Predicting Long-Term College Success through Degree Completion Using ACT Composite Score, ACT Benchmarks, and High School Grade Point Average

*050205120* Rev 1

Predicting Long-Term

College Success through

Degree Completion Using

ACT

Composite Score, ACT

Benchmarks, and High School

Grade Point Average

Justine Radunzel

Julie Noble

August 2012

ACT Research

Report Series

2012 (5)

For additional copies, write:

ACT Research Report Series

P.O. Box 168

Iowa City, IA 52243-0168

Predicting Long-Term College Success through Degree

Completion Using ACT

Composite Score, ACT

Benchmarks, and High School Grade Point Average

Justine Radunzel

Julie Noble

Abstract

This study compared the effectiveness of ACT

Composite score and high school grade

point average (HSGPA) for predicting long-term college success. Outcomes included annual

progress towards a degree (based on cumulative credit-bearing hours earned), degree completion,

and cumulative grade point average (GPA) at 150% of normal time to degree completion (year 6

and year 3 for four- and two-year institutions, respectively). The utility of the individual ACT

College Readiness Benchmarks for predicting long-term college success was also evaluated.

Data for this study included over 190,000 ACT-tested students who enrolled in college as

first-time entering students in fall, 2000 through 2006. Over 100 total two- and four-year

institutions were represented. Hierarchical logistic models were used to estimate institution-

specific probabilities of college success based on ACT scores and HSGPA. First-year college

GPA was also included as a predictor in the path analysis models. Accuracy and success rates

were calculated using the distributions of ACT scores and HSGPA for each institution’s

approximate applicant pool; rates were then summarized across institutions. Direct and indirect

effects of ACT score, HSGPA, and first-year academic performance on subsequent college

outcomes were also examined. Results were disaggregated by institution type.

Both ACT Composite score and HSGPA were effective for predicting long-term college

success at both four- and two-year institutions. Across the outcomes, test scores increased

prediction accuracy over that for HSGPA alone. ACT scores and HSGPA were primarily

indirectly related to subsequent college outcomes (through first-year college GPA). The ACT

Benchmarks were also found to be useful for predicting long-term college success, providing

further validity evidence for using them as measures of college readiness.

iii

Acknowledgments

The authors thank Jeff Allen, Jill Crouse, and Richard Sawyer for their helpful comments and

suggestions on an earlier draft of this report.

Predicting Long-Term College Success through Degree Completion Using ACT Composite

Score, ACT Benchmarks, and High School Grade Point Average

Introduction

There are several student measures that are typically considered in the admissions

process, largely because they have been found to identify accurately students who are ready for

college and to predict students’ eventual success in college. The top four measures identified by

four-year postsecondary institutions (Clinedinst, Hurley, & Hawkins, 2011) are academically

related and include grades in college preparatory courses, strength of high school curriculum,

standardized test scores (ACT or SAT), and high school grade point average (HSGPA). But,

many institutions also use other non-academic measures (e.g., extra-curricular activities and

demonstrated interest in the institution) in making admission decisions. They do this to meet

other goals that are not directly related to academic success but that closely align with their

educational mission, such as maintaining equal opportunity and diversity in student enrollments.

While most two-year institutions currently practice open admissions, about one-fifth of them use

standardized test scores or HSGPA as part of their admission process (Breland, Maxey, Gernand,

Cumming, & Trapani, 2002).

In this report, we focus on the use of pre-enrollment achievement measures to identify

students who are likely to be successful in college, recognizing that this addresses only one

aspect of the admission process. In order to evaluate the effectiveness of these measures, the

outcome(s) of interest need to be identified. For making admission decisions, one outcome that is

commonly used is first-year academic performance, as measured by first-year college grade

point average (GPA). But, as pressure for increased accountability in higher education and

higher graduation rates continues, institutions are considering outcomes beyond the first year of

college, including persistence, academic performance, and degree completion. For example, a

recent report from the Higher Education Research Institute (2011) refers institutions to a degree

completion calculator available on their website that calculates expected degree completion rates

based on student characteristics of their incoming freshman class. Another study (Saupe & Curs,

2008) discussed a procedure for developing four enrollment management scores, one of which

was a graduation score intended to predict whether a student would graduate within six years of

enrolling.

Two-year institutions also feel the pressure to increase graduation rates. Due to the

reduced resources available to them, some are having to prioritize access; restrict enrollment;

eliminate lower-level, remedial courses; and identify students who are likely to graduate or

transfer to a four-year institution (González, 2012). In addition, they are being encouraged to

evaluate intermediate outcomes that measure progress towards degree completion to help

determine the reasons why so many students are not completing degrees (Moore, Shulock, &

Offenstein, 2009).

Historically, numerous studies have consistently demonstrated that ACT scores and

HSGPA are valid measures of early college success, defined as first-year college GPA and/or

first to second year retention (Allen & Robbins, 2010; Allen, Robbins, Casillas, & Oh, 2008;

Noble & Radunzel, 2007; Robbins, Allen, Casillas, Peterson, & Le, 2006). In a recent study of

50 four-year institutions, Westrick, Le, Robbins, Radunzel, and Schmidt (2012) found that the

estimated mean correlation with first-year college GPA across institutions was 0.51 for ACT

Composite (ACTC) score and 0.58 for HSGPA, after adjusting for range restriction. Sackett,

Kuncel, Arneson, Cooper, & Waters (2009) found a similar correlation between SAT scores and

first-year college GPA, after controlling for socioeconomic status. Sawyer (2010) reported that

the multiple correlations of high school subject-area grade averages and ACT scores with first-

year college GPA were higher when scores and grades were used jointly than when they were

used separately.

Several studies have gone beyond examining correlations to evaluate the estimated

effects of using these two measures for making admission decisions, with first-year college GPA

as the outcome under consideration. For example, two studies (Sawyer, 2010; Noble & Sawyer,

2002) found that HSGPA was slightly more accurate (as measured by the estimated percentage

of correct classifications) for predicting first-year success at GPA thresholds of 2.00, 2.50, and

3.00, and the ACTC score was slightly more accurate for predicting success at thresholds of 3.50

and 3.75. Across the different college GPA thresholds, using ACTC score and HSGPA in

combination resulted in greater prediction accuracy, and was more effective for identifying

successful students among those who would be expected to be successful, relative to using them

separately. This latter finding demonstrates the incremental validity of test scores for predicting

first-year academic performance. Sawyer (2010) also pointed out that HSGPA was a much

stronger predictor of first-year GPA among students with higher ACTC scores than among those

with lower scores. A similar result also held for ACTC score among students with higher

HSGPAs.

The ACT College Readiness Benchmarks (in English, mathematics, reading, and science)

have also been shown to be predictive of early college success. The Benchmarks are the

minimum ACT test scores required for students to have a high probability of success in first-

year, credit-bearing college courses—English Composition, College Algebra, social sciences

courses, and Biology (Allen & Sconing, 2005), and provide an empirical definition for college

readiness. Students who meet the Benchmark have approximately a 50% chance of earning a B

or better and approximately a 75% chance of earning a C or better in the corresponding college

course or courses (ACT, 2010a). The Benchmarks were identified as the typical scores across

both two- and four-year postsecondary institutions that maximized the accuracy for predicting

success (defined as earning a grade of a B or higher) in the corresponding courses. Meeting the

Benchmarks has also been shown to be positively associated with early college outcomes, such

as immediately enrolling in college the fall following high school graduation, persisting to the

second year at the same institution, and achieving a 2.00 or higher, or 3.00 or higher, first-year

college GPA (ACT, 2010b). Students who meet the Benchmarks are also less likely than those

who do not meet the Benchmarks to take remedial coursework in English or mathematics.

A recent study (Radunzel & Noble, 2012) found that students who met the individual ACT

Benchmarks were substantially more likely than those who did not meet the Benchmarks to

persist in college through degree completion and to earn a degree in a timely manner. Moreover,

as the number of ACT Benchmarks met increased, students’ likelihood of achieving these

outcomes also increased. The study also found that students with higher ACTC scores had higher

success rates than those with lower scores; a similar result held for HSGPA. These findings were

seen for students attending four-year institutions, as well as for those attending two-year

institutions.

In a review of the literature, Moore and Shulock (2009) cited several studies documenting

that first-year college GPA is also predictive of degree completion. These studies suggested that

pre-college measures, such as standardized test scores and HSGPA, appear to influence degree

completion primarily by virtue of their effect on first-year college academic performance.

In this study, we extend prior research on the topic of using ACT scores and HSGPA for

making admission decisions by focusing on long-term college outcomes through degree

completion and applying the same methodology used in Sawyer (2010) and Noble and Sawyer

(2002). In particular, in this study we investigate

 the maximum accuracy of ACTC score and HSGPA used alone and jointly for predicting

long-term college success. We also estimate the percentages of students who would be

successful from among those who are expected to be successful (selected).

 the usefulness of the ACT College Readiness Benchmarks in each of the subject areas for

predicting long-term college success, thus providing further validity evidence for using

them as measures of college readiness. The percentages of successful students based on

those with scores at or above the Benchmarks are compared to those associated with

ACTC or HSGPA values that maximize the percentage of correct classifications.

 the utility of ACTC score and/or HSGPA for predicting long-term college success, given

first-year college GPA. For this objective, we estimate the direct, indirect, and total

effects of ACTC score, HSGPA, and first-year academic performance on subsequent

college outcomes.

Because a majority of two-year institutions have open admissions policies, prior studies of

this nature have included the results for four-year institutions only. But, in light of the growing

concerns for open access remaining the norm at two-year institutions (González, 2012), this

research topic is relevant and timely. Therefore, as an initial look, we also examine these issues

for two-year institutions. Though some of the long-term outcomes differ between the two types

of institutions (e.g., degree types), we compare their results at common achievement levels to

evaluate the utility of ACTC score, HSGPA, and the ACT Benchmarks as college readiness

indicators and for predicting long-term college success for all students, regardless of whether

they initially apply to a two- or four-year institution.

Data

Data for this study included approximately 194,000 ACT-tested students who enrolled in

college as first-time entering students in fall, 2000 through 2006. Over 100 institutions were

represented, including all public institutions from two state systems. Four-year institutions

included in the study were required to have at least six years of follow-up data available on their

students so that six-year bachelor’s degree completion rates could be evaluated for a given

cohort. Two-year institutions were required to have at least three years of follow-up data

available so that three-year associate’s degree completion rates could be evaluated for a given

cohort.

Postsecondary institutions make admission decisions about applicants. Therefore, to

study the usefulness of using ACTC score, HSGPA, and the ACT Benchmark scores for

informing college admission decisions, we also included over 505,000 students who sent their

ACT scores to study institutions during the same time frame but did not enroll there.

Nonenrolled students who sent scores to an institution, plus those who actually enrolled in an

institution, comprised the “applicant pool” for that institution. The applicant pools for the

institutions in this study approximate actual applicant pools. Students may send their ACT scores

to any number of institutions, but actually apply to only a subset of them. Conversely, some

students may apply to some institutions without submitting official ACT score reports. The

analyses in this report are based on data from all score senders; they are considered to be proxies

for “applicants.”

College outcomes included annual progress to degree (based on cumulative hours

earned), degree completion, and cumulative GPA at 150% of normal time to degree completion

Nonenrolled students were identified from the 2000 to 2006 ACT-tested high school graduate histories. These

students requested that their ACT scores be sent to at least one of the 104 institutions included in this study during

the same time period as that for enrolled students.

(at the end of year 6 for four-year institutions and the end of year 3 for two-year institutions).

Analyses were done separately by institution type, where type was defined at time of initial

enrollment. However, some institutions provided data for some but not all of the outcomes. As a

result, the number of institutions and enrolled students with available data differed by college

outcome (Table 1). Multiple freshman cohorts of students from an institution were combined

together in the analyses (cohorts 2000 to 2003 for four-year institutions and cohorts 2000 to 2006

for two-year institutions).

Table 1

Sample Sizes for Total Group of Students and Enrolled Students with Available College

Outcomes by Type of Institution

Outcome variable Two-year institutions Four-year institutions

Total

Number of institutions 43 61

Number of enrolled students 67,816 125,911

Number of nonenrolled students 84,407 420,629

Number of students in applicant pool 152,223 546,540

Cumulative hours earned

Number of institutions

Number of enrolled students

62,407

111,691

Cumulative GPA

Number of institutions

Number of enrolled students

28,868

68,662

Degree completion

Number of institutions

Number of enrolled students

67,816

125,911

Degree completion plus transfer

Number of institutions

Number of enrolled students

66,129

Note. Sample sizes by college outcome are for enrolled students with available data on the outcome and provide

counts of the numbers of students used to estimate the hierarchical logistic models. Slight fluctuations in the

numbers of enrolled students and numbers of institutions by outcome are due to missing data for individual students

or entire institutions.

Cumulative GPA was evaluated at year 3 for two-year institutions and at year 6 for four-year institutions.

