Predicting Success: An Examination of the Predictive Validity of a Measure of Motivational-Developmental Dimensions in College Admissions

Abstract

Amid the COVID-19 pandemic, an unprecedented number of higher education institutions adopted test-optional admissions policies. The proliferation of these policies and the criticism of standardized admissions tests as unreliable predictors of applicants’ postsecondary educational promise have prompted the reimagining of evaluative methodologies in college admissions. However, few institutions have designed and implemented new measures of applicants’ potential for success, rather opting to redistribute the weight given to other variables such as high school course grades and high school GPA. We use multiple regression to investigate the predictive validity of a measure of non-cognitive, motivational-developmental dimensions implemented as part of a test-optional admissions policy at a large urban research university in the United States. The measure, composed of four short-answer essay questions, was developed based on the social-cognitive motivational and developmental-constructivist perspectives. Our findings suggest that scores derived from the measure make a statistically significant but small contribution to the prediction of undergraduate GPA and 4-year bachelor’s degree completion. We also find that the measure does not make a statistically significant nor practical contribution to the prediction of 5-year graduation.

The current landscape of higher education, characterized by the changing demography of college students and public health concerns amid the COVID-19 pandemic, has prompted institutions to employ a more comprehensive approach to the evaluation of admissions applicants’ credentials, personal characteristics, and past experiences. The proliferation of test-optional admissions policies (FairTest National Center for Fair & Open Testing, 2023; Furuta, 2017) and holistic approaches to the evaluation of college admission applicants necessitates a more comprehensive understanding of applicants’ potential to succeed in postsecondary education. Admissions models vary widely depending on institutions’ mission and philosophical orientation. For example, some institutions hold that any student is entitled to a college education whereas others hold that college access is a reward for prior academic success (Perfetto et al., 1999). Regardless of an institution’s philosophical basis for admissions decision-making and the corresponding criteria it uses to determine applicants’ eligibility for admission, most admissions models can be improved by including measures that reliably, accurately, and comprehensively evaluate applicants’ potential for success in college with concern for the fair and equitable distribution of educational opportunities (Camara & Kimmel, 2005; Zwick, 2017).

Measures lack validity when they are not reproducible (or reliable, as defined in psychometrics) for applicants from all backgrounds (American Educational Research Association [AERA], American Psychological Association [APA], & National Council on Measurement in Education [NCME], 2014; Drost, 2011). For example, scholars have documented concerns about the use of standardized tests in college admissions, citing differential prediction by reported race (e.g., Blau et al., 2004), gender (e.g., Leonard & Jiang, 1999), and socioeconomic status (e.g., Rothstein, 2004; Sackett et al., 2012). Research has also demonstrated the limited predictive validity of tests beyond the first year of undergraduate study (e.g., Berry & Sackett, 2009; Sackett & Kuncel, 2018; Sackett et al., 2012).

Although an increasing number of institutions have adopted test-optional policies (FairTest National Center for Fair & Open Testing, 2023; Furuta, 2017) or embraced holistic review (Bastedo et al., 2018; Hossler et al., 2019), few institutions have replaced standardized tests with new, proprietary measures of applicants’ potential to succeed in college. Rather, many institutions have redistributed the weight given to standardized test scores to other aspects of the admissions application such as high school grade point average (HSGPA; Galla et al., 2019; Sweitzer et al., 2018), a practice which may perpetuate barriers to equitable access for students who are systemically marginalized and have limited access to educational opportunities such as enrollment in Advanced Placement courses (Kolluri, 2018). Accordingly, recent research has highlighted the need to provide admissions officers with richer information about prospective students. When used as part of a holistic admissions model, information about applicants’ high school and home context may mitigate bias against disadvantaged students and help to promote equitable access to postsecondary education (Bastedo et al., 2022).

As colleges and universities employ holistic assessments of applicants, the increased inclusion of motivational and developmental constructs in the admissions process stands out as a promising possibility. Sedlacek (2004) advanced the view that non-cognitive constructs, such as achievement motivation, may be used to expand the predictive potential of more frequently relied upon cognitive measures. Motivational constructs reflect the resources and abilities students need to effectively adapt to the various academic, emotional, and social challenges associated with attaining a postsecondary education (Allen et al., 2009; Camara & Kimmel, 2005; Kaplan et al., 2017; Sedlacek, 2017). Scholars suggest flexible admissions policies that incorporate a wide variety of student characteristics—including psychosocial and motivational constructs—will likely have the added benefit of increasing diversity among the student body by promoting more equitable access to postsecondary education (Soares, 2015; Zwick, 2002).

Despite evidence of considerable associations between several psychosocial constructs and college student outcomes, an adequately valid integrated inventory of these constructs has yet to be developed (Le et al., 2005; Sedlacek, 2004). Such an inventory could prove highly valuable, especially as student demographics shift and an increasing number of institutions adopt test-optional policies. We respond to the need for reliable and valid measures of postsecondary promise by investigating the predictive validity of a proprietary measure of admissions applicants’ motivational-developmental attributes. Our measure includes select non-cognitive constructs derived from the social-cognitive and developmental-constructivist domains. The measure was developed in 2015 to coincide with the implementation of a test-optional admissions policy at a large, public, urban research university located in the Mid-Atlantic region of the United States. Guided by motivational and student development theories, we addressed the following research questions:

1. Is the measure of motivational-developmental dimensions a reliable predictor of undergraduate GPA (UGPA), and 4- and 5-year bachelor’s degree completion?

2. Is the measure of motivational-developmental dimensions a statistically significant predictor of UGPA?

3. Is the measure of motivational-developmental dimensions a statistically significant predictor of 4- and 5-year bachelor’s degree completion?

4. Which specific motivational-developmental dimensions, if any, predict UGPA and four- and five-year bachelor’s degree completion?

Literature Review

Although evidence suggests that high-achieving, underserved students particularly benefit from post-college achievements such as increased lifetime earnings (Ma et al., 2019), admissions practices frequently benefit students with higher socioeconomic status (Bowman & Bastedo, 2018). As inequitable admissions practices have persisted, the criteria and methodologies colleges and universities use to evaluate admissions applicants have been the subject of substantial empirical investigation (e.g., Bowman & Bastedo, 2018; Hossler et al., 2019; National Association for College Admission Counseling, 2016; Robbins et al., 2004). Previous research has predominately focused on cognitive factors such as standardized test scores (i.e., SAT, ACT) and other variables with cognitive and non-cognitive attributes such as high school course grades, HSGPA, and class rank (Burton & Ramist, 2001; Galla et al., 2019; Sweitzer et al., 2018). However, these criteria have been criticized as unreliable and potentially biased. In recognition of these shortcomings, scholars have attempted to validate new measures of postsecondary educational promise (e.g., Le et al., 2005; Oswald et al., 2004; Robbins et al., 2006; Sedlacek, 2004; Thomas et al., 2007).

