DU Report Retention A Dillard Specific Regression Model

2011
Retention: A Dillard Specific
Regression Model

wkirkland
Dillard University
9/26/2011

Abstract
Declining student retention has been the subject of serious discussion among

decision-makers at Dillard University during the past two years. The most common

explanation suggests the cause for the low rate centers around the issue of student

academic preparation, especially the academic profile of admitted first-time freshmen.

This study analyzes the impact of nine independent variables in predicting retention for

the entering freshmen cohort group of Fall 2010. Despite expectations that academic

preparation would be a predictor, little evidence is found that standardized test score

(ACT) and/or high school grade point average (HSGPA) have a positive influence on

retention. The opposite is true for ACT composite score; it is negatively related to

retention. HSGPA has no influence. The most potent predictor of retention is the

amount of unmet financial aid need. It is also negatively related to retention but in a

positive way. As the amount declines retention increases. The second best predictor is

academic performance, or first semester grade point average. Thus, the evidence shows

that unmet financial needs play an equal or greater role as academic performance in

predicting retention.

2

Every fall semester, thousands of recent high school graduates flock to college

campuses around the country only to fail to re-enter the following year. During the

summer of 2010 senior administrators at Dillard asked various offices to identify areas in

which they might help improve retention at the university. The Office of Institutional

Research responded by proposing to conduct a retention study specific to Dillard. As

difficult decisions must be made about budget priorities, one latent function of this study

is to provide information to Dillard policymakers about issues driving retention that may

indirectly have budgetary implications. How does attrition affect the institution? Dillard

makes investments in recruiting students, and, when they do not return it loses a

percentage of that cost. What factors may be hindering or promoting retention at Dillard?

During the past two years, Dillard’s retention rate has declined nearly ten

percentage points. Knowing what influences students to return for their second year of

matriculation may be beneficial in numerous ways to administrators seeking to improve

retention. First, it may help them identify the types of pressures incoming freshmen face

during their initial foray into college. Second, it may assist administrators in designing

first year programs specifically tailored to the needs of first-time freshmen at Dillard.

Third, it may help administrators develop proactive strategies for reducing attrition,

including identifying “at risk” students. And, finally, it may point to strategic areas for

efficient and effective deployment of budgetary resources already appropriated to reduce

attrition.

3

Descriptive Differences Between Returnees and Non Returnees

The focus of this study is the Fall 2010 first-time freshmen cohort group. For the

purpose of this analysis, a returnee (retained student) is defined as an individual who

entered the university as a part of that group and re-enrolled in fall semester 2011. A non

returnee is someone from that cohort who did not re-enroll. Dillard University, Office of

Institutional Research tracked the retention of 341 cohort members. Of that group, 226

(66%) returned in fall 2011.

What are some differences between returnees and non returnees? Table 1 reports

descriptive differences between the groups based on academic indicators. Returnees tend

to have significantly higher first semester grade point averages but nearly identical high

school grade point averages and ACT composite scores. Table 2 reports differences by

residence indicators. There is little differences between the two groups on both

indicators. Similar proportions of each group are in-state and commuters. Table 3

compares the two groups by financial aid indicators. Returnees tend to have slightly less

original financial aid need and significantly less unmet financial aid need amount. Based

upon this initial analysis, the large differences in grade point averages and unmet

financial aid need suggest that these two variables may play a significant role in

retention. While the three tables show differences between the two groups, they do not

answer the central question, what are the predictors of retention at Dillard?