Progress to degree was based on cumulative credit-bearing hours earned at the end of

each spring term, and measured whether the student was making progress towards degree

completion. For dropouts and stopouts, the last value for cumulative hours earned was carried

forward. For four-year institutions, end-of-year cumulative hours thresholds were 24, 48, 72, and

96 earned credit hours for years 1, 2, 3, and 4, respectively, approximating bachelor’s degree

completion in about five years. For two-year institutions, end-of-year cumulative hours

thresholds were 18, 36, and 54 credit hours earned for years 1, 2, and 3, respectively,

approximating associate’s degree completion in slightly over three years.

For degree completion, we evaluated earning a bachelor’s degree within six years of

initial enrollment at a four-year institution and earning an associate’s degree within three years at

a two-year institution. In addition, for two-year institutions from two state systems, we evaluated

associate’s degree completion or transfer to an in-state four-year institution within three years of

initially enrolling in college. Given the data sources available for this study, we focused on

degree completion from the initial institution.

Cumulative GPA was evaluated at the end of year 6 for four-year institutions and at the

end of year 3 for two-year institutions (referred to in this report as the year 6/year 3 cumulative

GPA). Cumulative GPAs at degree completion were included in these analyses for students

graduating before the end of year 6 for four-year institutions or year 3 for two-year institutions.

Cumulative GPAs were initially evaluated at the following levels: 2.50 or higher, 3.00 or higher,

3.25 or higher, 3.50 or higher, and 3.75 or higher. However, results for the 2.50 or higher college

criterion were not included in this report because very few students had year 6/year 3 cumulative

GPAs below 2.50.

The sample for the study does not represent students or institutions nationally. A large

majority of both the two- and four-year institutions came from the North Central accrediting

region (Table 2). Moreover, most of the four-year institutions and all of the two-year institutions

were public institutions.

The four-year institutions varied in admissions selectivity, though the

majority (75%) had traditional or selective admissions policies.

Table 2

Percent of Two- and Four-Year Institutions by Institutional Characteristic

Institutional characteristic

Two-year institutions

(n = 43)

Four-year institutions

(n = 61)

Affiliation

Public

Private

Selectivity

Selective/highly selective

Traditional

Liberal/open

Unknown

Accrediting region

North Central

Southern

Northwestern

Middle States

Unknown

Locale

Urban

Suburban

Small city

Small town

Note. Percentages may not sum to 100 percent due to rounding.

The pre-enrollment measures used in this study included ACTC score, HSGPA, and the

ACT College Readiness Benchmarks. The ACT Composite score is the rounded arithmetic

In fall 2003, approximately three-fourths and nearly one-half of four- and two-year institutions in the United

States, respectively, were private institutions (Knapp, Kelly-Reid, Whitmore, Wu, Gallego, Cong, Berzofsky, Huh,

Levine, & Broyles, 2005).

average of the four subject area scores (English, Mathematics, Reading, and Science). Test

scores are reported on a scale of 1 to 36. If students took the ACT more than once, only the most

recent results were used. HSGPA was based on student’s self-reported coursework taken in 23

specific courses in English, mathematics, social studies, and science and the grades earned in

these courses. The ACT College Readiness Benchmarks correspond to scores of 18, 22, 21, and

24 on the ACT English, Mathematics, Reading, and Science tests, respectively (Allen &

Sconing, 2005).

Method

For each institution, mean ACTC scores and HSGPAs, as well as the corresponding

standard deviations (SDs) were computed for enrolled students and the entire applicant pool.

Mean cumulative GPAs and success rates were calculated by institution for enrolled students.

Distributions of the means and rates of these variables were then summarized across institutions

using minimum, median, and maximum values.

Hierarchical logistic models estimated progress to degree, cumulative GPA, and degree

completion rates for enrolled students from the pre-enrollment measures. Hierarchical models

account for students clustered within institutions and allow the estimated college outcome

success rates to vary across institutions. Separate models were developed by year for each

relevant outcome and by institution type (two- vs. four-year). In all cases, we estimated random

slope and intercept models.

Models were estimated for predicting college success based on (a) ACTC score, (b)

HSGPA, (c) ACTC score and HSGPA used jointly, and (d) individual ACT subject area scores.

The ACTC score and HSGPA joint model was evaluated with and without the interaction

between these two measures. Nearly all of the interaction terms were statistically significant at

the 0.01 level.

Clearly, a student’s likelihood of being successful in college is based on multiple

predictors, including cognitive and non-cognitive factors, as well as sociodemographic factors

(Allen & Robbins, 2010). ACT does not advocate making college success predictions solely on

the basis of a single measure, such as a test score. The use in this paper of one or two predictors

is a mathematical simplification. The methods used here, such as those used with the ACTC and

HSGPA joint model, could be generalized to multiple measures. The usefulness of these two

measures for predicting long-term success is evaluated from the perspective of accurately

distinguishing students who are likely to be successful from those who are not. In particular, the

methodology used here is based on statistical decision theory (Sawyer, 1996) for validating

educational selection decisions; the method frames validity evidence in terms of probable

outcomes, given the ACT score or HSGPA and the outcome criteria used. The methodology used

is the same as that used by ACT for helping institutions make course placement decisions.

For each predictor (or predictor combination) at institution-specific values we estimated

three decision-based statistics for making admission decisions:

1. the maximum percentage of correct classifications (maximum accuracy rate (maxAR)),

2. the percentage of successful students among those who would be expected to be

successful (success rate (SR)), and

3. the increase in the percentage of correct classifications over expecting all applicants to be

successful (increase in accuracy rate (∆AR)).

The latter two statistics were evaluated at the institution-specific predictor value that maximized the

percentage of correct classifications. In this report, we refer to the predictor value associated with

the maxAR as the “selection value.”

Correct classifications include students at or above a given predictor (selection) value who

were successful and students below the value who would have not been successful. For predictors

that are positively related to success, it can be shown that the predictor value that maximizes the

percentage of correct classifications corresponds to a 0.50 probability of success for a given model.

For the two-predictor model, multiple combinations of ACTC score and HSGPA corresponding to a

probability of success of 0.50 were identified. Probability distributions that cross 0.50 will yield

accuracy rate distributions that increase to a maximum and then decrease. If the probability

distribution for an institution does not cross 0.50, the maxAR is generally not interpretable, and the

model is therefore considered a “nonviable model” for an institution. Models for institutions with

probability curves crossing 0.50 are referred to here as “viable models.”

If there were no selection procedure (i.e., if all students were selected, regardless of their

ACTC score and/or HSGPA), a certain percentage of them would be successful. This percentage

is referred to as the “baseline” accuracy rate. The arithmetic difference between the maxAR and

the baseline accuracy rate represents the increase in accuracy rate (∆AR) that results from using

test scores or HSGPA. Large ∆ARs correspond to a greater contribution by the pre-enrollment

measures in increasing the percentage of correct classifications.

MaxARs, SRs, and ARs were calculated for each institution with a viable model. These

statistics were generated using the institution-specific parameter estimates from the hierarchical

models and the distributions of ACT scores and HSGPA for each institution’s applicant pool.

The institution-specific estimated conditional probabilities of success for nonenrolled students were assumed to be

the same as those for enrolled students.

Distributions of these statistics were then summarized across institutions using minimum,

median, and maximum values. Results across institutions with viable models for each individual

predictor/outcome combination are presented in this paper.

Results across institutions with

viable models for both predictors were similar to these. For comparison purposes, the median

percentages of students with scores below the selection values associated with the maxAR were

also reported. (Note: 100 minus this percentage gives the percentage of students in the applicant

pool at or above the selection value). For a more complete description of the methodology used

(including the assumptions being made) to evaluate the usefulness of pre-enrollment measures in

the admissions process, see the full ACT research report by Sawyer (2010).

To study the utility of the ACT College Readiness Benchmark scores for predicting long-

term college success, SRs were estimated at the Benchmark scores for each institution, regardless of

whether or not the probability curve for the institution crossed 0.50. Increases in SRs (denoted by

SRs) were also estimated to evaluate the usefulness of the predictor variable for increasing SRs

over baseline success rates. Distributions of SRs and SRs were summarized across institutions

using minimum, median, and maximum values. Median SRs at each ACT Benchmark were

compared to those at the institution-specific ACTC or HSGPA values associated with the maxAR

(the latter based on institutions with viable models only).

Path analysis was used to estimate the effects of ACTC score, HSGPA, and first-year

college GPA jointly on subsequent college outcomes for enrolled students. First-year GPA was the

cumulative GPA from the end of the spring term of the student’s first year of college. Figure 1

shows the hypothesized path model for each college outcome.

Across college outcomes, median decision-based statistics were comparable for the joint models with and without

the interaction. Therefore, median statistics based on the model with the interaction are presented in this report.

Figure 1. Path model evaluated. ACTC = ACT Composite; HSGPA = high school grade point

average; GPA = grade point average.

The path model included two regression models, as well as the correlation between

ACTC score and HSGPA. The first model regressed first-year college GPA on ACTC score and

HSGPA. The second model regressed the college outcome of interest (e.g., degree completion)

on ACTC score, HSGPA, and first-year GPA. Hierarchical linear regression was used to estimate

the models for continuous outcomes and hierarchical logistic regression was used for binary

outcomes (the only college outcome that was not binary in these analyses was year 6/year 3

cumulative GPA). The fixed effects coefficients from the hierarchical models were standardized

for comparative purposes. The method described by MacKinnon and Dwyer (1993) and Jasti,

Dudley, and Goldwater (2008) for standardizing coefficients was applied when the outcome

variable was binary.

An indirect effect for a specific path is found by taking the product of the standardized

coefficients from each of the regressions that comprise the path. The total indirect effect of a

predictor is found by summing all indirect effects across the various possible paths. Paths

connecting two correlated variables are included as indirect paths. Adding the total indirect effect

and the direct effect gives the total effect of the predictor.

HSGPA

First-year

GPA

College

success

ACTC

score

Some students omitted responses to high school coursework and grade items when they

completed the ACT registration materials. Multiple imputation was used to estimate missing

values; 12% of enrolled students and 11% of nonenrolled students had missing HSGPA. Five

data sets were imputed. Models were developed for all five imputed data sets; no differences in

parameter estimates (including standard errors) of practical significance were found across the

data sets. The results reported here for all analyses involving HSGPA are those based on the

initial imputed data set.

Results

Effectiveness of ACTC Score and HSGPA for Predicting Long-Term College Success

In this section, we describe the incremental benefit of using ACTC score and HSGPA

jointly for predicting college success through degree completion. We first present descriptive

statistics for ACTC scores and HSGPAs for enrolled students and the entire applicant pool, as

well as for college outcomes over time for enrolled students. Next, we present probability

distributions for the various college outcomes as functions of ACTC scores and HSGPAs.

Following this, we present maxARs, ∆ARs, and SRs for ACTC score and HSGPA used

separately and jointly to predict long-term college success.

Descriptive statistics. At four-year institutions, mean ACTC scores and HSGPAs were

typically higher among enrolled students than among students in the entire applicant pool (Table

3). The corresponding standard deviations were slightly smaller for enrolled students. The typical

mean ACTC score across four-year institutions in this study (21.5) was lower than the mean

score (22.6) of first-year ACT-tested college students nationally who enrolled in four-year

institutions in 2003 (ACT, 2004).

Table 3

Distributions of Means and Standard Deviations of ACTC Scores, HSGPAs, College Success

Rates, and College GPAs by Applicant/Enrollment Status across Four-Year Institutions

Enrollment

status

Predictor/outcome

variable

Number of students Mean SD

Med Min/Max Med Min/Max Med Min/Max

Applicant

pool

ACTC score 6,692 159/41,628 20.4 16.3/23.0 4.1 3.1/4.7

HSGPA 6,692 159/41,628 3.21 2.81/3.50 0.58 0.45/0.64

Enrolled

students

ACTC score 1,287 50/9,824 21.5 16.1/25.3 3.9 3.1/4.7

HSGPA 1,287 50/9,824 3.32 2.82/3.73 0.53 0.29/0.64

Progress year 1 1,541 50/9,824 68 27/89

Progress year 2 1,541 50/9,824 54 21/83

Progress year 3 1,529 49/9,824 46 18/79

Progress year 4 1,526 49/9,824 44 17/79

Bachelor’s degree 1,287 50/9,824 42 17/79

First-year GPA 1,170 44/9,225 2.80 2.35/3.18 0.81 0.54/0.96

Year 6 cum GPA

612 24/6,286 3.12 2.77/3.47 0.52 0.38/0.75

Note. SD = standard deviation; Med = median; Min = minimum; Max = maximum; ACTC = ACT Composite;

HSGPA = high school grade point average.

Student’s cumulative GPAs at degree completion were included in year 6 GPA analyses for students who

graduated with a bachelor’s degree before the end of year 6.

