The ACT: Biased, Inaccurate, and Misused

What is the ACT?

More than a million high school students take the ACT (formerly known as the American College Testing Program Assessment) each year. Like the SAT, the ACT is a standardized multiple-choice test meant to predict first-year college grades. While the SAT predominates on the East and West Coasts, the ACT is more common in the Midwest, Southwest, and Deep South.

The ACT consists of four individual tests: English, Math, Reading, and Science Reasoning. The score report for the “Enhanced ACT Assessment” includes a composite ranging from 1-36, a score for each individual test, and sub scores. There is also an optional “writing” test.

The ACT was developed as an alternative to the SAT, but is just a different test, not a better one. Like the SAT, the ACT has long-standing problems of bias, inaccuracy, coachability, and misuse. Because of these flaws, no test – neither the ACT nor SAT – should be required in the college admissions process.

How is the ACT biased?

Race, class and gender biases give White, affluent, and male test-takers an unfair edge.
ACT scores are directly related to family income: the richer students’ parents are, the higher are average scores. But score gaps between groups on the ACT cannot be explained away solely by differences in educational opportunity linked to social class. According to ACT research, when all factors are equal, such as course work, grades and family income, Whites still outscore all other groups. If the ACT were not biased, Asian Americans, who take more academic courses than any other group, would likely score even higher. Moreover, boys score slightly higher than girls across all races, despite boys’ lower grades in high school and college when matched for identical courses.

ACT has performed few studies of score differences in its test, making it difficult to pinpoint the sources of the score gaps. Here are a few likely candidates:

Biased format: Research shows that a fast-paced, multiple-choice format favors males over females. Guessing, a risk males are more likely to take, is rewarded. Since multiple-choice items do not allow for shades of meaning they work against the most typical female thinking style.

Biased language: Idiomatic terms such as “ball and chain” (to indicate a married partner) and “straight from the horse’s mouth” may not be familiar to many test-takers, particularly those whose first language isn’t English, causing them to choose wrong answers.

How accurate is the ACT?

ACT scores do not predict college performance effectively
Even the test-maker admits that high school grades predict first-year college grades better than ACT scores do. In fact, adding the ACT to the high school record does not significantly improve predictions.

One study at Chicago State University confirmed this trend. For the vast majority of the university’s graduates who scored in the middle range of the test as high school students, the ACT explained only 3.6% of the differences in cumulative college GPA. In fact, the exam over-predicted the performance of the class graduating in 1992, which had the highest average ACT score among the classes in the research study yet the poorest academic performance over four years at the university.

The ACT regularly underestimates the abilities of females, who earn higher grades than males in college, despite lower ACT scores. Recognizing the problem, the Massachusetts Institute of Technology routinely admits females with lower math scores because they find the women still perform as well as men.

The ACT also does a poor job of predicting the college performance for students of color. One study conducted at a medium-sized regional university in the Southeast showed that the ACT explained only 6.8% of the differences in first-semester college grades for African Americans, while for Whites the ACT predicted approximately 28% of the differences. High school grades predicted equally well for both groups, demonstrating that other measures of achievement are not as subject to differences across racial groups as are test scores.

ACT scores are imprecise
The individual tests have large margins of error, according to data from ACT. The margin of error – the inconsistency in ACT scores inherent in the testing process – on each subject’s 1-36 point scale is 1.55 points in English, 1.43 in Mathematics, 2.20 in Reading, and 1.75 in Science Reasoning. In other words, if a student were to retake the exam, there would be about a two-thirds chance that her score would be 1.55 points higher or lower on the English test than on a previous administration of the test. There is also a one-third chance the score difference would be even larger. The margins of error, while appearing to be small at 1.43 – 2.20, can actually have significant consequences for applicants when admissions offices or financial aid programs require minimum (or “cut-off”) scores.
Is the ACT coachable? ACT itself sells coaching products. ACT insists that “for students who have not studied the content or grasped it… short term review is not likely to be of much benefit.” While ACT acknowledges that familiarity and test-taking skills can affect a score, it also maintains that the descriptions and sample items included in the registration packet address them. If that claim is true, why is ACT selling coaching guides and software? Many commercial firms promote extensive lines of ACT test preparation materials, ranging from workbooks to classroom and on-line courses.
How is the ACT misused?

Cut-off scores on the ACT unfairly deny education and scholarships.
The ACT’s flaws have serious consequences. Despite its inaccuracies, biases, and coachability, ACT cut-off scores are often used to determine entrance into schools and allocate scholarships. A single point can decide whether a student is admitted or receives needed funds. Though these misuses violate ACT guidelines for proper test use, the test-maker has done nothing to stop them. ACT has the responsibility and the power to protect students from testing abuse by refusing to send scores to colleges, scholarship agencies, and educational systems which misuse their product.

Education denied
In Mississippi, ACT test scores have long been used as a tool of segregation in state colleges and universities. In 1962 Mississippi’s higher education board instituted a minimum ACT requirement for admission that was eight points above the average Black student’s score in the state. According to a federal court, this was done “soon after the court ordered admission of James Meredith to the University of Mississippi because it deterred black enrollment.” The U.S. Supreme Court agreed, specifically citing the role of the ACT cut-off as a key factor in denying African Americans access to the state’s major universities. Nonetheless, the state still relies on ACT cut-offs to determine eligibility for admissions and lucrative scholarships.

Scholarships denied
In Louisiana, high achieving students can have their tuition and fees waived for up to four years at state colleges and universities if they score above an ACT cut-off of 20. This minimum score is close to the average for White students in Louisiana, but more than three points above the African American average. Several other state scholarship programs also employ cut scores, including Florida’s “Bright Futures” and South Carolina’s “Palmetto Fellows” programs.

What’s the alternative? The weak predictive power of the ACT, its susceptibility to coaching, examples of test score misuse, and the negative impact test score use has on educational equity all lead to the same conclusion – test scores should be optional in college admissions. The nearly 400 colleges and universities that already admit substantial numbers of freshman applicants without regard to test scores have shown that class rank, high school grades, and rigor of classes taken are better tools for predicting college success than any standardized test. The SAT and SAT Subject Tests are often viewed as alternatives to the ACT, yet these exams have many of the same flaws and limitations as the ACT.

Approximately 1.3 million test takers


Female 21.0 0.0 Male 21.2 0.0 ETHNICITY African-American or Black 17.0 -0.1 American Indian or Alaskan Native 18.9 +0.1 Asian-American or Pacific Islander 22.6 +0.3 Caucasian-American, White 22.1 +0.1 Hispanic 18.7 +0.1 Other/No Response (14%) 21.6 +0.5 ALL TEST-TAKERS 21.2 +0.1

source: ACT