Predictive Validity of Traditional Criteria in College Admissions

Traditionally, the criteria most widely considered in the undergraduate admissions process measure cognitive and reasoning abilities and therefore may be used to assess subject-specific knowledge and skills (Camara & Kimmel, 2005). Research has consistently demonstrated that standardized test scores and HSGPA each contribute to the prediction of academic performance and student persistence (e.g., Allen et al., 2008; Bridgeman et al., 2008; Kobrin et al., 2008; Robbins et al., 2004, 2006; Westrick et al., 2015; Willingham et al., 2002; see also Mattern & Patterson, 2011a, 2011b for a series of reports on SAT validity for predicting grades and persistence). However, previous studies have identified limitations of these variables, suggesting, for example, that SAT predictions may overestimate first-year college GPA and obscure background characteristics that are more accurately predictive of college performance (e.g., Rothstein, 2004; Soares, 2015; Syverson, 2007). Others have identified problematic variability of high school grading standards and course rigor (Atkinson & Geiser, 2009; Bowers, 2011; Buckley et al., 2018; Burton & Ramist, 2001; Camara & Kimmel, 2005; Syverson, 2007; Thorsen & Cliffordson, 2012; Westrick et al., 2015; Zwick, 2002), which may diminish the reliability of HSGPA and high school course grades for high-stakes admissions decisions.

Personal essays are a common component of holistic admissions reviews, offering insight into applicants’ experiences, challenges, goals, and interests (Todorova, 2018). The inclusion of personal essays in college applications is often justified as a means by which applicants can demonstrate character strengths and talents that may not be evident in academic records or standardized test scores. Personal essays may be used to evaluate applicants’ non-cognitive attributes such as creativity and self-efficacy (Pretz & Kaufman, 2017). Although critics argue that personal essay assessments, like standardized tests, may reflect pervasive social class- and race-based inequities (Alvero et al., 2021; Rosinger et al., 2021; Todorova, 2018; Warren, 2013), foregoing sole reliance on high school grades and standardized test scores by including personal essays in admissions decision-making is often championed as a way to expand college access to traditionally underserved students (Hossler et al., 2019).

Predictive Validity of Psychosocial Factors in College Admissions

The use of holistic approaches to undergraduate admissions has emerged as a strategy for expanding the predictive potential of traditional admissions criteria while addressing the disparities these criteria present (Bastedo et al., 2018; Hossler et al., 2019). Incorporating non-cognitive psychosocial factors into the admissions process has the potential to incrementally enhance the predictive power achieved when relying solely on cognitive variables (Allen et al., 2009; Sedlacek, 2004, 2017), as these psychosocial constructs are largely distinct from commonly used cognitive measures (Camara & Kimmel, 2005). Additionally, many psychosocial factors are more malleable than student demographic characteristics and traditional measures of cognitive ability, allowing for the possibility of interventions that may make college success more likely for students once they are enrolled (Allen et al., 2009; Robbins et al., 2004).

Prior research has identified associations between several non-cognitive psychosocial attributes and outcomes related to educational success (Robbins et al., 2006). Research pertaining to specific psychosocial factors has revealed positive associations between self-efficacy and various components of college success, including academic adjustment (Chemers et al., 2001), academic performance (Bandura, 1986; Krumrei-Mancuso et al., 2013; Robbins et al., 2006; Vuong et al., 2010; Zajacova et al., 2005), college satisfaction (Chemers et al., 2001; DeWitz & Walsh, 2002), and persistence and retention (Davidson & Beck, 2006; Robbins et al., 2004, 2006). Conscientiousness, a personality trait that is part of the “Big Five” factor model (Digman, 1990; Goldberg, 1993), has also consistently been shown to predict academic performance to an even greater extent than standardized test scores and HSGPA (Lounsbury et al., 2003; Nguyen et al., 2005; Noftle & Robins, 2007). Additionally, academic self-concept, or how well someone feels they can learn, has been identified as a significant predictor of academic performance, particularly among students from minoritized racial and ethnic groups and low-income backgrounds (Astin, 1992; Bailey, 1978; Gerardi, 2005; Sedlacek, 2004, 2017). Despite inconsistencies among the findings, scholars found coping and attributional styles to be both directly and indirectly associated with outcomes including student motivation, academic performance (LaForge & Cantrell, 2003; Martinez & Sewell, 2000; Rowe & Lockhart, 2005; Struthers et al., 2000; Yee et al., 2003), health status (Sasaki & Yamasaki, 2007), and happiness (O’Donnell et al., 2013). Additionally, prior studies sought to determine the extent to which psychosocial constructs reflect social inequities and thus become potentially biased criteria in college admissions. For instance, evidence suggests that the development of self-efficacy and self-concept are influenced by one’s social class (Easterbrook et al., 2020; Usher et al., 2019; Wiederkehr et al., 2015).

Theoretical Framework

We examined the validity of a measure of several motivational-developmental dimensions designed to predict undergraduate students’ academic performance and degree completion. These dimensions included causal attributions, coping strategies, relevant experiences, self-awareness, self-authorship, self-concept, and self-set goals. These constructs reflect the need for college students to have motivational resources to facilitate effort and persistence as well as developmental maturity to apply these resources adaptively in the context of college (Kaplan et al., 2017). The social-cognitive motivational (Bandura, 1986, 2006) and constructivist-developmental (Kegan, 1994) perspectives provide corresponding complementary theoretical frameworks from which the motivational-developmental dimensions we examined were drawn.

Social-Cognitive Motivational Perspective

The social-cognitive motivational perspective underscores the contribution of the combined influence of students’ high competence perceptions to persistence, adjustment, coping, and performance; attributions of success and failure to internal, malleable, and controllable causes; self-setting of autonomous, challenging, specific, and realistic goals; and coping with difficulties and failure by focusing on analyzing the problem, regulating negative emotions, and applying context-specific strategies (Bandura, 2006). The dimensions we examined include four constructs based on the social-cognitive motivational perspective: self-concept (i.e., individuals’ self-perceptions of ability; Marsh & Martin, 2011); self-set goals (i.e., an individual’s personally determined goals for themselves and for others; Locke & Latham, 2002; Vansteenkiste & Ryan, 2013); causal attributions (i.e., cognitive-affective explanations of the causes of success and failure; Hong et al., 1999; Weiner, 2010); and coping strategies (i.e., purposeful behavioral, emotional, and cognitive actions for responding to situations perceived to challenge an individual’s resources; Compas et al., 2001). These social-cognitive motivational constructs are related. However, they constitute distinct attributes that combine to form an adaptive motivational mindset (Dweck & Leggett, 1988).