Among enrolled students, the typical six-year bachelor’s degree completion rate across

four-year institutions was 42% and ranged from 17% to 79%. Median progress to degree rates

declined over time from 68% by the end of year 1 (24 or more cumulative hours earned) to 44%

by the end of year 4 (96 or more cumulative hours earned). The typical year 6 college cumulative

GPA was above 3.00 (median = 3.12).

At two-year institutions, mean ACTC scores and HSGPAs were comparable between

enrolled students and the applicant pool (Table 4). The typical mean ACTC score of students

enrolled in two-year institutions in this study (18.3) was slightly lower than the mean score

(18.8) of first-year ACT-tested college students nationally who enrolled in two-year institutions

in 2003 (ACT, 2004).

Table 4

Distributions of Means and Standard Deviations of ACTC Scores, HSGPAs, College Success

Rates, and College GPAs by Applicant/Enrollment Status across Two-Year Institutions

Enrollment

status

Predictor/outcome

variable

Number of students Mean SD

Med Min/Max Med Min/Max Med Min/Max

Applicant

pool

ACTC score 2,137 120/16,472 18.3 16.7/20.2 3.6 3.2/4.1

HSGPA 2,137 120/16,472 3.04 2.81/3.21 0.59 0.51/0.62

Enrolled

students

ACTC score 834 95/9,551 18.3 16.9/20.6 3.5 3.0/4.0

HSGPA 834 95/9,551 3.02 2.79/3.25 0.58 0.49/0.62

Progress year 1 832 79/8,804 50 18/77

Progress year 2 831 95/8,866 40 8/61

Progress year 3 830 95/8,808 34 4/54

Associate’s

degree

834 95/9,551 14 4/34

Associate’s

degree or transfer

1,053 157/9,551 23 7/41

First-year GPA 385 65/7,321 2.63 2.18/3.01 0.90 0.76/1.10

Year 3 cum GPA

343 25/4,729 2.81 2.55/3.08 0.71 0.53/1.00

Note. SD = standard deviation; Med = median; Min = minimum; Max = maximum; ACTC = ACT Composite;

HSGPA = high school grade point average.

Student’s cumulative GPAs at degree completion were included in year 3 GPA analyses for students who

graduated with an associate’s degree before the end of year 3.

Among enrolled students, the typical three-year associate’s degree completion rate for

two-year institutions was relatively low at 14% and ranged from 4% to 34% across institutions.

The typical rate for completing an associate’s degree or transferring to an in-state four-year

institution by year 3 was higher at 23% and ranged from 7% to 41% across institutions. Progress

to degree rates for two-year institutions also declined over time from 50% by the end of year 1

(18 or more cumulative hours earned) to 34% by the end of year 3 (54 or more cumulative hours

earned). The typical year 3 college cumulative GPA was less than 3.00 (median = 2.81).

Hierarchical logistic regression results. Figures A-1 to A-10 in Appendix A provide

estimated probabilities of completing a degree, progressing towards a degree, or achieving

different levels of year6/year3 cumulative GPA as a function of ACTC score or HSGPA. The

probabilities in the figures were estimated using the fixed effects parameter estimates from the

hierarchical logistic models. Across college outcomes, as ACTC score or HSGPA increased, the

estimated probabilities of success at either a typical two- or four-year institution increased.

Degree completion. At four-year institutions, the probability of earning a bachelor’s

degree by year 6 for students with an ACTC score of 25 (the maximum average ACTC score

across institutions) was substantially higher than that for students with an ACTC score of 16 (the

minimum average ACTC score across institutions; 0.54 vs. 0.31, see Figure A-1). At two-year

institutions, the chances of earning an associate’s degree or transferring to an in-state four-year

institution by year 3 were greater than those of earning an associate’s degree by year 3 by at least

10 percentage points for students with ACTC scores of 19 or higher or HSGPAs of 3.3 or higher

(Figures A-1 and A-2). Regardless of the student’s HSGPA, the probabilities of degree

completion at two-year institutions (including or not including transfer) were less than 0.50.

Progress to degree. The estimated probabilities of progressing towards a degree

associated with a given ACTC score or HSGPA decreased over time with the largest decline in

probabilities occurring between years 1 and 2 (Figures A-3 and A-4 for four-year institutions and

Figures A-5 and A-6 for two-year institutions). For example, at a typical four-year institution, the

estimated probability of progressing towards a degree for a student with an ACTC score of 21

was 0.68 at the end of year 1 (earned 24 or more hours), and decreased to 0.54 (48 or more

hours), 0.48 (72 or more hours), and 0.45 (96 or more hours) by the end of years 2, 3, and 4,

respectively (Figure A-3).

The corresponding probabilities for a student with a HSGPA of 3.20

were estimated to be 0.64, 0.50, 0.44, and 0.41, respectively (Figure A-4). Estimated progress-to-

degree probabilities also declined over time for two-year institutions (Figures A-5 and A-6).

Year 6 cumulative GPA at four-year institutions. The chances of earning a 3.75 or

higher year 6 cumulative GPA were at least 30 percentage points lower than the chances of

earning a 3.00 or higher GPA for students with ACTC scores of 16 to 32 (Figure A-7) or

HSGPAs of 2.80 or higher (Figure A-8). For example, an ACTC score of 21 corresponded to a

0.58 estimated probability of earning a 3.00 or higher year 6 cumulative GPA at a typical four-

year institution (Figure A-7). The corresponding probabilities for the other criterion levels were

0.38 (3.25), 0.20 (3.50), and 0.07 (3.75), respectively.

A HSGPA of 3.20 was associated with a 0.52 estimated probability of achieving a 3.00 or

higher year 6 cumulative GPA at a typical four-year institution (Figure A-8). The corresponding

probabilities for the other criterion levels were 0.30 (3.25), 0.13 (3.50), and 0.03 (3.75),

respectively. For the 3.75 criterion, a HSGPA of 4.00 corresponded to a typical probability of

success of only 0.33.

ACTC score of 21 and HSGPA of 3.20 correspond to the approximate median average ACTC score and HSGPA

across four-year institutions.

Year 3 cumulative GPA at two-year institutions. Probability curves for two-year

institutions were comparable to those for four-year institutions. An ACTC score of 18

corresponded to a 0.42 estimated probability of earning a 3.00 or higher year 3 cumulative GPA

at a typical two-year institution (Figure A-9).

The corresponding probabilities for the other

criterion levels were 0.26 (3.25), 0.14 (3.50), and 0.05 (3.75), respectively.

A HSGPA of 3.00 was associated with an estimated probability of 0.39 of achieving a

3.00 or higher year 3 cumulative GPA at a two-year institution (Figure A-10). The corresponding

probabilities for the other criterion levels were 0.24 (3.25), 0.12 (3.50), and 0.04 (3.75),

respectively. For the criterion levels of 3.50 or higher and 3.75 or higher, a HSGPA of 4.00

corresponded to a typical probability of success of less than 0.50 at two-year institutions (0.45

and 0.26, respectively).

Accuracy and success rates for ACTC score and HSGPA. In this section, we

summarize median maxARs, ∆ARs, and SRs across institutions with viable models using ACTC

score or HSGPA as predictors of long-term college outcomes.

Minimum and maximum values

for these statistics are provided in Appendix B, Tables B-1 to B-4. Results for progress to degree

outcomes are also provided in the tables in Appendix B.

Degree completion. For four-year institutions, 58 of the 61 institutions had viable ACTC

models and 56 had viable HSGPA models for predicting degree completion (Table 5).

ACTC score of 18 and HSGPA of 3.00 correspond to the approximate median average ACTC score and HSGPA

across two-year institutions.

For a viable model, the probability distribution must cross 0.50.

Table 5

Median Results for Predicting Degree Completion

Predictor variable

Number

institutions

Maximum

accuracy

rate

(maxAR)

Increase in

accuracy

rate

(∆AR)

Success

rate

(SR)

Bachelor’s degree completion by year 6 -- four-year institutions (n = 61)

ACTC 58 64 24 57

HSGPA 56 65 23 58

ACTC & HSGPA

61 67 26 60

Associate’s degree completion by year 3 -- two-year institutions (n = 43)

ACTC 25 81 63 55

HSGPA 5 72 43 52

ACTC & HSGPA

25 82 64 53

Associate’s degree completion or transfer to four-year institution by year 3 -

- two-year institutions (n = 40)

ACTC 38 77 54 55

HSGPA 14 69 37 54

ACTC & HSGPA

38 77 54 55

Note. Results across institutions with viable models for both predictors were similar to those presented here.

Statistics were evaluated at selection values associated with the maxAR. ACTC = ACT Composite; HSGPA = high

school grade point average.

The three institutions with nonviable ACTC models had relatively high six-year bachelor’s

degree completion rates (66% to 79%). The five institutions with nonviable HSGPA models had

relatively low six-year bachelor’s degree completion rates (17% to 25%). Joint ACTC and

HSGPA models were viable for all 61 four-year institutions included in this study.

The median ACTC and HSGPA selection values associated with the maxAR for

predicting six-year bachelor’s degree completion were relatively high (25 for ACTC and 3.57 for

HSGPA; Appendix B, Table B-1), and these selection values varied substantially across

institutions (ranging from 9 to 31 for ACTC and from 2.20 to 4.00 for HSGPA). Multiple

combinations of ACTC score and HSGPA corresponded to a probability of 0.50 for the joint

model and are therefore not listed in the tables in Appendix B. Median maxARs and SRs were

comparable for ACTC and HSGPA (64% vs. 65% and 57% vs. 58%, respectively). The median

maxAR and SR for the joint model were higher than those based on the single-predictor models

(by 2 to 3 percentage points). The typical maxAR associated with using both ACTC score and

HSGPA jointly for predicting bachelor’s degree completion was 26 percentage points higher

than the baseline AR. These findings demonstrate the incremental benefit of using ACTC score

and HSGPA for predicting bachelor’s degree completion by year 6.

To better understand the results based on the joint model and the incremental usefulness

of ACTC score beyond HSGPA, Figure 2 provides the estimated probabilities of completing a

bachelor’s degree by year 6 associated with different values of HSGPA and ACTC scores. As

both HSGPA and ACTC score increased, probabilities of success also increased. The ACTC

score differential was larger for students with higher HSGPAs than those with lower HSGPAs.

The same was true for the HSGPA differential when comparing students with higher and lower

ACTC scores.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0

HSGPA

Probability

ACTC = 35

ACTC = 30

ACTC = 25

ACTC = 20

ACTC = 15

ACTC = 10

Figure 2. Estimated probabilities of bachelor’s degree completion by year 6 based on HSGPA

and ACTC score at four-year institutions.

HSGPA = high school grade point average; ACTC =

ACT Composite.

For associate’s degree completion by year 3, 25 of the 43 two-year institutions had viable

ACTC models and only five had viable HSGPA models (three-year associate’s degree

completion rates were relatively low). As a result, selection values associated with the maxAR

were relatively high across the institutions with viable models (Appendix B, Table B-2); these

values were higher than those for predicting bachelor’s degree completion by year 6 at four-year

institutions. Across the 25 institutions with viable models, the typical maxAR and SR associated

with ACTC scores were relatively high (81% and 55%, respectively), while the percentages of

students at or above the ACTC selection values were relatively low (ranged from about 1% to

8% across institutions; Appendix B, Table B-2).

The probabilities in Figure 2 were estimated using the fixed effects parameter estimates from the hierarchical

logistic model that included an interaction term between ACTC score and HSGPA.

When examining associate’s degree completion or four-year transfer by year 3 as the

outcome, 38 of 40 two-year institutions had viable ACTC models, but only 14 of the 40

institutions had viable HSGPA models (the typical probability of associate’s degree completion

or transfer by year 3 for students with a 4.00 HSGPA was less than 0.50; Figure A-2). The

median selection values associated with the maxAR were slightly higher than those for

bachelor’s degree completion by year 6 at four-year institutions (27 vs. 25 for ACTC score and

3.75 vs. 3.57 for HSGPA). The median maxAR and ∆AR across two-year institutions based on

ACTC score were both higher than those associated with predicting bachelor’s degree

completion across four-year institutions. The median SR, however, was comparable. The median

maxAR for HSGPA viable models was considerably lower than that based on ACTC viable

models. Median maxARs, ∆ARs, and SRs for the ACTC and HSGPA joint models were

comparable to those for the ACTC models, reflecting the negligible incremental benefit of

HSGPA for predicting associate’s degree completion by year 3 alone or in combination with

transferring to an in-state four-year institution.

Progress to degree over time. Over 85% of the two- and four-year institutions with

cumulative hours earned available had viable models for evaluating the progress to degree

outcomes (44 out of 50 four-year institutions and 36 out of 42 two-year institutions; Appendix B,

Tables B-1 and B2). Median selection values for predicting progress to degree from either ACTC

or HSGPA increased over time (Appendix B, Tables B-1 and B-2). For example, for four-year

institutions, the median ACTC score associated with the maxAR increased from 18 at year 1 to

24 at year 4, compared to a score of 25 for predicting completion of a bachelor’s degree by year

6 (usually requiring 120 or more cumulative credit hours earned). For two-year institutions, the

typical score associated with the maxAR for completing an associate’s degree by year 3 was

much higher than that for predicting progress to degree at year 3 (29 vs. 23; Table B-2).

