UGA Findings Support Controversial SAT Writing Test

The first independent, academic study of the redesigned SAT college entrance exam finds that the new writing portion is a much better predictor of academic success than the verbal and math portions of the exam.

Three economists in the University of Georgia Terry College of Business analyzed data from more than 4,300 test takers and, unlike a recent study published by the group that administers the test, accounted for factors – such as level of parental education and the high school the student attended – that strongly influence success in college.

The writing section of the test was introduced in March 2005, and the researchers note that the lack of data on its effectiveness has led nearly half of the nation’s colleges and universities to disregard the scores. David Mustard, the associate professor of economics who co-authored the study with professor Christopher Cornwell and student Jessica Van Parys, said the team’s finding suggests that schools shouldn’t ignore SAT writing scores.

“Schools who don’t use the writing portion of the SAT are foregoing an opportunity to choose a class of students that will score higher GPAs, enroll in more credit hours, will be less likely to withdraw from classes and are likely to do well in a whole array of different variables,” Mustard said. “They’re really throwing out information that will help them choose a more qualified class.”

The researchers found that with each 100-point increase in SAT writing scores, first-year students:

Earn GPAs that are, on average, .07 points higher;
Earn .18 points higher in freshman English classes; and
Earn .54 more credit hours.
The SAT verbal (now known as the critical-reading section) was also a significant predictor of collegiate success, but not nearly as powerful as the writing section. With each 100-point increase on the SATV, students earned freshman GPAs that were .03 points higher, less than one half the .07 increase for the SATW.

Cornwell points out that the significance of SATV scores diminishes when the writing score is taken into account.

“Statistically speaking, the verbal section doesn’t add much predictive value beyond the writing section,” Cornwell said.

Still, the researchers said it is too soon to eliminate the verbal portion of the test. “The writing section could be a better measure of academic ability than the verbal,” Van Parys said, “but another explanation is that it’s so new that students haven’t yet learned to game the test.”

The researchers found high school GPA is still a much stronger predictor of collegiate success than any individual or combined sections of the SAT, but note that the test is clearly useful in differentiating between students who have equal GPAs.

The team examined data from incoming students at the University of Georgia in 2006, the year in which the first cohort of students to take the redesigned test enrolled. They said their results are likely to apply to other large, selective public universities.

The UGA study comes on the heels of one conducted by The College Board, the group that administers the SAT. The College Board study used a larger sample, Mustard said, but only took into account broad institutional characteristics like size and selectivity and whether the college or university was public or private. The College Board study only looked at the predictive value of SAT scores with regard to one variable – freshman GPA – while the UGA study looked at several outcome variables in addition to freshman GPA, including grades in freshman English and math classes, credit hours enrolled, credit hours earned and the probability that a student will lose Georgia’s merit-based HOPE scholarship. Notably, the researchers found that predictive effect of the SATW is higher than the SATV in all of the outcomes, excluding those related to mathematics.

The UGA study also controlled for personal variables, such as the number of advanced placement classes taken in high school, and family characteristics, such as the level of parental education. These statistical controls are critical, the researchers note, because such factors have a significant influence on collegiate success. Controlling for the high school the student attended, which takes into account the quality of the school, community demographic variables and the effects of peers, increases the explanatory power of their model by 50 percent, for example.

“There are so many variables that influence success in college that simply correlating SAT scores with first-year GPAs doesn’t tell you much about the predictive value of the test,” Mustard said. “By holding constant a range of factors, we were essentially able to take a person that has the same family characteristics, education, high school GPA, race, ethnicity and gender and then see how well variations in SAT scores predict success in college. Despite what critics might say, the test scores really do matter.”