Developmental-Constructivist Perspective

The developmental-constructivist perspective reflects the role of cross-contextual capacities for intentional and purposeful self-reflection and self-regulation of knowledge, relationships, goals, and actions related to coping and growth (Kegan, 1994). The motivational-developmental dimensions we examined include two constructs based on the developmental-constructivist perspective: self-awareness (i.e., the ability to consider oneself as an object for reflection, monitoring, and learning; Silvia & Duval, 2001) and self-authorship (i.e., the agentic capacity for an individual to generate and regulate their beliefs, decisions, identity, and social relationships; Baxter-Magolda et al., 2010).

Methodology

Informed by the literature as well as theoretical, empirical, ethical, and logistical factors, Kaplan et al. (2017) advanced the operational definitions of the motivational-developmental dimensions presented in Table 1. These definitions guided the development of an essay-based measure of motivational-developmental dimensions implemented as part of a test-optional admissions policy at a large, public, urban research university located in the Mid-Atlantic region of the United States. We investigated the predictive validity of this measure using multiple regression to analyze data collected from 886 first-year undergraduate students who applied for test-optional admissions and subsequently enrolled at the participating institution. Table 2 presents descriptive statistics on the demographic characteristics of these students.

Table 1.

Definitions of the motivational-developmental dimensions (Kaplan, 2015; Kaplan et al., 2017)

Motivational-developmental dimension	Definition
Attributions	Attributions reflect the characteristics of someone who accredits success and failure to controllable and malleable causes (e.g., effort as well as strategies). Attributions are cognitive explanations of the causes of events in the world that reflect the person’s ontological and epistemological explicit and implicit assumptions. When concerned with the person’s own experiences of success and of failure, attributions have been related to emotions, future expectations of success and failure, motivation, and achievement.
Coping	Coping reflects the characteristics of someone who manifests problem-focused coping that is specific and situation appropriate as well as adaptive recruitment of social supports during experiences or challenge. Coping strategies refer to active and purposeful processes by which an individual responds to taxing events or situations that are perceived to exceed his or her resources. Coping includes behavioral, emotional, and cognitive attempts to manage the demands imposed by a stressor.
Relevant experiences	Relevant experiences refer to characteristics of someone who describes pertinent understandings and life events and is able to explicate realistic and sophisticated connections from these experiences to future imagined college experience.
Self-authorship	Self-authorship refers to the characteristics of someone who manifests agency in decision-making that involves attending to norms and others’ advice as information and a strong personal voice in self-exploration and decision-making. Self-authorship also refers to a comprehensive outlook on one’s role in life as an ‘author’—being active in taking responsibility for choosing beliefs, constructing a sense of self in context, setting a course of action, and forming and influencing social relationships. It manifests in having an internal voice that evaluates and coordinates meaning of experiences about one’s life in areas of beliefs, role in different contexts, and social relationships.
Self-awareness	Self-awareness refers to the characteristics of someone who is self-reflective and aware of the multidimensionality of self and of context. Self-awareness also refers to the state of considering oneself as an object for reflection, examination, and learning. A developed self-awareness involves elaborate reflection on one’s own traits, strengths and weaknesses, thoughts, emotions, behavior, certainties and uncertainties, and changes (e.g., during a given length of time and as influenced by context, situations, and social interactions).
Self-concept	Self-concept refers to the characteristics of someone who is optimistic and confident but not unrealistically overconfident or self-aggrandizing. This dimension also refers to someone who holds a set of adaptive beliefs about success and recognizes their own limitations. Additionally, self-concept encompasses a person’s self-perceptions of ability in a domain (e.g., academics, social, physical). Self-concept relates to an individual’s self-esteem and perceived self-worth. Self-concept also relates to an individual’s self-efficacy and perception that he or she has the ability to organize and execute successfully a course of action required to attain a certain level of performance on a particular task.
Self-set goals	Self-set goals refer to the characteristics of someone who is motivated by specific personal growth goals of learning and developing, intrinsic goals of meaning and interest, extrinsic goals of accomplishment, as well as other-oriented goals. This dimension also refers to the ability to connect past experiences to future imagined college experience. Self-set goals also refer to a personally determined and desired goal, objective, or purpose for the self (self-oriented) and for the effect of one’s actions on others (other-oriented). Adaptive self-set goals are well-defined, relatively specific, realistic, and are defined in ways that allow assessing progress towards their accomplishment. Self-set goals are meaningful to an individual and involve the desire to make a positive difference in the world.

Table 2.

Descriptive statistics on test-optional admissions applicants

	Frequency	Percent
Gender
Female	606	68.5
Male	279	31.5
Race
African American	210	23.7
Asian	97	10.9
Hispanic	89	10.1
Other	84	9.5
White	405	45.8
Pell grant recipient status
Pell grant recipient	391	44.2
Pell grant non-recipient	494	55.8
Median household income
$35,102–$41,909	221	25.0
$41,910–$62,129	234	26.5
$62,130–$80,512	246	27.9
$80,513–$112,844	182	20.6
Academic program
Arts and Humanities	308	34.8
Business and Social Sciences	270	30.5
Health Sciences	113	12.8
Sciences and Mathematics	194	21.9

Procedures

As part of the test-optional admissions process, students provided responses to four short-answer essay questions developed by Kaplan et al. (2017). Table 3 presents descriptions of the essay questions and the primary motivational-developmental dimensions measured within each question. These essays were presented to students as a part of their initial admissions application. Students did not have access to the essay questions in advance and were expected to complete them without preparation. Following several rounds of training, norming, and rubric calibration, two readers used a rubric created by Kaplan (2015) and Kaplan et al. (2017; Table 4) to independently score the primary motivational-developmental dimensions within each essay question according to the following scale: 1 point (does not articulate the dimension), 4 points (narrowly articulates the dimension), 7 points (generally articulates the dimension), 10 points (explicitly articulates the dimension). Each essay question received a total score between 4 and 40 points per reader. A third reader scored an essay response if there was a variance of 5 points or more between the scores produced by the two initial readers on a given essay question. For 11.3% of the essays, a third reader’s score was accepted and the scores produced by both initial readers were rejected. Using this methodology, a total motivational-developmental dimension score (MDS) from 4 to 40 was produced by averaging the scores produced for each essay question. The MDS was included as part of an admissions index that the institution used to make undergraduate admission decisions. This index included a high school academic performance rating (HSGPA and course grades), MDS or standardized test score (depending on whether the applicant applied under the test-optional policy), and an admissions counselor rating. We obtained all student data from the Institutional Research department at the participating institution. Recognizing that college participation and completion varies by family income (Ma et al., 2019), we collected estimated county-level median household income data from the U.S. Department of Commerce Bureau of the Census Small Area Income and Poverty Estimates Program (2022).