Median maxARs, ∆ARs, and SRs for HSGPA were slightly higher than those for ACTC

for four-year institutions, but these median values were more comparable between these two

predictors for two-year institutions. However, at both two- and four-year institutions, the median

maxARs, ∆ARs, and SRs based on the joint models were slightly higher than those based on the

single-predictor models (by 1 to 3 percentage points). These findings are consistent with those

seen for degree completion.

Year 6 cumulative GPA at four-year institutions. All but one of the 57 four-year

institutions with cumulative GPA data available had viable ACTC models for predicting college

GPA at year 6 for criterion levels at or above 3.00 (Table 6). In contrast, only 40 and 2 of the 57

institutions had viable models for HSGPA for the 3.50 and 3.75 criterion levels, respectively (the

typical probability of a 3.75 or higher GPA for students with a 4.00 HSGPA was less than 0.50;

Figure A-8). For predicting college GPA at year 6, median ACTC scores and HSGPAs

associated with the maxAR increased across GPA criterion levels from 3.00 to 3.75 (Table B-3

of Appendix B). For example, the median ACTC selection value for a GPA level of 3.00 or

higher was 20; the corresponding selection values for the other criterion levels were 24, 27, and

31, respectively. For both ACTC score and HSGPA, selection values associated with the maxAR

varied substantially across institutions (see minimum and maximum values in Table B-3 of

Appendix B).

Table 6

Median Results for Predicting Levels of Year 6 College Cumulative GPA at Four-Year

Institutions

Predictor variable

Number

institutions

Maximum

accuracy

rate

(maxAR)

Increase in

accuracy

rate

(∆AR)

Success

rate

(SR)

3.00 or higher college GPA

ACTC 57 67 12 68

HSGPA 57 70 16 70

ACTC & HSGPA

57 71 18 71

3.25 or higher college GPA

ACTC 57 70 33 64

HSGPA 57 73 35 63

ACTC & HSGPA 57 75 39 67

3.50 or higher college GPA

ACTC 57 79 56 61

HSGPA 40 79 56 56

ACTC & HSGPA 57 83 62 63

3.75 or higher college GPA

ACTC 56 91 82 58

HSGPA 2 84 67 50

ACTC & HSGPA 57 92 84 59

Note. Students’ cumulative GPAs at degree completion were included in year 6 GPA analyses for students who

graduated with a bachelor’s degree before the end of year 6.

Statistics were evaluated at selection values associated

with the maxAR. GPA = grade point average; ACTC = ACT Composite; HSGPA = high school grade point average.

The HSGPA models were slightly more accurate than the ACTC models for criterion

levels of 3.00 and 3.25 (based on the median maxARs and ∆ARs), but the HSGPA and ACTC

models were comparable for the criterion level of 3.50. For the GPA criterion level of 3.75, the

typical maxAR and ∆AR were relatively high for ACTC score (91% and 82%, respectively).

Across the GPA criterion levels (at or above 3.00), the median maxARs, ∆ARs, and SRs for the

ACTC and HSGPA joint model generally exceeded those for both single-predictor models.

Figure 3 contains the estimated probabilities for achieving a year 6 cumulative GPA of 3.00 or

higher at a typical four-year institution based on different values of HSGPA and ACTC score.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0

HSGPA

Probability

ACTC = 35

ACTC = 30

ACTC = 25

ACTC = 20

ACTC = 15

ACTC = 10

Figure 3. Estimated probabilities of achieving a year 6 college cumulative GPA of 3.00 or higher

based on HSGPA and ACTC score at four-year institutions.

HSGPA = high school grade point

average; ACTC = ACT Composite.

ACTC differentials in estimated probabilities were greater for students with higher

HSGPAs than for those with lower HSGPAs. Larger HSGPA differentials were also seen for

students with higher ACTC scores than for those with lower scores. Similar results held for the

other GPA criterion levels.

Year 3 cumulative GPA at two-year institutions. Only 5 of the 42 two-year institutions

had viable HSGPA models for predicting year 3 college GPA at the 3.50 criterion level, and

none of the institutions had viable HSGPA models for the 3.75 criterion level (Table 7). In

contrast, all but two of the institutions had viable ACTC models across GPA criterion levels.

The probabilities in Figure 3 were estimated using the fixed effects parameter estimates from the hierarchical

logistic model that included an interaction between ACTC score and HSGPA.

Table 7

Median Results for Predicting Levels of Year 3 College Cumulative GPA at Two-Year

Institutions

Predictor variable

Number

institutions

Maximum

accuracy

rate

(maxAR)

Increase in

accuracy

rate

(∆AR)

Success

rate

(SR)

3.00 or higher college GPA

ACTC 42 66 22 63

HSGPA 42 68 25 62

ACTC & HSGPA 42 70 27 66

3.25 or higher college GPA

ACTC 42 74 46 62

HSGPA 42 75 48 56

ACTC & HSGPA 42 77 50 64

3.50 or higher college GPA

ACTC 42 84 66 60

HSGPA 5 78 55 51

ACTC & HSGPA 42 85 68 61

3.75 or higher college GPA

ACTC 40 92 84 58

HSGPA 0

ACTC & HSGPA 41 93 85 58

Note. Students’ cumulative GPAs at degree completion were included in year 3 GPA analyses for students who

graduated with an associate’s degree before the end of year 3. Statistics were evaluated at selection values associated

with the maxAR.

GPA = grade point average; ACTC = ACT Composite; HSGPA = high school grade point average.

Similar to the results seen for four-year institutions, median selection values associated

with the maxAR increased across GPA criterion levels for two-year institutions (Appendix B,

Table B-4). The ACTC selection scores for two-year institutions were generally within 2 score

points of those for four-year institutions, and the HSGPA selection values for two-year

institutions were generally higher than those for four-year institutions. Selection values

associated with the maxAR also varied substantially across institutions (Appendix B, Table B-4).

Median maxARs, ∆ARs, and SRs for the ACTC and HSGPA joint models tended to be slightly

higher than those for both single-predictor models for GPA criterion levels of 3.00 or higher. In

general, typical SRs for two-year institutions were lower than those for four-year institutions for

GPA criterion levels of 3.00 and 3.25.

Usefulness of ACT College Readiness Benchmarks for Predicting Long-Term College

Success

In this section, we evaluate the effectiveness of the ACT College Readiness Benchmarks

for predicting college success through degree completion. We first present descriptive statistics

on ACT Benchmark attainment for enrolled students, as well as for the entire applicant pool. We

also briefly describe the probability distributions for college success as functions of the

individual ACT subject area scores. Following this, we present the typical probabilities of

success, SRs, and ∆SRs associated with the ACT College Readiness Benchmark scores.

Descriptive statistics. At four-year institutions, the typical percentages of students

meeting the ACT Benchmarks were higher among enrolled students than among students in the

applicant pool (Table 8). In contrast, at two-year institutions, the typical Benchmark attainment

percentages were comparable for these two student groups, and were lower than those for

students from four-year institutions.

Table 8

Distributions of Percentages of Students Meeting ACT Benchmarks across Institutions by

Applicant/Enrollment Status and Type of Institution

Institution

type ACT Benchmark

Applicant pool Enrolled students

Med Min/Max Med Min/Max

Four-year

English 70 31/84 78 31/96

Mathematics 31 5/58 38 5/85

Reading 51 16/69 59 15/83

Science 21 3/44 26 3/56

Two-year

English 52 38/67 53 38/72

Mathematics 14 7/38 13 5/38

Reading 33 20/48 33 19/50

Science 10 3/20 9 2/23

Note. For the typical numbers of students per institution see those reported in Tables 3 and

4. Med = median; Min = minimum; Max = maximum.

Across college outcomes for both institution types, as the ACT subject area score

increased, the typical probabilities of college success also increased. This finding is illustrated in

Figure 4 for six-year bachelor’s degree completion at four-year institutions.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACT subject area score

Probability

English

Mathematics

Reading

Science

Figure 4. Probability of completing a bachelor’s degree by year 6 as a function of ACT subject

area score.

The logistic curve associated with the ACT Mathematics score was steeper than those

associated with the other subject area tests (that is, there were greater differences in the

probabilities of completing a bachelor’s degree by year 6 between students with higher and lower

scores in mathematics than in the other subject areas). Across the outcomes, the typical

probabilities of success at the ACT Mathematics (22) and Science (24) Benchmark scores were

higher than those at the ACT English (18) and Reading (21) Benchmark scores. This finding is

further illustrated in the next section where the typical probabilities of success (across

institutions) at the Benchmark scores are provided.

Success rates for ACT College Readiness Benchmarks. Estimated SRs and ∆SRs

associated with ACT Benchmark scores across all institutions with available outcome data were

The probabilities in Figure 4 are based on fixed effects parameter estimates from the hierarchical logistic models.

calculated. For these analyses, cumulative GPA was evaluated for only two criterion levels: 3.00

or higher and 3.50 or higher.

Degree completion. For four-year institutions, median probabilities of completing a

bachelor’s degree by year 6 ranged from 0.35 to 0.47 across the Benchmarks (Table 9). For two-

year institutions, median probabilities of success ranged from 0.12 to 0.23 across the

Benchmarks for predicting associate’s degree completion, and ranged from 0.21 to 0.36 for

predicting associate’s degree completion or transfer to an in-state four-year institution by year 3.

Table 9

Median Results for Predicting Degree Completion by ACT College Readiness Benchmarks

Subject area

Probability

of success at

Benchmark

Success

rate

(SR)

Increase in

success rate

(∆SR)

Bachelor’s degree completion by year 6 (n = 61 four-year

institutions)

English 0.35 46 5

Mathematics 0.46 53 13

Reading 0.39 47 5

Science 0.47 52 11

Associate’s degree completion by year 3 (n = 43 two-year

institutions)

English 0.12 18 4

Mathematics 0.22 28 14

Reading 0.15 19 5

Science 0.23 26 12

Associate’s degree completion or transfer to four-year

institution by year 3 (n = 40 two-year institutions)

English 0.21 29 6

Mathematics 0.36 42 19

Reading 0.25 30 7

Science 0.35 38 16

Note. For the typical percentages of students at or above the ACT Benchmark scores, see Table 8.

The probabilities of degree completion for each Benchmark varied substantially across

institutions (Appendix C, Tables C-1 and C-2). For example, the estimated probability of

completing a bachelor’s degree by year 6 at the ACT Mathematics Benchmark score ranged from

0.23 to 0.79 across institutions, further illustrating the high variability in institutional degree

completion rates and admission requirements across institutions.

For both two- and four-year institutions, typical SRs associated with the ACT Benchmark

scores were higher than baseline SRs as evidenced by the positive ∆SRs (ranging from 4 to 19

percentage points). Median SRs and ∆SRs associated with the ACT Mathematics and Science

Benchmarks were higher than those associated with the ACT English and Reading Benchmarks

(by as little as 5 percentage points to as much as 13 percentage points; larger differences were

seen for two-year institutions). To help provide context for these SRs, median SRs for the

Mathematics Benchmark score were only slightly lower than those corresponding to ACTC or

HSGPA values that maximized prediction accuracy for bachelor’s degree completion by year 6

for four-year institutions (within 4 to 5 percentage points; see Tables 5 and 9). A similar result

was not seen for the two degree completion outcomes at two-year institutions: the typical SRs

associated with the Mathematics Benchmark were considerably lower than those corresponding

to ACTC scores that maximized prediction accuracy (by at least 13 percentage points).

This

latter finding is a result of the institution-specific associate’s degree completion rates being

relatively low (ranged from 4% to 34% when transfers were not included and from 7% to 41%

when transfers were included).

The SRs corresponding to Benchmark scores were based on all 61 four-year institutions and 40 two-year

institutions, while SRs based on institution-specific ACTC selection scores (those associated with the maxAR) were

based on 58 four-year institutions and 25 or 38 two-year institutions with viable ACTC models. For ACTC score we

also evaluated SRs across all institutions at specific scores. But, the conclusions were similar to those already noted

for results based on the institution-specific selection scores.

Progress to degree over time. The typical probabilities of success and SRs associated

with using the ACT Benchmarks to predict progress towards a degree decreased over time

(Tables C-1 for four-year institutions and C-2 for two-year institutions). For example, for four-

year institutions, the typical chances of completing 24 or more credit hours by the end of year 1

for students with an ACT Mathematics score of 22 was 73%. For students with the same ACT

Mathematics score, the typical chances of completing 96 or more credit hours by the end of year

4 decreased to 51%. And, the corresponding median SRs decreased from 82% to 57% from year

1 to year 4. These findings agree with an earlier result where the median ACTC score and

HSGPA values associated with the maxAR for predicting progress to degree increased over time

(see Tables B-1 and B-2).

