Thursday, June 23, 2011

New Study Questions Validity of Two Parts of ACT

This from Ed Week:
The validity of the ACT in predicting college success has come under scrutiny in a new paper out by the National Bureau of Economic Research.
The study, Improving College Performance and Retention the Easy Way: Unpacking the ACT Exam, by Eric P. Bettinger, Brent J. Evans, and Devin G. Pope, suggests that two of the four sub tests of the ACT, English and mathematics, are highly predictive of positive college outcomes, while the other two, science and reading, provide little or no additional predictive power.
Colleges rely on the ACT exam in their admission decisions to increase their ability to differentiate between students likely to succeed and those that have a high risk of under-performing and dropping out. We show that two of the four sub tests of the ACT, English and Mathematics, are highly predictive of positive college outcomes while the other two subtests, Science and Reading, provide little or no additional predictive power. This result is robust across various samples, specifications, and outcome measures. We demonstrate that focusing solely on the English and Mathematics test scores greatly enhances the predictive validity of the ACT exam.
Officials at the ACT, based in Iowa City, Iowa, maintain the test is supported by decades of research.

1 comment:

Skip Kifer said...

This is a study done by economists who seem to have little knowledge of the ACT or how it has been used historically. The prediction models they emphasize do not include the strongest predictor - high school record, although it is entered into their equations after they separate the subtests. There notion of "greatly enhances" is an arguably small statistical difference. ACT could have great answers to these assertions.