Table 3.

Test-optional essay questions and motivational-developmental dimensions

Test-optional essay question description	Motivational-developmental dimensions
Essay 1 required applicants to reflect on past success and sources of pride. Applicants discussed how these experiences might relate to future experiences in college.	Self-concept Self-set goals Attribution of successes Relevant experiences
Essay 2 required applicants to reflect on experiences of failure. Applicants discussed how the experience might relate to future experiences in college.	Self-concept Attribution of failures Coping Relevant experiences
Essay 3 required applicants to reflect on experiences in which they faced a significant challenge. Applicants discussed how they approached this challenge.	Self-awareness Self-set goals Coping Self-authorship
Essay 4 required applicants to reflect on alternatives to college participation. Applicants discussed what they might do instead if not admitted to college.	Self-awareness Self-set goals Coping Self-authorship

Table 4.

Rubrics to assess motivational-developmental dimensions (Kaplan, 2015)

Dimension	Rating
	1	4	7	10
Attribution	Attributes failure/success to uncontrollable (e.g., luck) and/or unchanging causes (e.g., innate ability, difficulty of subject)	Attributes failure/success to a combination of controllable (e.g., effort) and uncontrollable (e.g., ability, domain difficulty) causes	Attributes failure/success to controllable and malleable causes (e.g., effort) but without situation differentiation	Attributes failure/success to controllable and malleable causes (e.g., effort as well as inappropriate strategies), makes differentiated situation-specific attributions
Coping	Helpless, vague, or denial, or avoidant coping (i.e., “it will sort itself out”)	Relies heavily on significant others (e.g., parents, peers), or uses inappropriate strategies	Problem-focused coping, but general and not catered to the specific situation	Agentic, problem-focused coping that is specific and situation appropriate, involves regulation of emotions, adaptive recruitment of social supports
Relevant experiences	No relevant experiences	Describes relevant experiences but with little elaboration	Describes relevant experiences with elaboration	Describes relevant experiences as well as insights and extrapolations to other contexts
Self-authorship	Relies uncritically on others’ prescriptions and/or norms in the situation for decision-making; or manifests conformity to norms	Manifests uncertainty and anxiety about decision-making	Manifests personal voice that is self-centered and does not consider others or the norms in the situation in decision-making	Attends to norms and advice from others as information, manifests a strong personal voice in personal exploration and decision-making
Self-awareness	None	Simplistic, cliché statements	Reflective, but narrow in scope	Reflective, complex, concerns perceptions as well as emotions
Self-concept	None or pessimistic, self-deprecating	Overconfident and unrealistic, or very simplistic	Optimistic but unidimensional, concerns only one domain	Multidimensional, concerns more than one domain, situation-specific
Self-set goals	None, or overly narrow or simplistic	Very general, vague, distal, unrealistic, or narrowly extrinsic	Specific, intrinsic and extrinsic (grades, high paying career), self-oriented	Specific, clear, proximal, realistic, personal-growth and other-oriented

Variables

To reduce the possibility of omitted variable bias, the data we analyzed included a range of student demographic, financial, admissions, and academic information collected as part of the undergraduate admissions process at the participating institution. However, as is consistent with single institution studies, our data did not include all possible variables identified in the literature that may explain our outcomes of interest.

Our predictor variable MDS was a composite score of seven motivational-developmental dimensions (attributions of successes and failures, coping, relevant experiences, self-authorship, self-awareness, self-concept, and self-set goals). UGPA and 4- and 5-year graduation served as our outcome variables. The UGPA variable reflected the cumulative UGPA earned as of a student’s final semester enrolled. Four-year graduation was a dichotomous variable that represented bachelor’s degree completion within eight or fewer consecutive academic semesters. Five-year graduation was a dichotomous variable that represented bachelor’s degree completion in nine or ten consecutive academic semesters.

Based on previous studies that investigated the correlations between student characteristics (e.g., race, gender, and socioeconomic status), academic performance (e.g., HSGPA and UGPA), and baccalaureate degree completion (Mayhew et al., 2016; Pascarella & Terenzini, 2005), we included students’ race, gender, and socioeconomic status (approximated by Pell Grant receipt status and county-level median household income) as covariates in our regression analyses to account for the effects these variables may have on college outcomes. We dummy coded the categorical covariate for students’ race given the five racial groups included in our institutional data. Students with a self-reported race of American Indian, Multiple Ethnicities, Pacific Islander, or Unknown or a status of International were categorized as “Other Race” due to the limited racial representativeness of the sample (see Table 2). Additionally, we retained incongruent designations of race (e.g., African American and White) as these categories reflect those in the institutional dataset. We utilized Pell Grant receipt and a standard score of estimated median household income for the county in which students resided at the time of their application as an approximation of students’ socioeconomic status because Expected Family Contribution (EFC) data were missing for 75 students (8.5%) at the time of their matriculation. We also included HSGPA as a covariate in our regression analyses to account for students’ prior academic performance and an admissions counselor rating of students’ extracurricular activities, personal essay, and high school context. We included these variables in our analyses because of the associations between these commonly utilized admissions criteria and relevant postsecondary outcomes such as undergraduate GPA and graduation (Allensworth & Clark, 2020; Bastedo et al., 2018; Galla et al., 2019; Huang et al., 2017).

The admission counselor rating variable was recorded on a scale of 1 to 10, with 10 reflecting a counselor’s highest positive rating of the applicant. Additionally, we included a categorical variable for students’ academic program at matriculation to account for differences in the rigor and grading standards across academic disciplines and fields (Arcidiacono et al., 2012; Martin et al., 2017). The academic program categories in our dataset included Arts & Humanities, Business & Social Sciences, Health Sciences, and Sciences & Mathematics. Table 5 includes the means, standard deviations, and correlations for all variables.

Table 5.