Across time points and types of institutions, the highest median SRs and ∆SRs for the

progress to degree outcomes were generally associated with the ACT Mathematics Benchmark,

but typical ∆SRs were positive for all of the Benchmarks demonstrating the incremental value of

these indicators over baseline SRs. The median SRs for the ACT Mathematics and Science

Benchmarks were higher than those corresponding to the institution-specific ACTC selection

values for the first two years, and were comparable for later years (compare Table B-1 to Table

C-1 and Table B-2 to Table C-2). Typical ∆SRs associated with the ACT Mathematics and

Science Benchmark scores for predicting the progress to degree outcomes ranged from 12 to 16

percentage points at four-year institutions and from 19 to 27 percentage points at two-year

institutions. The median SRs for the ACT English and Reading Benchmarks were comparable to

those associated with the institution-specific ACTC selection values for earlier years, but were

slightly lower for later years.

Year 6 cumulative GPA at four-year institutions. The typical probabilities of achieving a

year 6 college GPA of 3.00 or higher ranged from 0.46 to 0.72 across the Benchmarks (Table

10). The corresponding probabilities for the 3.50 criterion ranged from 0.13 to 0.34. Probabilities

of success varied substantially across institutions (Table C-3).

Table 10

Median Results for Predicting Levels of Year 6/Year 3 College Cumulative GPAs using ACT

College Readiness Benchmarks, by Institution Type

ACT

Benchmark

Four-year institutions Two-year institutions

Probability

of success

Benchmark

Success

rate

(SR)

Increase in

success

rate

(∆SR)

Probability

of success

Benchmark

Success

rate

(SR)

Increase

in success

rate

(∆SR)

3.00 or higher college GPA

English 0.46 65 10 0.42 55 13

Mathematics 0.66 76 19 0.63 70 26

Reading 0.59 69 12 0.50 60 16

Science 0.72 77 19 0.63 68 24

3.50 or higher college GPA

English 0.13 29 7 0.14 25 8

Mathematics 0.28 40 18 0.30 39 23

Reading 0.22 34 10 0.21 29 12

Science 0.34 43 19 0.34 41 23

Note. ACT Benchmarks are 18 for English, 22 for Mathematics, 21 for Reading, and 24 for Science. Cumulative

GPA was evaluated at year 6 for four-year institutions and at year 3 for two-year institutions. Students’ cumulative

GPAs at degree completion were included in GPA analyses for students at four-year institutions who graduated with

a bachelor’s degree before the end of year 6 or students at two-year institutions who graduated with an associate’s

degree before the end of year 3.

Median SRs were substantially higher and ∆SRs were slightly higher for the 3.00 or

higher criterion than for the 3.50 or higher criterion. Typical probabilities of success, SRs, and

∆SRs associated with the ACT Mathematics and Science Benchmark scores were higher than

those associated with the ACT English and Reading Benchmark scores at both GPA criterion

levels. Median SRs for the Mathematics and Science Benchmarks were also slightly higher than

those associated with the institution-specific ACTC and HSGPA selection values that maximized

prediction accuracy for the 3.00 or higher criterion, but were substantial lower for the 3.50 or

higher criterion (see Tables 6 and 10). The latter finding is due to the typical ACTC selection

value being considerably higher than the ACT Benchmark scores (by at least 3 scale score

points). Moreover, a relatively small percentage of students achieved a 3.50 or higher year 6

cumulative GPA (median across four-year institutions was 27%).

Year 3 cumulative GPA at two-year institutions. Findings noted for year 6 cumulative

GPA at four-year institutions generally held true for year 3 cumulative GPA at two-year

institutions. Typical probabilities of success associated with the ACT Benchmarks were lower

for two-year institutions than for four-year institutions for the 3.00 or higher criterion, but were

somewhat more comparable for the 3.50 or higher criterion (Table 10). Similar to those noted for

four-year institutions, probabilities of success and SRs varied across the institutions (Table C-4).

Median SRs were lower for two-year institutions than for four-year institutions, but the opposite

was true for ∆SRs. Differences in typical SRs and ∆SRs between the two types of institutions

were larger for the 3.00 or higher criterion than for the 3.50 or higher criterion.

Path Analysis

In this section we investigate the effects of ACTC score, HSGPA, and first-year college

GPA jointly for predicting subsequent college outcomes. The indirect effects of ACTC scores

and HSGPA on college outcomes mediated through first-year college GPA, as well as their

direct effects on college outcomes, are estimated (see Figure 1). Only enrolled students with

college GPAs at the end of the first year were included in the analyses (83% and 62% of the

samples for four- and two-year institutions, respectively).

Degree completion at four-year institutions. The path model for bachelor’s degree

completion by year 6 for four-year institutions is shown in Figure 5.

Figure 5. Path model for bachelor’s degree completion by year 6 at four-year institutions.

ACTC = ACT Composite; HSGPA = high school grade point average; GPA = grade point

average; Bach = Bachelor’s.

ACTC score was found to be only indirectly related to bachelor’s degree completion; the

direct path was not statistically significant at the 0.01 level. The direct effect of first-year GPA

on bachelor’s degree completion was over 7 times that for HSGPA. The paths from ACTC score

and HSGPA on first-year GPA were both significant. The total effect of HSGPA on degree

completion (direct and indirect) was slightly greater than that for ACTC score (0.31 vs. 0.26).

The total effect on bachelor’s degree completion for each of the pre-enrollment measures was

lower than the direct/total effect of first-year GPA (Table D-1 from Appendix D).

Degree completion at two-year institutions. The path model for associate’s degree

completion by year 3 at two-year institutions is shown in Figure 6, and the path model for

associate’s degree completion or transfer to four-year institution by year 3 is shown in Figure 7.

HSGPA

First-year

GPA

Bach degree

completion

ACTC

score

0.52

0.07

0.48

0.35

0.26

Figure 6. Path model for associate’s degree completion by year 3 at two-year institutions.

ACTC = ACT Composite; HSGPA = high school grade point average; GPA = grade point

average; Assoc = Associate’s.

Figure 7. Path model for associate’s degree completion or transfer to an in-state four-year

institution by year 3 at two-year institutions. ACTC = ACT Composite; HSGPA = high school

grade point average; GPA = grade point average; Assoc = Associate’s.

The direct effect of first-year GPA on each outcome was greater than the direct effects of

ACTC score or HSGPA. The total effect of HSGPA on degree completion was greater than that

of ACTC score (0.27 vs. 0.36 when not including transfers and 0.29 vs. 0.34 when including

transfers; Table D-2 from Appendix D). They were each smaller than the direct/total effect of

first-year GPA on these two outcomes.

HSGPA

First-year

GPA

Assoc degree

or transfer

ACTC

score

0.44

0.13

0.46

0.30

0.18

0.09

HSGPA

First-year

GPA

Assoc degree

completion

ACTC

score

0.46

0.16

0.46

0.29

0.18

0.06

Progress to degree. For four-year institutions, similar to that seen for degree completion,

ACTC score was only indirectly related to the progress to degree outcomes. The significant

direct effects of HSGPA on these outcomes were relatively small in comparison to those for

first-year GPA (0.07 vs. 0.58 to 0.62; see Table D-1 in Appendix D). For two-year institutions,

the direct effects of ACTC score and HSGPA on the progress to degree outcomes were

comparable and relatively small in comparison to those of first-year GPA on these outcomes (see

Table D-2 in Appendix D). Across these outcomes over time, the total effects of each of the pre-

enrollment measures (direct and indirect) were smaller than the direct/total effects of first-year

GPA on the progress to degree outcomes.

Year 6 cumulative GPA at four-year institutions. The direct effects of ACTC score

and HSGPA on year 6 cumulative GPA were somewhat comparable, but were substantially

smaller than the direct effect of first-year GPA on the outcome (Figure 8).

Figure 8. Path model for year 6 cumulative GPA at four-year institutions. Analyses included

only those students who either graduated prior to year 6 or were still enrolled at the end of year

6. ACTC = ACT Composite; HSGPA = high school grade point average; GPA = grade point

average; cum = cumulative.

For students who either graduated prior to year 6 or were still enrolled at year 6, the

direct effects on first-year GPA, as well as the total effects on year 6 cumulative GPA, were

HSGPA

First-year

GPA

Year 6

cum GPA

ACTC

score

0.68

0.12

0.45

0.32

0.31

0.09

comparable for ACTC score and HSGPA (Table D-1 in Appendix D). The total effects for both

pre-enrollment measures were smaller than the direct/total effect of first-year GPA on this

outcome.

Year 3 cumulative GPA at two-year institutions. Path model results for year 3

cumulative GPA at two-year institutions were similar to those for year 6 cumulative GPA at

four-year institutions (Figure 9). The one exception was that the direct effect of HSGPA on first-

year GPA was greater than the direct effect of ACTC score on the same outcome.

Figure 9. Path model for year 3 cumulative GPA at two-year institutions. Analyses included only

those students who either graduated prior to year 3 or were still enrolled at the end of year 3.

ACTC = ACT Composite; HSGPA = high school grade point average; GPA = grade point

average; cum = cumulative.

Discussion

Long-term student success is clearly an important goal for all postsecondary institutions.

And, in light of the increased pressure to improve degree completion rates, institutions may be

more likely to admit students who have a reasonable chance of progressing towards and

completing a degree. Four-year institutions often use multiple measures in making admission

decisions. And, even though most two-year institutions have open admission policies, they often

look at students’ high school records and require students to take course placement exams to help

HSGPA

First-year

GPA

Year 3

cum GPA

ACTC

score

0.71

0.12

0.47

0.35

0.22

0.10

determine which courses they will need to take. Most students also use test scores and HSGPA to

help them identify institutions to which they want to apply for admissions (Sawyer, 2010).

In this study, for both four- and two-year institutions, we evaluated the utility of ACT

scores and HSGPAs for identifying among possible applicants those who are likely to be

successful in college beyond the first year. For each outcome, estimated decision-based statistics

associated with ACTC score and HSGPA used alone and jointly were compared at values that

maximized the percentage of correct classifications. The ACTC and HSGPA selection values

(that maximized prediction accuracy) identified in this study were used for comparative purposes

only. In general, institutions rarely use strict selection cutoffs in making their admission

decisions.

Some researchers have suggested that standardized tests like the ACT are not useful and

not predictive of long-term college success (Soares, 2012). However, results from this study

refute that notion. For example, typical maximum accuracy rates for the progress to degree

outcomes over time through degree completion were moderately high (64% to 71% at four-year

institutions and 65% to 77% at two-year institutions). In general, typical maximum accuracy and

corresponding success rates were slightly higher for HSGPA than for ACTC score at four-year

institutions, but were comparable at two-year institutions. However, across college outcomes at

both types of institutions, using both ACTC score and HSGPA was generally more beneficial for

improving prediction accuracy and success rates over those based on single-predictor models,

providing evidence of the incremental benefit of using both measures for predicting college

success beyond the first year. Our estimate of the typical maximum accuracy rate based on the

joint model for predicting six-year bachelor’s degree completion (67%) is in line with results

from another study (Schmitt, Keeney, Oswald, Pleskac, Billington, Sinha, & Zorzie, 2009) that

found that four-year bachelor’s degree completion was successfully predicted by SAT/ACT

scores and HSGPA jointly for 63% of students.

Results based on the ACTC and HSGPA joint model also suggest that the effect of

HSGPA on long-term college success depends on the student’s ACTC score. For example, for

students with a HSGPA of 4.0, the typical chances of completing a bachelor’s degree by year 6

from the same initial four-year institution were a little over 60% (Figure A-2). But, these chances

were greater for students with higher ACTC scores and smaller for those with lower ACTC

scores (ranging from 40% to 80%; Figure 2). The ACTC score differential was also found to be

larger for students with higher HSGPAs. Two other independent studies (ACT, 2012) based on

observed degree completion rates also found this to be the case. Allowing higher ACTC scores to

compensate for lower HSGPAs and vice versa contributes to the increase in the percentage of

correct classifications based on the joint model.

Another important finding from the first part of this study was the apparent inability of

HSGPA to predict higher levels of later college GPA (at year 6 for four-year institutions and at

year 3 for two-year institutions). For example, across four-year institutions, the typical chances

of achieving higher levels of year 6 cumulative GPA associated with a HSGPA of 4.00 was less

than 50% (for criterion level of 3.75; Figure A-8). On the other hand, the typical chances of

achieving a year 6 cumulative GPA of 3.75 or higher were relatively high for students with

higher ACTC scores (Figure A-7). Sawyer (2010) found a similar result for first-year college

GPA. Moreover, the typical ACTC and HSGPA values that maximized prediction accuracy for

achieving different levels of year 6 cumulative GPA were similar to those reported in Noble and

Sawyer (2002) for predicting first-year college GPA levels. However, the typical maximum

accuracy rates found in this study were slightly lower than those found for first-year college

GPA. As one would expect, this finding suggests that these two pre-enrollment measures are

more strongly related to more proximal college outcomes than to the more distal ones.