Table 1. Retention Status of Fall 2010 First-time Freshmen Cohort by Academic Indicators

High First
School Semester
Grade Grade Act
Point Point Composite
Retention Status Average Average Average

4

Returnee 3.02 2.70 18.60
Non Returnee 2.92 2.05 18.60
N=341
Source: Dillard University, Office of Institutional Research

Table 2. Retention Status of Fall 2010 First-time Freshmen Cohort by Residence Indicators

Percent Percent
Retention Status Commuter In-State

Returnee 52% 66%
Non Returnee 48% 65%
N=341

Table 3. Retention Status of Fall 2010 First-time Freshmen Cohort by Financial Aid Indicators

Average Average
Amount of Amount of
Original Unmet
Retention Status Need Need

Returnee $21,386 $2,527
Non Returnee $22,765 $6,355
N=341

Approach

This study approaches retention from a predictive perspective that assumes there

are factors that have varied and independent influences on retention. It also assumes that

these independent influences exist at the margins. It is not intended to yield a “perfect

solution” to the retention issue, but to provide decision-makers with a framework that

explains some of the forces contributing to the problem At best this framework may

5

assist decision-makers in developing strategies that attack the problem at the margins as a

prelude to getting at the core problem.

A previous study of retention at Dillard, funded by Pew, (Fugar1998) focused on

the issue from a comparative framework looking at differences among students in

learning communities versus non-learning communities. That study focused on retention

at Dillard within a programmatic framework, looking at the effect of a particular program

within a classroom environment.

While this study does not cover the full parameters of retention issues, never the

less, it incorporates many of the assumptions found in traditional predictive retention

models (Porter 1990; McGrath and Braunstien 1997; Deberard, Speilmans and Julka

2004). In addition, it incorporates assumptions based on the understanding and

experiences of Dillard personnel. Finally, the study incorporates an approach that views

Dillard in a unique context as a private “historically black” institution serving an

underserved population with financial challenges. In other words, some things related to

retention may be different from what is assumed in traditional models.

Traditional explanations of student retention have centered on student

achievement and predictors of achievement as relevant variables for study. In keeping

with that approach our model includes high school grade point average (GPA) and ACT

composite score (Daughtery and Lane 1999). Antidotal accounts suggest the relevance of

the traditional approach in reporting retention data to the public. An article referring to

retention at local institutions in Galesburg Illinois stated, “Retention rates at three local

colleges are linked to admission requirements and average ACT scores, school officials

said Monday” (Essig 2010 p. 1).

6

Others have approached the problem from a personal perspective- that is- focus

on the role personality and personal behavior plays in influencing retention (Lu 1994;

Musgrave-Marquart et al. 1997; Jeynes 2002). Time and resources were not sufficient to

incorporate this aspect into this study. Such an approach would have required the

selection of a sample and the development and distribution of a survey instrument. Never

the less, the personality approach is widely used

In addition to variables used in traditional models, this model tapped the

experience of Dillard staff members. Some members from the first year program, over

the years, have consistently alluded to their feeling that there are differences between in-

state and out-of-state students as well as between commuter and residential students.

Staff in the Office of Records and Registration suggested that credit hours attempted may

be affecting retention, They noted the high credit hour load taken in the first semester by

first-time freshmen. Officials here appear to share the same view of officials from

Temple University’s enrollment management office; they indicated that the credit hours

attempted variable was a major player in getting students to re-enroll (Scannnell 2011).

A third set of variables were incorporated to account for the unique context in

which Dillard students matriculate. Predicting academic success for African-Americans

has usually focused on retention within the context of majority institutions (Seidman

2007). Traditional models may miss the unique features of understanding retention in a

homogenous predominate African-American setting. Consequently, our model takes

this into account by focusing on the role financial aid may play in retention. National

financial aid data indicate that 65 percent of all undergraduates receive financial aid and

79 percent of full-time/full year students receive aid (National Center for Educational

7

Statistics, 2009). On the other hand, 94 percent of first-time freshmen enrolled at Dillard

in 2009-2010 received financial aid (Dillard University, 2011). Therefore three financial

aid indicators are included in our model.