Means, standard deviations, and correlations for study variables

Variable	M	SD	MDS	HSGPA	Counselor rating	UGPA	4-year graduation
MDS	25.7	2.6
HSGPA	3.4	.35	− .10**
Counselor rating	6.0	1.03	− .092**	− .092**
UGPA	3.0	.70	.13**	.17**	.041
4-year graduation	.5	.50	.14**	.05	.000	.56**
5-year graduation	.7	.48	.12**	.05	.034	.63**	.72**

4-year and 5-year graduation are dichotomous variables where 1 = graduated within specified timeframe and 0 = did not graduate within specified timeframe. Point-biserial correlation coefficients are presented for these variables

**n = 885. p < .01, *p < .05

Data Analysis

We used SPSS version 28 (IBM, 2020) to compute descriptive statistics and to conduct our correlation and regression analyses. We also used Lenhard and Lenhard’s (2014) calculator to compare correlations from independent samples. Prior to conducting our analyses, we examined our dataset to identify systematically missing cases and tested our data to ensure they met the assumptions associated with the analytical techniques we used (see “Appendix” for the results of our assumption tests). We removed five cases (0.6%) for students who matriculated at the participating institution but withdrew before earning course grades in their first semester. Additionally, we removed one case (0.1%) with a missing MDS, one case (0.1%) with a missing HSGPA, and 16 cases with missing admissions counselor ratings (1.8%). Accordingly, our analyses included all cases for which there were complete data. We indicate the analytical sample size for each analysis in table notes.

To test the reliability of the MDS measure, first we computed a Light’s kappa statistic (Light, 1971) to measure interrater reliability between the essay readers as there was not a fixed number of readers for each essay question. We computed Light’s kappa values by calculating Cohen’s kappa and averaging these values across all rater pairs. Second, we computed Pearson correlation coefficients to determine the strength and direction of the associations between the MDS scores and our outcome variables by student demographic characteristics. Third, we compared the resulting correlation coefficients to test for statistically significant differences across student subgroups (Diedenhofen & Musch, 2015; Lenhard & Lenhard, 2014). Specifically, we ran three separate tests to compare the nine correlation coefficients for each of our outcome variables across the student demographic characteristics in our dataset (race, gender, Pell Grant receipt). Fourth, we ran separate regression analyses using interaction terms to determine if the MDS was moderated by student demographic characteristics including race, gender, and Pell Grant receipt status. To create our interaction terms, we centered the MDS to reduce multicollinearity caused by higher-order terms. Lastly, to nuance our findings, we entered the individual motivational-development dimensions as separate variables in stepwise and combined regression models to examine which dimensions, if any, were statistically significant predictors of our outcome variables. We used multivariate linear and logistic regression to investigate the accuracy of the MDS in predicting UGPA and four- and five-year degree completion, respectively. We used the following regression equation to predict our outcome variables:

$$Y_i={\beta }_{\text{i}}{X}_{\text{i}}+{\beta }_{\text{i}}MD{S}_{\text{i}}+{\epsilon }_i,$$

where Y_i is our outcome variable of interest (UGPA, 4-year, and 5-year graduation); β_i is the coefficient of X_i, a given covariate in the model (e.g., HSGPA); β_i is the slope of the line for the MDS, our coefficient of interest; MDS_i is the value of the MDS for student i; and ε_i are the residuals or errors in the model.

Results

Research Question 1

For Research Question 1, we asked whether the MDS is a reliable predictor of UGPA and 4- and 5-year bachelor’s degree completion. The results of our interrater reliability analysis indicated slight agreement between readers across the individual motivational-development dimensions that comprise our measure. Light’s Kappa values ranged from κ = .132 (Coping) to κ = .238 (Relevant Experiences). Table 6 presents these results.

Table 6.

Interrater reliability of motivational-developmental score

Motivational-developmental dimension	Light’s Kappa
Attributions	.191
Self-authorship	.218
Self-awareness	.205
Coping	.132
Relevant experiences	.238
Self-concept	.155
Self-set goals	.146

n = 885. Cohen (1960) suggested kappa results be interpreted as follows: < 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement (Cohen, 1960; McHugh, 2012)

Pearson correlational analyses indicated statistically significant associations between the MDS score and the outcome variables for several student subgroups. We identified statistically significant correlations at the p < .01 level between the MDS and UGPA for female students (r = .12) and Pell Grant recipients (r = .14); 4-year graduation for female students (r = .14) and Pell Grant recipients (r = .17); and 5-year graduation for Asian students (r = .28), male students (r = .16), and Pell Grant recipients (r = .15). Table 7 presents these results.

Table 7.

Correlations between MDS and outcome variables by student demographic characteristics

	UGPA	4-year graduation	5-year graduation
Self-reported race
African American	.11	.16*	.15*
Asian	.16	.21*	.28**
Hispanic	− .17	− .13	− .15
Other race	.20	− .02	− .05
White	.08	.13*	.06
Self-reported gender
Female	.12**	.14**	.09*
Male	.13*	.15*	.16**
Pell grant recipient status
Pell grant recipient	.14**	.17**	.15**
Pell grant non-recipient	.05	.05	.02

n = 885. **p < .01, *p < .05

Given these findings, we tested the equality of these correlation coefficients since they were obtained from independent samples (Cohen & Cohen, 1983; Preacher, 2002). The tests did not identify two-tailed p values less than .05 between all correlation coefficients. By convention, this indicates that the differences between the correlation coefficients are not statistically significant (Cohen & Cohen, 1983).

Our moderator analyses using interaction terms between the MDS and student demographic characteristics (race, gender, Pell Grant recipient) yielded nonsignificant results except for the interaction between Asian students and the MDS for 5-year graduation, p = .009. Table 8 presents these results.

Table 8.

Regression analyses with interaction terms

Variable	UGPA		4-year graduation		5-year graduation
	β	p	P	Exp(β)	p	Exp(β)
Constant	1.235	< .001	.011	.114	.029	.143
African American × MDS	− .014	.471	.278	.936	.126	.910
Asian × MDS	− .026	.341	.054	.833	.009	.784
Hispanic × MDS	.048	.105	.059	1.190	.110	1.168
Other race × MDS	− .045	.180	.394	1.091	.312	1.115
Male × MDS	− .003	.851	.849	.989	.304	.940
Pell grant non-recipient × MDS	− .005	.539	.088	.958	.733	.991
Median household income	2.374	.081	< .001	1.356	< .001	1.381
HSGPA	.390	< .001	.004	1.838	.004	1.869
Counselor rating	.025	.263	.895	1.009	.287	1.080

n = 862 for UGPA, n = 867 for 4-year and 5-year graduation. Reference categories: White, female, Pell Grant recipient

Research Question 2

For Research Question 2, we asked whether the MDS is a statistically significant predictor of UGPA. The MDS was statistically significant in the model (p = .013). However, the small β coefficient estimate (β = .022) suggests that a 1-point increase in the MDS is associated with a .022 increase in UGPA. The full linear regression model was statistically significant, R² = .165, F(13, 861) = 12.876, p < .001, adjusted R² = .152. Table 9 presents these results.