The findings from the path models examined in this study highlight the importance of

students being ready for college and performing well academically during their first year to

improve their chances of progressing towards and completing a degree. Both ACTC score and

HSGPA were primarily indirectly related to subsequent college outcomes (through first-year

college GPA), but both contributed unique information towards predicting first-year academic

performance. Across the outcomes, the total effect of each of the pre-enrollment measures (direct

and indirect) was smaller than the direct/total effect of first-year GPA. One limitation of these

analyses was that students who dropped out prior to the end of the first year were not included.

Another recent study (Allen & Robbins, 2010) also examined path models for predicting

timely degree attainment (by year 4 for bachelor’s degrees and by year 2 for associate’s degrees).

The study also found that first-year academic performance had the largest effect on degree

completion. It also found that ACTC score and HSGPA were more predictive of first-year

academic performance than other noncognitive and sociodemographic characteristics. Compared

to our study, differences in their standardized path coefficients for predicting first-year academic

performance between HSGPA and ACTC score were substantially smaller for four-year

institutions, but they were considerably larger for two-year institutions. They adjusted for

measurement error in their analyses; we did not account for range restriction or measurement

error in our analyses.

Results from the path analysis also suggest that while ACTC score and HSGPA are

correlated, there are differences in what they measure. ACT scores are reported on a score scale

that maintains the same meaning across years and across high schools and are therefore not

affected by differential grading standards as is the case for HSGPA (Woodruff & Ziomek, 2004).

ACT scores reflect level of educational achievement at a moment in time, often at the end of a

student’s junior year or beginning of the senior year in high school. HSGPA, on the other hand,

reflects performance in courses over the duration of high school, and is not only affected by level

of content mastery, but is also affected by a student’s personal behaviors, such as whether the

student is prudent about taking good notes, putting forth effort and participating in class,

completing homework assignments, and preparing well for course exams.

From part 2 of this study, we found that the ACT College Readiness Benchmarks are also

useful for predicting long-term college success through degree completion for applicant pools,

providing further validity evidence for using the Benchmarks as an empirical definition of

college readiness. Typical success rates for predicting college outcomes beyond the first year

were generally higher for the ACT Mathematics and Science Benchmarks than for the ACT

English and Reading Benchmarks. In addition, typical success rates for the ACT Mathematics

Benchmark were generally comparable to those based on ACTC scores that maximized

prediction accuracy. The exceptions to this finding were for outcomes with relatively low

success rates consistently seen across all institutions (e.g., associate’s degree completion by year

3, year 6/year 3 cumulative GPA of 3.50 or higher).

The finding associated with the ACT Mathematics Benchmark is supported by the results

from another study: Adelman (2006) found that the highest level of high school mathematics

coursework is an important factor associated with bachelor’s degree completion. A policy brief

by Achieve (2008) suggests that the reason high school mathematics preparation is so important

for college success is related to the higher-order thinking and critical reasoning skills that

students learn beginning in Algebra I and continue to build upon in subsequent higher-level

mathematics courses. Students who develop these skills are better equipped for their future

career pathways, whichever path they may choose to follow.

The typical estimated success rates associated with the Benchmark scores and with the

ACTC or HSGPA values that maximized prediction accuracy were only slightly lower for

applicants at two-year institutions than for those at four-year institutions. This result was seen for

the progress to degree outcomes based on cumulative hours earned over time, as well as the year

6/year 3 cumulative GPA outcomes.

These findings indicate that in order for all students to

have a reasonable chance of progressing towards and completing a degree at either type of

institution, they need to graduate from high school with a core set of academic skills that help put

them on a more direct path towards long-term college success.

Unfortunately, too many high school graduates are underprepared for college-level

coursework and need to take remedial coursework. For example, only 66% of the 1.6 million

ACT-tested 2011 high school graduates met the ACT English Benchmark (ACT, 2011). The

corresponding percentages were 45%, 52%, and 30% in mathematics, reading, and science,

respectively. But, degree completion is often delayed for students taking remedial courses

(Adelman, 2004). States, districts, and high schools are increasingly implementing policies to

help address this need to prepare all students for college and career. For example, 45 states and

the District of Columbia have formally adopted the Common Core State Standards (National

Governors Association Center for Best Practices & Council of Chief State School Officers,

2012) in an effort to help improve the college and career readiness of their high school graduates.

The typical ACTC or HSGPA values that maximized prediction accuracy (that is, the

values associated with at least a 50% chance of being successful) were relatively high for degree

For these outcomes, the ACTC and HSGPA values that maximized prediction accuracy were similar between two-

and four-year institutions so these comparisons are appropriate and meaningful. In addition, the corresponding

typical maximum accuracy rates were comparable between the two types of institutions for these outcomes.

completion from the same initial institution, especially for two-year institutions.

Using these

high selection values would result in a substantial percentage of applicants being rejected for

admissions. However, in another study where students were followed across institutions, we

(Radunzel & Noble, 2012) found that among students with ACTC scores or HSGPAs below the

selection values identified in this study, a significant percentage of these students completed a

degree by year 6. For example, 65% of students who initially enrolled in four-year institutions

and had an ACTC score of 22 to 24 completed a bachelor’s degree within six years of enrolling

in college. It is interesting to note that for the institutions included in this study, the ACTC and

HSGPA values that maximized prediction accuracy varied substantially across institutions, and

were related to an institution’s degree completion rate (lower selection scores were generally

seen for institutions with higher degree completion rates). In part, the typical selection values are

so high because six-year degree completion rates from the same initial institution are generally

low: it is common for students to transfer to another institution (Hossler, Shapiro, Dundar,

Ziskin, Chen, Zerquera, & Torres, 2012). It is also common for students from two-year

institutions to take longer than three years to complete a degree (Green & Radwin, 2012).

In this study, we were limited to evaluating degree completion from the initial institution.

But institutions are primarily interested in identifying applicants who are most likely to graduate

from their institution. For two-year institutions, we also accounted for those students who

transferred to an in-state four-year institution. Besides degree completion, we also considered

progress to degree outcomes over time that were based on cumulative credit-bearing hours

earned. These outcomes might be useful indicators of success for two-year institutions,

These selection values are similar to those one might expect to see as criteria for admission at highly selective

four-year institutions or for merit-based scholarships.

especially as they move towards establishing intermediate markers that help track students’

progress along the pathway to degree completion (Moore et al., 2009).

Most institutions admit 50% to 85% of their applicants (Clinedinst et al., 2011).

Institutions are able to compensate for lower admissions standards with effective support

programs and interventions (Lotkowski, Robbins, & Noeth, 2004; Tinto, 2002). For example,

ACT scores and HSGPA, along with measures of psychosocial factors can help campuses

identify students who are most in need of academic support programs (Robbins et al., 2006).

Moreover, using multiple measures, including augmenting pre-enrollment measures with

information collected early in college (such as mid-term grades during a student’s first term) to

predict later college success enables colleges to identify and intervene with high-risk students in

appropriate ways.

In this study, we could only approximate applicant pools for the institutions using data

for all students who sent their ACT scores to these institutions over the study time period. In

addition, the approach of evaluating one or two pre-enrollment measures at a time is a

simplification of the admissions process. However, the methods used can be extended to include

additional measures. Another limitation of the study was that the study sample is not a nationally

representative sample. In spite of this limitation, this study was based on a large number of

institutions and a large number of students. Another strength of the study was that we included

results for ACT-tested students from both two- and four-year institutions. And, although the

ACT is generally not required for admission to two-year institutions, in states that administer the

ACT statewide, most if not all public high school graduates will have ACT scores available for

use by institutions. Future research might consider examining these same research questions for

COMPASS

-tested students at two-year institutions.

Future research might also examine the

effects of using ACTC score and HSGPA for predicting long-term college success across student

demographic groups to ensure equity in the admissions process.

In conclusion, findings from this study suggest that if institutions wish to admit students

with the highest likelihood of success, both ACTC score and HSGPA should be considered, as

both measures are related to college success during the first year and in subsequent years through

degree completion. The ACT Benchmarks are also effective at identifying students who are

ready for college and likely to succeed beyond the first year of college. For four-year institutions

wanting to incorporate a student’s likelihood of long-term college success into their admission

decisions, the results from this study suggest that HSGPA should carry

 slightly greater weight than ACTC score for evaluating a student’s likelihood of

progressing towards or completing a bachelor’s degree within six years of initially

enrolling (irrespective of students’ final cumulative GPAs)

, and

 approximately the same weight as ACTC score for predicting a student’s

likelihood of achieving moderate levels of year 6 cumulative GPA, and slightly

less weight for predicting a student’s likelihood of achieving higher levels of year

6 cumulative GPA.

For two-year institutions, HSGPA should carry approximately the same weight as ACTC score

for predicting the progress to degree outcomes considered in this study, as well as moderate

levels of year 3 cumulative GPA. For evaluating degree completion and higher levels of year 3

cumulative GPA, HSGPA should carry less weight than ACTC score.

COMPASS also has College Readiness Benchmarks (ACT, 2010a).

This recommendation does not take into account students’ cumulative GPAs; it addresses only whether students

will graduate in six years (some students may graduate with lower cumulative GPAs). Findings from this study

support the notion that ACTC score should carry the same or slightly greater weight than HSGPA for predicting

final cumulative GPA.

References

Achieve. (2008). The building blocks of success: Higher-level math for all students. (Achieve

Policy Brief). Washington, DC: Author.

ACT. (2012). ACT scores and postsecondary degree attainment: What’s the connection? (ACT

Information Brief No. 2012-11). Iowa City, IA: Author.

ACT. (2011). ACT profile report–national: Graduating class 2011–national. Iowa City, IA:

Author.

ACT. (2010a). What are ACT’s college readiness benchmarks? Iowa City, IA: Author.

ACT. (2010b). Mind the gaps: How college readiness narrows achievement gaps in college

success. Iowa City, IA: Author.

ACT. (2004). ACT class profile report, 2003-2004 freshmen. (Available from ACT, Inc., P.O.

Box 168, Iowa City, Iowa 52243).

Adelman, C. (2006). The toolbox revisited: Paths to degree completion from high school through

college. Washington, DC: U.S. Department of Education.

Adelman, C. (2004). Principal indicators of student academic histories in postsecondary

education. Washington, DC: U.S. Department of Education.

Allen, J., & Robbins, S. (2010). Effects of interest-major congruence, motivation, and academic

performance on timely degree attainment. Journal of Counseling Psychology, 57(1), 23-

35.

Allen, J., Robbins, S., Casilla, A., & Oh, I. (2008) Third-year college retention and transfer:

Effects of academic performance, motivation, and social connectedness. Research in

Higher Education, 49, 647-664.

Allen, J., & Sconing, J. (2005). Using ACT Assessment scores to set benchmarks for college

readiness. (ACT Research Report No. 2005-3). Iowa City, IA: ACT, Inc.

Breland, H., Maxey, J., Gernand, R., Cumming, T., & Trapani, C. (2002). Trends in college

admission admission 2000: A report of a survey of undergraduate admissions policies,

practices, and procedures. Retrieved May 10, 2012 from the Association for Institutional

Research website: http://www.airweb.org/images/trendsreport.pdf.

Clinedinst, M.E., Hurley, S.F., and Hawkins, D.A. (2011). 2011 State of College Admission.

Alexandria, VA: National Association for College Admissions Counseling.

González, J. (2012, April 22). Education for all? 2-year colleges struggle to preserve their

mission. The Chronicle of Higher Education. Retrieved on May 10, 2012 from

http://chronicle.com/article/2-Year-Colleges-Fight-to-Save/131608/.

Green, C., & Radwin, D. (2012). Characteristics of associate’s degree attainers and time to

associate’s degree. (Web Tables NCES 2012-271). Washington, DC: U.S. Department of

Education.

Higher Education Research Institute (2011, November). Completing college: Assessing

graduation rates at four-year institutions. HERI Research Brief, UCLA, Los Angeles,

CA: HERI.

Hossler, D., Shapiro, D., Dundar, A., Ziskin, M., Chen, J., Zerquera, D., Torres, V. (2012).

Transfer & Mobility: A national view of pre-degree student movement in postsecondary

institutions. (Signature Report™ 2). Herndon, VA: National Student Clearinghouse

Research Center.

Jasti, S., Dudley, W.N., & Goldwater, E. (2008). SAS macros for testing statistical mediation in

data with binary mediators or outcomes. Nursing Research, 57, 118-121,

Knapp, L.G., Kelly-Reid, J.E., Whitmore, R.W., Wu, S., Gallego, L., Cong, J., Berzofsky, M.,

Huh, S., Levine, B., & Broyles, S.G. (2005). Postsecondary institutions in the United

States: Fall 2003 and degrees and other awards conferred: 2002-03. (NCES 2005-154).