Retention Model

A retention regression model specific to Dillard was developed to predict

retention of first-time freshmen. The model includes nine independent variables. They

are: (1) in state versus out-of-state, (2) first semester grade point average, (3) hours taken

first semester, (4) on campus versus commuter, (5) high school grade point average (6)

ACT composite score, (7) original financial aid need amount, (8) unmet financial aid

need amount, and (9) percent of unmet financial aid need. The independent variables in-

state and off-campus are treated as dummy variables. In-state and off-campus students

are coded 1 and out-of-state and on-campus students are coded 0. The dependent

variable retention is also treated as a dummy variable. Persons who returned were coded

1 and non returnees were coded zero,

Variables Influencing Retention at Dillard

Three variables are found to influence retention at Dillard. They are: first

semester grade point average, ACT composite score, and amount of unmet financial aid

need. The most potent predictor is the amount of unmet financial aid need (beta weight

-.368) followed by grade point average (beta weight .324) and ACT score (beta weight

-.176). While grade point average is positively related to retention, unmet need and ACT

score are negatively related to retention. In other words, for every unit increase in

retention there is an increase in grade point average. On the other hand, for every unit

increase in retention unmet need decreases, the same can be said about ACT score,

8

although the latter has one-half the predictive value. As ACT scores decline retention

increases. The unstandardized coefficient identifies the threshold at which unmet need is

likely to influence retention. As one moves from the category non returnee to returnee

the amount of unmet need declines by $3,257. The remaining six independent variables

have little influence on retention and fail to reach statistical significance. For a detailed

table of the regression results see Appendix A.

Our findings corroborate previous research findings yet ours also differs from

them significantly. That fact is substantiated in other published material:

According to University Business Magazine, “the research shows there are a
number of other drivers that influence re-enrollment trends.” It further states, “First and
foremost is the level of academic success (e.g., term 1 GPA) followed by variables such
as entry qualifications (GPA in high school, standardized test scores, etc,) residential
versus commuter status; attempted hours; participation in intercollegiate athletics or other
extracurricular activities; gender; and race. Variables such as amount of borrowing,
unmet need, and level of grant sometimes emerge as statistically significant variables in
predictive retention models, but their influence on behavior is often minor” (Scannell
2011 p.1).

Our results show that GPA is a significant predictor. On the other hand, in

contrast to other findings, unmet financial need is the best predictor while standardized

test is a weaker predictor. The results raise the question, why is there a negative

relationship between ACT score and retention? This seems counter intuitive. Evidence

presented earlier in Table 1 showing returnees and non returnees with identical average

ACT score may hold a clue. This suggests that high achievers are returning at the same

rate as low achievers. Perhaps higher achieving students have high expectations that are

not being met by the institution. No doubt, this issue needs further study.

Why is there a weaker than expected relationship between ACT score and

retention? The answer may be related to the inherent nature of the relationship between

9

GPA and retention that may reduce the influence of ACT score. ACT score impact on

retention is probably indirect as evidenced by its correlation (.352) with first semester

grade point average (see Appendix B). If one considers the intuitive nature between

ACT score and retention versus that between grade point average and retention the

surprise may wane. In fact, retention is a function of grade point average. If one fails to

obtain a specific level one is dismissed by the institution. On the other hand, a low ACT

score is likely to affect ones admission to the institution, but will not result in a student’s

dismissal after enrolling.

Conclusion

This report began by asking what factors influence retention at Dillard. After

developing and analyzing a regression model specific to Dillard, it is clear that the model

did not identify a “silver bullet” to explain retention. Never the less, it identified unmet

financial aid need as the most potent predictor in the model. This is contrary to national

trends. That in itself probably validates the need to use a Dillard specific predictive

model when approaching retention.

The potential budgetary ramifications exposed by this study are significant. Non

returnees were awarded more than $1.9 million in aid during their matriculation. One

may infer that much of the aid followed the students when they left. What if fifty percent

of them had returned? As the proportion of first-time freshmen in need of some type of

financial aid at the institution consistently hovers at 90 percent or above, and student

attendance is sensitive to financial aid needs, unmet need will probably continue to

influence retention in a significant way.

10

What options do students with substantial unmet need have? The most plausible

answer is probably the need to fill that gap in order to remain in school. Those who are

able to do so stand an increased chance of continuing their matriculation. Those who are

unable to do so may find it difficult to remain at the institution. Those who stay may fill

the gap by securing employment on a full-time or part-time basis or securing more aid.