Table 9.

Summary of regression analysis predicting UGPA

Variable	Unstandardized coefficients		Standardized coefficients		R2	F
	β	SE	β	p
Constant	.724	.376			.152	12.876
African American	− .278	.061	− .169	< .001
Asian	− .169	.079	− .076	.034
Hispanic	− .070	.079	− .030	.381
Other race	.028	.083	.011	.739
Male	− .197	.049	− .130	< .001
Pell grant non-recipient	.151	.050	.107	.003
Median household income	.044	.024	.062	.069
HSGPA	.464	.068	.228	< .001
Counselor rating	.022	.022	.032	.323
Business and Social Sciences	− .132	.055	− .087	.017
Health Sciences	− .234	.073	− .112	.001
Sciences and Mathematics	− .305	.064	− .179	< .001
MDS	.022	.009	.082	.013

Reference categories: White, female, Pell Grant recipient, Arts and Humanities

n = 862

SE standard error of β

For robustness, we reran our analysis using stepwise regression to produce a covariate-only model and a model that includes the MDS. The addition of the MDS resulted in a statistically significant change in the F-statistic, F(1, 848) = 6.175, p = .013. This suggests that adding the MDS to the model marginally improved the prediction of UGPA compared to the covariate-only model, ΔR² = .006.

Research Question 3

For Research Question 3, we asked whether the MDS is a statistically significant predictor of four- and five-year bachelor’s degree completion. To nuance the results, we reran our analyses for graduation using stepwise regression produce a covariate-only model and a model that includes the MDS.

Four-Year Graduation

The full logistic regression model to predict four-year graduation was statistically significant, X²(13) = 109.061, p < .001. The model explained 15.8% (Nagelkerke R² = .158) of the variance in four-year graduation and correctly classified 64.6% of cases. Sensitivity was 50.5%, specificity was 45.6%, positive predictive value was 62.9%, and negative predictive value was 33.6%. The MDS was statistically significant in the model (p = .013). An increase in the MDS score was associated with a small increase in the likelihood of 4-year graduation (Exp(β) = 1.076). The addition of the MDS to the regression model resulted in a statistically significant yet minimal contribution to the explanation of the variance in four-year graduation, Nagelkerke ΔR² = .009. Table 10 presents these results.

Table 10.

Summary of logistic regression analysis predicting 4-year graduation

Variable	β	SE	Wald	p	Exp(β)
Constant	− 4.366	1.243	12.339	< .001	.013
African American	− .198	.199	.993	.319	.820
Asian	− .790	.275	8.244	.004	.454
Hispanic	− .120	.257	.219	.640	.887
Other race	.029	.268	.012	.914	1.030
Male	− .377	.163	5.348	.021	.686
Pell grant non-recipient	.448	.164	7.471	.006	1.565
Median household income	.225	.079	8.060	.005	1.252
HSGPA	.831	.225	13.629	< .001	2.295
Counselor rating	− .030	.071	.178	.673	.971
Business and Social Sciences	.079	.178	.196	.658	1.082
Health Sciences	− .303	.237	1.634	.201	.739
Sciences and Mathematics	− .989	.215	21.238	< .001	.372
MDS	.073	.030	6.121	.013	1.076

Reference categories: White, female, Arts and Humanities

n = 867

SE standard error of β

Five-Year Graduation

The full logistic regression model to predict 5-year graduation was statistically significant, X²(13) = 75.464, p < .001. The model explained 11.4% (Nagelkerke R² = .114) of the variance in 5-year graduation and correctly classified 69.2% of cases. Sensitivity was 89.8%, specificity was 32.3%, positive predictive value was 70.4%, and negative predictive value was 63.7%. The MDS was not statistically significant in the model (p = .114). An increase in the MDS was associated with a small increase in the likelihood of 5-year graduation (Exp(β) = 1.05). The addition of the MDS to the regression model resulted in a statistically significant yet minimal contribution to the explanation of the variance in 4-year graduation, Nagelkerke ΔR² = 0.003. Table 11 presents these results.

Table 11.

Summary of logistic regression analysis predicting 5-year graduation

Variable	β	se	Wald	p	Exp(β)
Constant	− 3.529	1.272	7.697	.006	.029
African American	− .374	.202	3.440	.064	.688
Asian	− .450	.259	3.026	.082	.637
Hispanic	− .217	.263	.677	.411	.805
Other race	− .180	.278	.418	.518	.835
Male	− .359	.165	4.740	.029	.699
Pell grant non-recipient	.483	.167	8.372	.004	1.621
Median household income	.228	.081	7.870	.005	1.256
HSGPA	.689	.229	9.060	.003	1.992
Counselor rating	.046	.074	.398	.534	1.047
Business and Social Sciences	.044	.190	.054	.816	1.045
Health Sciences	− .195	.246	.630	.427	.823
Sciences and Mathematics	− .589	.207	8.053	.005	.555
MDS	.047	.030	2.495	.114	1.048

Reference categories: White, female, Pell Grant recipient, Arts and Humanities

n = 867

SE standard error of β

Research Question 4

For Research Question 4, we asked whether any specific motivational-developmental dimensions predicted UGPA and 4- and 5-year bachelor’s degree completion. Among the individual motivational-developmental dimensions we examined, only coping was a statistically significant predictor of UGPA (p = .009), 4-year graduation (p = .014), and 5-year graduation (p = .016). Table 12 presents these results.

Table 12.