U.S. Department of Education, Washington, DC: National Center for Education

Statistics.

Lotkowski, V.A., Robbins, S.B., & Noeth, R.J. (2004). The role of academic and non-academic

factors in improving college retention. Iowa City, IA: ACT, Inc.

MacKinnon, D.P., & Dwyer, J.H. (1993). Estimating mediated effects in prevention studies.

Evaluation Review, 17, 144-158.

Moore, C., & Shulock, N. (2009). Student progress toward degree completion: Lessons from the

research literature. Sacramento, CA: Institute for Higher Education Leadership & Policy.

Moore, C., Shulock, N., & Offenstein, J. (2009). Steps to success: analyzing milestone

achievement to improve community college student outcomes. Sacramento, CA: Institute

for Higher Education Leadership & Policy.

National Governors Association Center for Best Practices & Council of Chief State School

Officers. (2012). Common Core State Standards Initiative: In the States. Retrieved June

12, 2012 from http://www.corestandards.org/in-the-states.

Noble, J., & Radunzel, J. (2007). College readiness = college success beyond the first year. Paper

presented at the Annual Forum of the Association for Institutional Research, June 2–6,

Kansas City, Missouri.

Noble, J., & Sawyer, R. (2002). Predicting different levels of academic success in college using

high school GPA and ACT Composite score. (ACT Research Report No. 2002-4). Iowa

City, IA: ACT, Inc.

Radunzel, J., & Noble, J. (2012). Tracking 2003 ACT

-tested high school graduates: College

readiness, enrollment, and long-term success. (ACT Research Report No. 2012-2). Iowa

City, IA: ACT, Inc.

Robbins, S., Allen, J., Casillas, A., Peterson, C., & Le, H. (2006). Unraveling the differential

effects of motivational and skills, social, and self-management measures from traditional

predictors of college outcomes. Journal of Educational Psychology, 98, 598-616.

Sackett, P.R., Kuncel, N.R., Arneson, J.J., Cooper, S.R., & Waters, S.D. (2009). Does

socioeconomic status explain the relationship between admission tests and post-

secondary academic performance? Psychological Bulletin, 135(1), 1-22.

Saupe, J.L., & Curs, B.R. (2008). Deriving enrollment management scores from ACT data. (IR

Applications No. 16). Tallahassee, FL: The Association for Institutional Research.

Sawyer, R. (2010). Usefulness of high school average and ACT scores in making college

admission decisions. (ACT Research Report No. 2010-2). Iowa City, IA: ACT, Inc.

Sawyer, R.L. (1996). Decision theory models for validating course placement tests. Journal of

Educational Measurement, 33, 271-290.

Schmitt, N., Keeney, J, Oswald, F.L., Pleskac, T.J., Billington, A.Q., Sinha, R., & Zorzie, M.

(2009). Prediction of 4-year college student performance using cognitive and

noncognitive predictors and the impact on demographic status of admitted students.

Journal of Applied Psychology, 94(6), 1479-1497.

Soares, J.A. (Ed.) (2012). SAT wars: The case for test-optional college admissions. New York,

NY: Teachers College Press.

Tinto, V. (2002, October). Enhancing student persistence: Connecting the dots. Paper presented

at the Optimizing the Nation’s Investment: Persistence and Success in Postsecondary

Education conference sponsored by the Wisconsin Center for the Advancement of

Postsecondary Education.

Westrick, P., Le, H., Robbins, S., Radunzel, J., & Schmidt, F. (2012). “Can Do” vs. “Will Do”:

The Relative Effects of ACT Scores and High School GPA on College Performance and

Retention. Manuscript in preparation.

Woodruff, D.J., & Ziomek, R.L. (2004). High school grade inflation from 1991 to 2003. (ACT

Research Report No. 2004-4). Iowa City, IA: ACT, Inc.

Appendix A

Figures A-1 to A-10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACTC score

Probability

Bachelor's degree by year 6

Associate's degree or transfer to four-year institution by year 3

Associate's degree by year 3

Figure A-1. Estimated probabilities of degree completion based on ACTC score. ACTC = ACT

Composite.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.5 3.0 3.5 4.0

HSGPA

Probability

Bachelor's degree by year 6

Associate's degree or transfer to four-year institution by year 3

Associate's degree by year 3

Figure A-2. Estimated probabilities of degree completion based on HSGPA. HSGPA = high

school grade point average.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACTC score

Probability

Year 1 (24+ hours)

Year 2 (48+ hours)

Year 3 (72+ hours)

Year 4 (96+ hours)

Figure A-3. Estimated probabilities of progressing towards a degree based on ACTC score for

four-year institutions. ACTC = ACT Composite.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.5 3.0 3.5 4.0

HSGPA

Probability

Year 1 (24+ hours)

Year 2 (48+ hours)

Year 3 (72+ hours)

Year 4 (96+ hours)

Figure A-4. Estimated probabilities of progressing towards a degree based on HSGPA for four-

year institutions. HSGPA = high school grade point average.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACTC score

Probability

Year 1 (18+ hours)

Year 2 (36+ hours)

Year 3 (54+ hours)

Figure A-5. Estimated probabilities of progressing towards a degree based on ACTC score for

two-year institutions. ACTC = ACT Composite.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.5 3.0 3.5 4.0

HSGPA

Probability

Year 1 (18+ hours)

Year 2 (36+ hours)

Year 3 (54+ hours)

Figure A-6. Estimated probabilities of progressing towards a degree based on HSGPA for two-

year institutions. HSGPA = high school grade point average.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACTC score

Probability

3.00 or higher

3.25 or higher

3.50 or higher

3.75 or higher

Figure A-7. Estimated probabilities of achieving specific year 6 cumulative GPAs based on

ACTC score for four-year institutions. ACTC = ACT Composite.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.5 3.0 3.5 4.0

HSGPA

Probability

3.00 or higher

3.25 or higher

3.50 or higher

3.75 or higher

Figure A-8. Estimated probabilities of achieving specific year 6 cumulative GPAs based on

HSGPA for four-year institutions. HSGPA = high school grade point average.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 15 20 25 30 35

ACTC score

Probability

3.00 or higher

3.25 or higher

3.50 or higher

3.75 or higher

Figure A-9. Estimated probabilities of achieving specific year 3 cumulative GPAs based on

ACTC score for two-year institutions. ACTC = ACT Composite.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2.0 2.5 3.0 3.5 4.0

HSGPA

Probability

3.00 or higher

3.25 or higher

3.50 or higher

3.75 or higher

Figure A-10. Estimated probabilities of achieving specific year 3 cumulative GPAs based on

HSGPA for two-year institutions. HSGPA = high school grade point average.

Appendix B

Tables B-1 to B-4

Appendix B

Table B-1

Results for Bachelor’s Degree Completion and Progress to Degree at Four-Year Institutions based on ACTC Score and HSGPA

Models

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

Bachelor’s degree completion by year 6

ACTC 58 25 9/31 64 55/84 24 0/67 57 52/73 84 0/100

HSGPA 56 3.57 2.20/4.00 65 58/79 23 0/57 58 50/77 67 3/96

ACTC &

HSGPA

61 67 58/84 26 0/67 60 53/76 71 1/98

Progress to degree year 1

ACTC 44 18 10/23 71 62/84 4 0/58 70 61/84 22 0/89

HSGPA 48 2.80 1.39/3.72 73 66/86 6 0/51 73 58/86 25 0/89

ACTC &

HSGPA

48 74 66/83 6 0/59 74 67/83 28 0/87

Progress to degree year 2

ACTC 49 20 10/26 65 58/84 12 0/66 63 58/77 52 0/94

HSGPA 50 3.13 2.07/3.91 69 61/81 14 0/61 66 54/79 45 1/95

ACTC &

HSGPA

50 69 61/84 15 0/67 67 62/75 50 2/93

Progress to degree year 3

ACTC 50 22 9/28 65 55/84 19 0/68 60 55/76 66 0/97

HSGPA 50 3.39 2.38/3.97 67 60/82 21 0/64 62 51/76 57 2/97

ACTC &

HSGPA

50 68 59/85 22 0/68 64 59/74 62 2/95

Table B-1 (cont.)

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

Progress to degree year 4

ACTC 48 24 10/29 64 55/83 23 0/65 59 53/72 71 0/98

HSGPA 49 3.46 2.46/4.00 67 59/81 23 1/61 60 50/75 60 4/97

ACTC &

HSGPA

50 68 58/83 25 0/65 62 57/74 65 2/96

Note. The selection value is the predictor value associated with the maxAR. There were 61 and 50 institutions with available data for bachelor’s degree

completion and progress to degree analyses, respectively. Med = Median; Min = Minimum; Max = Maximum; ACTC = ACT Composite; HSGPA = high school

grade point average.

Multiple combinations of ACTC score and HSGPA corresponded to a probability of 0.50 for the joint model.

Table B-2

Results for Associate’s Degree Completion and Progress to Degree at Two-Year Institutions based on ACTC Score and HSGPA

Models

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

Associate’s degree completion by year 3

ACTC 25 29 24/33 81 66/93 63 32/86 55 51/62 99 92/100

HSGPA 5 3.92 3.82/3.98 72 69/74 43 36/48 52 51/54 90 83/94

ACTC &

HSGPA

25 82 66/93 64 31/86 53 50/56 99 82/100

Associate’s degree completion or transfer to four-year institution by year 3

ACTC 38 27 21/33 77 63/88 54 22/77 55 50/60 98 76/100

HSGPA 14 3.75 3.43/3.96 69 63/75 37 24/51 54 51/60 86 63/97

ACTC &

HSGPA

38 77 63/88 54 25/77 55 51/62 96 65/100

Progress to degree year 1

ACTC 41 19 11/31 66 57/80 17 0/59 65 52/73 52 0/99

HSGPA 41 3.03 1.66/3.68 65 62/77 15 0/47 64 56/77 50 1/90

ACTC &

HSGPA

42 68 62/79 18 0/58 67 52/77 51 1/100

Progress to degree year 2

ACTC 41 21 15/27 65 59/79 25 1/57 60 54/65 75 16/96

HSGPA 41 3.38 2.44/3.89 66 60/78 26 2/56 59 53/67 72 19/97

ACTC &

HSGPA

41 67 61/80 28 3/58 62 55/68 71 22/93

Table B-2 (cont.)

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

Progress to degree year 3

ACTC 41 23 17/30 67 58/82 34 5/64 57 53/61 88 37/99

HSGPA 36 3.62 2.85/3.96 68 58/77 33 6/54 56 51/65 83 40/98

ACTC &

HSGPA

41 69 59/83 35 7/65 58 54/66 81 42/98

Note. The selection value is the predictor value associated with the maxAR. There were 43 and 42 institutions with available data for associate’s degree

completion and progress to degree analyses, respectively. There were 40 institutions with data available for associate’s degree completion or transfer to a four-

year institution by year 3. Med = Median; Min = Minimum; Max = Maximum; ACTC = ACT Composite; HSGPA = high school grade point average.

Multiple combinations of ACTC score and HSGPA corresponded to a probability of 0.50 for the joint model.

Table B-3

Results for Achieving Levels of Year 6 College Cumulative GPA at Four-Year Institutions based on ACTC Score and HSGPA Models

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

3.00 or higher college GPA

ACTC 57 20 16/23 67 61/73 12 3/33 68 62/75 43 18/85

HSGPA 57 3.17 2.69/3.69 70 65/75 16 5/36 70 60/77 41 21/78

ACTC &

HSGPA

57 71 63/75 18 6/38 71 63/78 49 24/82

3.25 or higher college GPA

ACTC 57 24 20/27 70 64/81 33 15/60 64 56/70 76 45/96

HSGPA 57 3.59 3.03/3.96 73 64/80 35 10/59 63 53/70 68 33/94

ACTC &

HSGPA

57 75 67/82 39 12/63 67 61/73 74 41/94

3.50 or higher college GPA

ACTC 57 27 22/31 79 69/92 56 33/83 61 55/69 92 77/99

HSGPA 40 3.86 3.53/4.00 79 69/87 56 28/73 56 50/62 86 65/97

ACTC &

HSGPA

57 83 70/92 62 30/84 63 57/69 90 71/99

Table B-3 (cont.)

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

3.75 or higher college GPA

ACTC 56 31 26/34 91 78/98 82 57/96 58 52/65 99 94/100

HSGPA 2 4.00 4.00/4.00 84 79/89 67 57/77 50 50/50 90 87/92

ACTC &

HSGPA

57 92 80/98 84 57/96 59 54/64 98 89/100

Note. The selection value is the predictor value associated with the maxAR. Students’ cumulative GPAs at degree completion were included in year 6 GPA

analyses for students who graduated with a bachelor’s degree before the end of year 6. There were 57 institutions with data available for the year 6 college GPA

analyses. Med = Median; Min = Minimum; Max = Maximum; ACTC = ACT Composite; HSGPA = high school grade point average.

