ACT Research & Policy | Technical Brief | July 2019
As shown in Figure 1, retesters performed better
in college than what was expected based on
their test scores. And this prediction error was
minimized when superscores were used, as
compared to the other scoring methods. If
superscores reflected positive measurement
error—that is, an overestimate of one’s true
achievement level—then superscores would
predict students to earn higher grades in college
than what they actually earned, and this
overprediction would increase as the number of
retests increases. However, the results of the
study suggested exactly the opposite.
Why is this the case? One hypothesis is that
superscores and number of retests reflect not only
academic preparation but also a motivational
component. Specifically, students who are willing
to forgo multiple Saturdays to sit for a multiple-
hour test with the hope of maybe increasing their
score are also the students who are likely to ask
questions in their college courses, visit their
professor during office hours, and take advantage
of any extra credit opportunities to ensure the best
possible grade. Future research should explore
these hypotheses.
Another contribution of this study is the
evaluation of the diversity implications of
employing one scoring method versus another.
Interestingly, despite the fact that underserved
students are less likely to retest (Harmston &
Crouse, 2016), the superscoring method did not
result in a less diverse admitted class as
compared to the other three methods. In fact,
the gender, racial, and parental income
distributions of a simulated admitted class were
identical across the four scoring methods.
Current Study
The focus of the current study is to extend the
previous research with an emphasis on further
exploring the diversity implications of
superscoring. As mentioned above, underserved
students are less likely to retest as compared to
their affluent peers. For students who test only
once, superscoring has no impact on their ACT
Composite score. Only students who retest have
the potential to increase their ACT Composite
score through superscoring, and the magnitude of
this difference should be related to the number of
times the student retests, in general. With that in
mind, one potential concern or unintended
consequence of superscoring is that subgroup
differences will be exacerbated under this scoring
policy. The focus of the current study is to
investigate the extent to which superscoring
increases, decreases, or has no impact on
subgroup differences.
Method
Using data on the 2018 ACT-tested graduating
class, we compared the average ACT
Composite score for various student subgroups
based on their most recent ACT Composite
score as well as a superscore ACT Composite
score. We estimated subgroup performance
differences in terms of both:
1. Mean differences or unstandardized
differences (USTD): the difference between
the mean value in two groups
2. Standardized differences (STD): the difference
between the mean value in two groups,
divided by the overall standard deviation.
We estimated performance differences by the
following student characteristics: race/ethnicity,
gender, household income, and parental
education. These characteristics were self-
reported by students at the time they registered
to take the ACT.
Results
Retesting Rates. Samples sizes and retesting
rates for the 2018 ACT-tested graduating high
school class are summarized in Table 1.
Results are presented for the overall sample and
by the student subgroups of interest. The total
group consisted of over 1.9 million students. Of
those students, 44% took the ACT more than
once. As previously documented, we find that
minority students and students from lower socio-
economic households are less likely to retest.