Summary of regression analyses with motivational-developmental dimensions as predictors of student outcomes

Variable	UGPA		4-year graduation		5-year graduation
	β	p	p	Exp(β)	p	Exp(β)
Constant	.326	.428	< .001	.003	.002	.014
African American	− .273	< .001	.334	.822	.065	.684
Asian	− .159	.046	.005	.458	.079	.631
Hispanic	− .070	.379	.699	.904	.456	.819
Other race	.034	.686	.852	1.051	.519	.835
Male	− .193	< .001	.023	.685	.037	.705
Pell grant non-recipient	.143	.005	.010	1.537	.007	1.578
Median household income	2.340	.081	.005	1.254	.005	1.262
HSGPA	.471	< .001	< .001	2.367	.002	2.058
Counselor rating	.019	.376	.624	.966	.566	1.044
Business	− .136	.014	.712	1.069	.833	1.041
Health sciences	− .240	.001	.179	.725	.410	.815
Sciences and Mathematics	− .310	< .001	< .001	.361	.004	.550
Attributions	.017	.570	.184	1.147	.355	1.100
Coping	.066	.009	.014	1.226	.016	1.229
Relevant experiences	− .010	.599	.988	1.001	.186	.918
Self-authorship	.013	.550	.974	1.002	.751	1.023
Self-awareness	.015	.527	.381	1.073	.793	1.022
Self-concept	.013	.568	.464	1.058	.102	1.137
Self-set goals	.017	.652	.752	.962	.523	.923

n = 862 for UGPA, n = 867 for 4-year and 5-year graduation. Reference categories: White, female, Pell Grant recipient, Arts and Humanities

Discussion

Summary of the Findings

The results of our reliability tests indicated slight agreement between raters on the scoring of the motivational-developmental dimensions that comprise our measure. This finding suggests that variation in students’ MDS may be a function of variation in raters’ scores rather than true differences across students in the motivational-developmental constructs our measure was designed to assess. This is demonstrated by the slight agreement between raters on the Coping dimension, yet Coping was identified as the only statistically significant predictor of our outcomes of interest. Therefore, readers should interpret our results with caution as potential measurement error may lead to incorrect conclusions regarding the reliability and efficacy of our measure.

Our between-group and moderator analyses did not identify statistically significant subgroup differences between the MDS and our outcomes of interest, nor did we identify statistically significant results when we entered interaction terms of student demographic variables and the MDS into our regression models, except for Asian students. The relationships that do exist are small in magnitude as demonstrated by the correlations between MDS and 4-year graduation (r = .21) and 5-year graduation (r = .28) among Asian students. Despite the absence of statistically significant differences in the correlation coefficients between student subgroups, the MDS is associated with varying levels of validity across student racial groups as suggested by the results of our moderation analysis.

Our findings are generally consistent with prior research that has explored the relationship between non-cognitive variables and student outcomes. For example, using the Student Readiness Inventory, Komarraju et al. (2013) found that the non-cognitive variable Academic Discipline incrementally predicted UGPA over HSGPA and standardized test scores. Additionally, a study of an admissions innovation employing essay-based non-cognitive assessments at DePaul University found that non-cognitive variables helped to predict first-year success and retention, particularly for students from lower income and minoritized backgrounds (Sedlacek, 2017). Consistent with our findings, the predictive power of non-cognitive variables in prior studies was small. Taken together, the similarities across such studies hold that incorporating psychosocial-based assessments remains a promising direction. However, strengthening the reliability of our measure is a necessary first step before it has the potential to effectively promote more holistic, equitable admissions decisions.

Importance of the Findings

Holistic review is an intentional approach for expanding the predictive utility of traditional admissions criteria by considering the non-cognitive characteristics of applicants to make more accurate and equitable decisions about postsecondary educational opportunities (Bastedo et al., 2018; Hossler et al., 2019). However, literature has demonstrated the need for clear and consistent understanding of the validity of non-cognitive factors for predicting students’ success during and beyond college. An effective shift to more holistic admissions processes requires new, validated measures of postsecondary educational promise that meaningfully incorporate psychosocial attributes into admissions models. This is of particular importance as test-optional admissions policies proliferate. To this end, we examined the predictive validity of one such measure of students’ motivational-developmental dimensions.

Our measure did not meet what Cohen (1960) deemed an acceptable threshold of reliability to be considered a valid measure. Although our results show that the MDS makes a small contribution to the explanation of the variance in UGPA and 4-year graduation rates, we do not recommend its use for high-stakes decision-making given the propensity for measurement error in the absence of additional steps to improve interrater reliability. Measures, such as the one used in our study, must demonstrate reliability and consistent predictive validity for all groups of students. Assessments used for high-stakes decision making should be designed and implemented with care to avoid perpetuating inequitable admissions outcomes and presenting barriers in the college admissions process.

Nevertheless, we remain encouraged that the assessment of applicants’ psychosocial attributes may be a worthy component of the admissions process. Assessing non-cognitive dimensions may encourage admissions offices, and by extension institutions, to think holistically about their philosophical bases for admission decision-making (Perfetto et al., 1999) and how these philosophies pertain to their institution’s mission and values. Furthermore, research has demonstrated that even in admissions offices committed to holistic review, officers tend to predominantly rely on traditional academic criteria to make admissions decisions (Bowman & Bastedo, 2018). Although administering and evaluating non-cognitive assessments may require more time and effort from both admissions officers and prospective students, this approach is likely worthwhile if it promotes a more truly holistic, reliable, and accurate assessment of students’ potential.

Limitations

Despite the longitudinal nature of our study, we used degree completion outcome variables that are potentially influenced by a variety of factors not accounted for in our analyses. Consequently, we acknowledge that our study may be subject to omitted variable bias. For example, studies have suggested that variations in students’ tuition expenses net of financial aid affect retention and degree completion rates (Goldrick-Rab et al., 2016; Hossler et al., 2009; Nguyen et al., 2019; Welbeck et al., 2014; Xu & Webber, 2018). However, we endeavored to limit this bias by including variables that allowed us to comprehensively consider factors related to students’ admissions and outcomes, including background characteristics, academic variables, and admissions counselor ratings.

Student motivation and developmental characteristics are not fixed attributes; they are the product of self-reflection and accumulated life experiences (Bandura, 1994) and evolve over time (Mayhew et al., 2016; Pascarella & Terenzini, 2005). Scholars have argued that a holistic evaluation of students' previous academic performance and psychosocial attributes, developed over time in different contexts, cannot be captured by a single assessment (Sedlacek, 2004, 2017). However, our study measures students’ motivation and development at a specific time in their educational careers (i.e., during the college admissions process) and not as longitudinal constructs that may be continuously predictive of behaviors positively associated with educational outcomes.

Compared to the gender and race demographics of college students nationally, the relative homogeneity of the students in our study should frame any interpretation of the findings. For example, to feasibly conduct quantitative analyses using all participants’ data, several races had to be grouped into a single category (“Other”). This grouping may obscure important measurement and educational differences between student subgroups. Additionally, our analytic sample consists only of students who (a) applied for test-optional admissions, and (b) subsequently enrolled at the participating institution. This narrow sample further limits the generalizability of our findings.