Multiple combinations of ACTC score and HSGPA corresponded to a probability of 0.50 for the joint model.

Table B-4

Results for Achieving Levels of Year 3 College Cumulative GPA at Two-Year Institutions based on ACTC Score and HSGPA Models

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

3.00 or higher college GPA

ACTC 42 20 17/23 66 62/71 22 5/38 63 58/70 67 34/92

HSGPA 42 3.30 2.70/3.54 68 63/72 25 6/42 62 59/68 64 32/85

ACTC &

HSGPA

42 70 63/74 27 7/44 66 61/71 67 39/84

3.25 or higher college GPA

ACTC 42 23 21/26 74 65/80 46 24/61 62 55/67 88 62/98

HSGPA 42 3.73 3.41/3.95 75 64/80 48 24/58 56 52/61 86 68/96

ACTC &

HSGPA

42 77 67/81 50 27/61 64 58/67 85 64/95

3.50 or higher college GPA

ACTC 42 26 23/29 84 74/89 66 44/78 60 53/66 97 86/100

HSGPA 5 3.98 3.93/4.00 78 76/81 55 51/63 51 50/53 93 88/95

ACTC &

HSGPA

42 85 76/90 68 48/79 61 55/64 95 84/99

Table B-4 (cont.)

Predictor

variable

Number

institutions

with viable

models

Selection value

(SV)

Maximum

accuracy rate

(maxAR)

Increase in AR

(∆AR)

Success rate

(SR)

Observed

percentage below

Med Min/Max Med Min/Max Med Min/Max Med Min/Max Med Min/Max

3.75 or higher college GPA

ACTC 40 29 27/33 92 85/96 84 70/91 58 50/69 99 98/100

HSGPA 0

ACTC &

HSGPA

41 93 86/96 85 72/91 58 52/80 99 97/100

Note. The selection value is the predictor value associated with the maxAR. Students’ cumulative GPAs at degree completion were included in year 3 GPA

analyses for students who graduated with an associate’s degree before the end of year 3. There were 42 institutions with available data for the year 3 cumulative

GPA analyses. Med = Median; Min = Minimum; Max = Maximum.

Multiple combinations of ACTC score and HSGPA corresponded to a probability of 0.50 for the joint model.

Appendix C

Tables C-1 to C-4

Table C-1

Results for Bachelor’s Degree Completion and Progress to Degree at Four-Year Institutions

based on ACT College Readiness Benchmarks

Subject

area

ACT

Benchmark

score

Probability of

success at

Benchmark

Success rate

(SR)

Increase in SR

(∆SR)

Med Min/Max Med Min/Max Med Min/Max

Bachelor’s degree completion by year 6

English 18 0.35 0.12/0.76 46 21/80 5 1/12

Mathematics 22 0.46 0.23/0.79 53 30/82 13 3/30

Reading 21 0.39 0.17/0.78 47 22/80 5 1/13

Science 24 0.47 0.24/0.80 52 29/81 11 2/25

Progress to degree year 1

English 18 0.59 0.26/0.84 71 44/88 6 2/21

Mathematics 22 0.73 0.51/0.87 82 66/90 15 4/47

Reading 21 0.67 0.35/0.86 74 45/89 7 2/21

Science 24 0.76 0.53/0.89 81 61/91 12 3/42

Progress to degree year 2

English 18 0.45 0.15/0.74 59 31/81 6 3/15

Mathematics 22 0.61 0.36/0.77 70 49/83 15 5/38

Reading 21 0.53 0.24/0.79 61 34/83 7 2/15

Science 24 0.62 0.39/0.83 69 47/85 14 4/33

Progress to degree year 3

English 18 0.39 0.12/0.73 51 25/79 6 3/13

Mathematics 22 0.53 0.29/0.76 62 42/81 16 4/33

Reading 21 0.46 0.19/0.77 53 27/81 7 3/14

Science 24 0.56 0.31/0.80 61 39/82 14 4/30

Progress to degree year 4

English 18 0.37 0.12/0.73 48 24/79 6 2/12

Mathematics 22 0.51 0.27/0.76 57 38/81 15 4/28

Reading 21 0.43 0.18/0.77 51 26/80 7 2/13

Science 24 0.53 0.29/0.80 58 36/82 13 4/25

Note. These analyses were based on all institutions with available data for each outcome (61 institutions for

bachelor’s degree completion and 50 for progress to degree outcomes). Med = Median; Min = Minimum; Max =

Maximum.

Table C-2

Results for Associate’s Degree Completion and Progress to Degree at Two-Year Institutions

based on ACT College Readiness Benchmarks

Subject

area

ACT

Benchmark

score

Probability of

success at

Benchmark

Success rate

(SR)

Increase in SR

(∆SR)

Med Min/Max Med Min/Max Med Min/Max

Associate’s degree completion by year 3

English 18 0.12 0.04/0.34 18 6/41 4 1/9

Mathematics 22 0.22 0.07/0.50 28 10/59 14 3/26

Reading 21 0.15 0.05/0.36 19 7/42 5 2/10

Science 24 0.23 0.07/0.49 26 9/54 12 4/21

Associate’s degree completion or transfer to four-year institution by year 3

English 18 0.21 0.09/0.41 29 13/51 6 3/11

Mathematics 22 0.36 0.15/0.61 42 19/70 19 11/28

Reading 21 0.25 0.10/0.45 30 13/52 7 4/11

Science 24 0.35 0.14/0.58 38 16/63 16 8/22

Progress to degree year 1

English 18 0.49 0.16/0.78 62 26/83 12 6/21

Mathematics 22 0.70 0.23/0.87 79 29/90 27 9/44

Reading 21 0.56 0.20/0.80 64 27/83 14 5/26

Science 24 0.70 0.30/0.86 74 38/88 23 11/38

Progress to degree year 2

English 18 0.38 0.10/0.62 49 14/69 9 2/16

Mathematics 22 0.57 0.13/0.78 64 15/83 24 4/35

Reading 21 0.44 0.12/0.64 50 15/70 10 3/17

Science 24 0.57 0.16/0.75 62 19/78 20 7/29

Progress to degree year 3

English 18 0.32 0.06/0.55 41 9/63 8 2/12

Mathematics 22 0.49 0.08/0.71 57 10/77 22 3/31

Reading 21 0.37 0.07/0.58 43 10/64 9 3/12

Science 24 0.49 0.09/0.69 53 11/73 19 4/24

Note. These analyses were based on all institutions with available data for each outcome (43 institutions for

associate’s degree completion, 40 for associate’s degree completion or transfer to a four-year institution, and 42 for

progress to degree outcomes). Med = Median; Min = Minimum; Max = Maximum.

Table C-3

Results for Achieving Levels of Year 6 College Cumulative GPA at Four-Year Institutions based

on ACT College Readiness Benchmarks

Subject

area

ACT

Benchmark

score

Probability of

success at

Benchmark

Success rate

(SR)

Increase in SR

(∆SR)

Med Min/Max Med Min/Max Med Min/Max

3.00 or higher college GPA

English 18 0.46 0.33/0.62 65 49/78 10 5/19

Mathematics 22 0.66 0.48/0.78 76 60/86 19 9/35

Reading 21 0.59 0.43/0.71 69 53/81 12 7/22

Science 24 0.72 0.54/0.82 77 60/86 19 11/33

3.50 or higher college GPA

English 18 0.13 0.05/0.35 29 19/43 7 3/11

Mathematics 22 0.28 0.14/0.52 40 24/60 18 8/29

Reading 21 0.22 0.12/0.41 34 22/48 10 6/16

Science 24 0.34 0.20/0.53 43 30/56 19 10/25

Note. Students’ cumulative GPAs at degree completion were included in Year 6 GPA analyses for students who

graduated with a bachelor’s degree before the end of year 6. These analyses were based on all 57 institutions with

year 6 college cumulative GPA data available. Med = Median; Min = Minimum; Max = Maximum; GPA = grade

point average.

Table C-4

Results for Achieving Levels of Year 3 College Cumulative GPA at Two-Year Institutions based

on ACT College Readiness Benchmarks

Subject

area

ACT

Benchmark

score

Probability of

success at

Benchmark

Success rate

(SR)

Increase in SR

(∆SR)

Med Min/Max Med Min/Max Med Min/Max

3.00 or higher college GPA

English 18 0.42 0.33/0.59 55 43/70 13 8/15

Mathematics 22 0.63 0.51/0.79 70 57/86 26 17/31

Reading 21 0.50 0.40/0.64 60 46/73 16 10/21

Science 24 0.63 0.51/0.78 68 55/82 24 18/31

3.50 or higher college GPA

English 18 0.14 0.09/0.25 25 17/40 8 5/12

Mathematics 22 0.30 0.21/0.50 39 28/63 23 15/35

Reading 21 0.21 0.13/0.32 29 21/44 12 8/16

Science 24 0.34 0.22/0.51 41 28/59 23 15/31

Note. Students’ cumulative GPAs at degree completion were included in Year 3 GPA analyses for students who

graduated with an associate’s degree before the end of year 3. These analyses were based on all 42 institutions with

year 3 college cumulative GPA data available. Med = Median; Min = Minimum; Max = Maximum; GPA = grade

point average.

Appendix D

Tables D-1 to D-2

Table D-1

Direct, Indirect, and Total Effects on Long-Term College Outcomes for Four-Year Institutions

Predictor

First-year

college GPA Long-term college outcome

Direct

effect

Direct

effect

Indirect

effect

Total

effect

Bachelor’s degree completion by year 6

ACTC 0.26 NS 0.26 0.26

HSGPA 0.35 0.07 0.24 0.31

First-year GPA – 0.52 – 0.52

Year 6 college cumulative GPA

ACTC 0.31 0.09 0.36 0.45

HSGPA 0.32 0.12 0.35 0.47

First-year GPA – 0.68 – 0.68

Progress to degree year 2

ACTC 0.26 NS 0.29 0.29

HSGPA 0.35 0.07 0.29 0.36

First-year GPA – 0.62 – 0.62

Progress to degree year 3

ACTC 0.26 NS 0.28 0.28

HSGPA 0.35 0.07 0.28 0.35

First-year GPA – 0.59 – 0.59

Progress to degree year 4

ACTC 0.26 NS 0.28 0.28

HSGPA 0.35 0.07 0.27 0.34

First-year GPA – 0.58 – 0.58

Note. The direct effects of ACTC score and HSGPA on first-year college GPA for the analyses based on year 6

college cumulative GPA differ from those based on the other outcomes, because the former case includes only those

students with a year 6 cumulative GPA (i.e., those who either graduated prior to year 6 or were still enrolled at year

6). In comparison, the latter group includes all students from four-year institutions with the needed data available to

calculate the outcome of interest (see Table 1). ACTC = ACT Composite; HSGPA = high school grade point

average; GPA = grade point average; NS = not significant at the 0.01 level.

Table D-2

Direct, Indirect, and Total Effects on Long-Term College Outcomes for Two-Year Institutions

Predictor

First-year

college GPA Long-term college outcome

Direct

effect

Direct

effect

Indirect

effect

Total

effect

Associate’s degree completion by year 3

ACTC 0.18 0.06 0.21 0.27

HSGPA 0.29 0.16 0.20 0.36

First-year GPA – 0.46 – 0.46

Associate’s degree completion or transfer to four-year institution by year 3

ACTC 0.18 0.09 0.20 0.29

HSGPA 0.30 0.13 0.21 0.34

First-year GPA – 0.44 – 0.44

Year 3 college cumulative GPA

ACTC 0.22 0.10 0.34 0.44

HSGPA 0.35 0.12 0.38 0.50

First-year GPA – 0.71 – 0.71

Progress to degree year 2

ACTC 0.18 0.14 0.20 0.34

HSGPA 0.29 0.13 0.24 0.37

First-year GPA – 0.47 – 0.47

Progress to degree year 3

ACTC 0.18 0.10 0.20 0.30

HSGPA 0.29 0.14 0.22 0.36

First-year GPA – 0.46 – 0.46

Note. The direct effects of ACTC score and HSGPA on first-year college GPA for the analyses based on year 3

cumulative GPA differ from those based on the other outcomes, because the former case includes only those

students with a year 3 cumulative GPA (i.e., those who either graduated prior to year 3 or were still enrolled at year

3). In comparison, the latter group includes all students from two-year institutions with the needed data available to

calculate the outcome of interest (see Table 1). ACTC = ACT Composite; HSGPA = high school grade point

average; GPA = grade point average; NS = not significant at the 0.01 level.

*050205120* Rev 1

Predicting Long-Term

College Success through

Degree Completion Using

ACT

Composite Score, ACT

Benchmarks, and High School

Grade Point Average

Justine Radunzel

Julie Noble

August 2012

ACT Research

Report Series

2012 (5)