the SAT because it’s less subjective and universal in scope. One study found that student study habits and attitudes in the eighth grade were more effective than the SAT in predicting college success. Tests are usually good predictors of how people will do on other tests, but when it comes to predicting how people will function in the whole educational environment they are somewhat less effective. The general public has been led to believe that the higher a student scores on a test, the more successful their college life. Nothing could be further from the truth. The above charts show that over 66% of college success is attributable to other factors not measured with the SAT or GPA.

A Basketball Metaphor

I’ll use a metaphor to explain the SAT selection process. In the selection of a basketball player for your team, if you chose free-throw shooting as the major criteria for making the team, then you would never pick Shaquille O’Neal. Most big men don’t make a high percentage of free throws. A player’s ability to shoot free throws is easy to measure and standardized when compared to other, more complex basketball skills. We all know it takes more than making free throws to be successful in basketball. Running, jumping, aggressiveness, quickness, strength, and leadership are just as important as free throw shooting. Taken individually or collectively, they are more difficult to measure because they are usually done in combination. Foul shooting is done in a more sterile environment, which only minimally reflects the other basketball traits.

Students in Virginia usually take the Stanford 9 Test and the new SOL (Standards of Learning) tests. Students who do well on the Stanford 9 usually do the best on the SOL tests. Also, there is a very high correlation between other standardized reading tests and student performance on the SOL tests. The correlation coefficient for reading tests and the SOL tests is approximately r = .9. This extremely high relationship indicates the student’s reading score can explain the 81% of the variance in the SOL tests. The other 19% is attributed to other factors. Students who read the best will score the highest on the SOL tests. The SOL tests will do little to improve the predicting of college success, because the student’s GPA and SAT scores already include factors measured by the SOL tests.

Success Depends on Many Factors