Transfer students, graduate students, and denied admission applicants were excluded from the sample. Therefore, our findings may not generalize to these student populations, despite the need to validate predictors of success for students of all types (e.g., transfer students, international students, returning students) and at all levels (e.g., graduate, professional). Additionally, we conducted our study at a single institution located in a specific geographic region and of a particular institutional classification (i.e., public, urban, comprehensive research university). Therefore, our findings may not generalize to other institutional types.

Research has identified associations between essay content, SAT scores, and household income (Alvero et al., 2021). The essay readers in our study were trained to specifically score articulations of the motivational-developmental constructs as opposed to other aspects of analytical writing such as grammar, syntax, and mechanics. We believe this approach allowed us to more accurately capture the dimensions of interest rather than students’ background characteristics such as their socioeconomic status or writing ability. However, because our study used human readers with inherent subjectivity, the reliability of the MDS is subject to their level of agreement on the articulations of each dimension measured within the essay questions.

Implications for Practice and Future Research

Enrollment management leaders and other higher education professionals must weigh the practical significance of accounting for a nominal percentage of the variance in educational outcomes (e.g., UGPA and degree completion) against the introduction of additional requirements in the undergraduate admissions process. Because our results demonstrated slight agreement between readers on the MDS measure and our outcomes of interest are moderated by MDS for Asian students, the MDS should not be used for high-stakes decision making in the absence of other variables with empirically demonstrated reliability and validity. However, we remain encouraged that scores derived from more reliable measures may enrich applicant portfolios undergoing holistic review by providing admissions officers with more comprehensive information about how students may approach and adapt to challenges, insights that are of particular importance at a time when many admissions policies have been disrupted (Bastedo et al., 2022). Still, institutions should be mindful that the use of an essay-based measure in the college admissions process may limit application completion and present workload constraints for admissions officers.

Successfully transitioning to college, especially in times of social and economic uncertainty, requires the possession of coping skills and psychosocial resources that exist apart from a student’s cognitive ability. Therefore, researchers should continue to develop and work to validate measures of non-cognitive psychosocial factors relevant to higher education and related contexts. For example, we believe in the promise of alternative measures of psychosocial factors such as the development of an empirically validated integrated inventory or scale.

While our study found that psychosocial factors made a small contribution to the explanation of UGPA and degree completion, greater explanatory power might be obtained from more reliable measures of applicants’ coping skills and similar variables related to educational success such as perseverance and resiliency. Therefore, future research should examine the predictive value of psychosocial factors concerning additional outcomes that may be both constitutive of and indirectly related to educational success (e.g., mental wellness). Future research should also be conducted to discern the predictive validity of psychosocial factors among diverse student populations in various higher education contexts including different types of institutions and degree levels.

Throughout the development and implementation of our measure, steps were taken to mitigate the effect of bias. For instance, raters responsible for scoring the essays were required to attend several training sessions, throughout which the rubric was calibrated and normed. Furthermore, multiple raters read and scored each applicant essay. While these measures were employed to reduce bias and ensure the reliability of our instrument, continued research on the efficacy of these steps—and on the utility of the personal essay format in general—is warranted. With this in mind, we encourage future research that employs alternative methods of data collection such as situational judgement tests that present prospective students with a realistic scenario they may encounter in college and ask them to indicate how they would respond. Such efforts should be scaled to include multiple study sites to maximize external validity. Finally, future research should examine student outcomes at multiple institutions that have integrated proprietary measures of psychosocial variables into their admissions processes. Such studies could examine whether these measures effectively address the limitations of traditional admissions criteria, reduce predictive bias, and expand postsecondary educational opportunities while allowing the institution to enroll a highly qualified and diverse student body.

Conclusion

Many colleges and universities have adapted their admissions criteria and shifted away from the traditional reliance on standardized test scores as a key predictor of student outcomes. Institutions must also carefully consider which criteria most reliably, accurately, and equitably predict students’ college performance and persistence. Equitable access to postsecondary education and the benefits it confers may be advanced using novel measures that represent not only applicants’ prior academic achievement but their personalities, backgrounds, and the challenges they have overcome. Understanding applicants’ non-cognitive psychosocial attributes such as those assessed using our instrument may contribute to a more holistic understanding of how students can succeed in postsecondary education.

Appendix

Assumption Test Results

Prior to conducting our linear and logistic regression analyses, we tested relevant assumptions. An analysis of standard residuals was conducted, which showed that the data contained no outliers after removing 24 cases (2.7%) for missing UGPA (Std. Residual Min =  − 5.361, Std. Residual Max = 2.019), and one outlier (.11%) for both 4-year (Std. Residual Min =  − 1.81, Std. Residual Max = 2.49) and 5-year graduation (Std. Residual Min =  − 2.34, Std. Residual Max = 1.45). By examining the outlier for graduation, we determined that this student participated in AP and dual enrollment coursework and graduated in less than 4 years. We confirmed that despite the removal of these cases, the resulting analytical sample sizes of 862 for UGPA and 867 for graduation provided adequate statistical power given the number of predictor variables included in our analyses (Tabachnick et al., 2007). The assumption of singularity was met as the predictor variables were determined to not be a combination of other predictor variables. The data met the assumption of independent errors (Durbin–Watson value = 1.942 for UGPA). An examination of correlations identified statistically significant correlations between several of the predictor variables (see Table 5). However, collinearity statistics were within acceptable limits (Tolerance = .705–.966, VIF = 1.035–1.417; Hair et al., 2018). Therefore, we deemed the assumption of multicollinearity to have been met. We determined our data were a good fit for the regression model based on Hosmer and Lemeshow Test (Hosmer et al., 1997) values for 4-year graduation (p = .213) when Pell Grant receipt is excluded from the analysis, and 5-year graduation (p = .221). However, we retained Pell Grant receipt in our regression analysis for 4-year graduation given the strength of the correlation between Pell Grant receipt and 4-year graduation (p = 0.72). Histograms of standardized residuals indicated that the data contained approximately normally distributed errors, as did the normal P–P plots of standardized residuals, which showed that all points were on or near the line. Therefore, the assumptions of normality, linearity, and homoscedasticity were satisfied (Hair et al., 2018).

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Data Availability

The dataset analyzed in the current study is not publicly available due to the proprietary nature of the measure examined. However, the dataset is available from the corresponding author on reasonable request.

Code Availability

Not applicable.

Declarations

Conflict of interest

The authors have no conflicts of interest to declare that are relevant to the content of this article.

Ethical Approval

This study was exempt by the Temple University Institutional Review Board as all data were anonymized and provided to the researchers by the Institutional Research department at the participating institution.

Consent to Participate

Not applicable.

Consent for Publication

The publisher has the authors’ permission to publish the relevant contribution.