Consequently, any future efforts aimed at stabilizing or increasing retention may need to

incorporate strategies that address this issue.

As traditional strategies focusing on academic success appear to be the

predominant approach at Dillard for addressing retention, and the model provides validity

for continuing this approach, perhaps it needs to be broadened to include a co-equal

strategy that focuses on unmet financial aid need as well. The institution has long

employed the tactic of “early warning” based on academic performance as an

intervention strategy. Perhaps now is the time to implement a tandem process that

focuses on both issues.

Students with high levels of unmet need may require as much monitoring as those

with academic issues. This may require decision-makers to re-think current retention

strategies and include tactics that allow for flexible and expanded class schedules. Rigid

schedules may preclude these students from seeking or obtaining employment. A second

tactic may include targeting institutional need based grants to at risk students. Those with

sufficient grade point averages but high levels of unmet need may benefit most from such

an effort. Given current budget constraints, the university may have to consider re-

directing resources to more effective strategies.

11

The study results also provide the university with the opportunity to be more

specific in exploration of grant opportunities related to retention. Now that it is known

that certain factors influence retention at Dillard the institution is in a better position to

articulate its retention needs to agencies that fund retention initiatives.

At this point a handful of ideas have been promulgated; it is expected that

officials from various entities across the campus may use the results presented in this

study to develop and launch an array of retention initiatives. Those efforts may result in

the development of novel new strategies to address the problem. If and when those

strategies are implemented they may create the need for a continuous monitoring

mechanism to assess and evaluate the effectiveness of those programs.

This study represents the first step in spurring attempts to find a solution(s) to the

recent decline in retention. Perhaps in the future retention studies on Dillard’s student

population may be expanded to focus on individual personal behavior.

12

Coefficientsa

Model
Unstandardized Coefficients
Standardized Coefficients
t
Sig.

B
Std. Error
Beta

1
(Constant)
.877
.260

3.373
.001

High School GPA
.032
.049
.034
.651
.516

Instate
-.016
.054
-.016
-.298
.766

Residency
-.040
.051
-.042
-.776

14

Appendix B

Correlations
High
ACT Original Unmet School
Hours Comp GPA Need Pack Need Residency Instate GPA
Hours Pearson
Correlation 1 .339(**) .203(**) -0.065 -.141(**) 0.030 0.055 .177(**)
Sig. (2-tailed) 0.000 0.000 0.237 0.010 0.587 0.315 0.001
N 341 341 339 335 335 341 341 341
ACT Pearson -.220(**
Comp Correlation .339(**) 1 .352(**) -.139(*) 0.006 0.004 .299(**)
)
Sig. (2-tailed) 0.000 0.000 0.000 0.011 0.912 0.939 0.000
N 341 341 339 335 335 341 341 341
GPA Pearson
Correlation .203(**) .352(**) 1 -0.080 -.262(**) 0.047 0.000 .332(**)
Sig. (2-tailed) 0.000 0.000 0.147 0.000 0.389 0.995 0.000
N 339 339 339 334 334 339 339 339
Original Pearson
Need Correlation -0.065 -.220(**) -0.080 1 .296(**) 0.065 0.083 -0.061
Sig. (2-tailed) 0.237 0.000 0.147 0.000 0.235 0.127 0.268
N 335 335 334 335 335 335 335 335
Unmet Pearson -.141(** -.262(**
Pack Correlation -.139(*) .296(**) 1 -0.053 -0.055 -0.064
) )
Need Sig. (2-tailed) 0.010 0.011 0.000 0.000 0.338 0.314 0.245
N 335 335 334 335 335 335 335 335
Residency Pearson
Correlation 0.030 0.006 0.047 0.065 -0.053 1 .486(**) 0.011
Sig. (2-tailed) 0.587 0.912 0.389 0.235 0.338 0.000 0.845
N 341 341 339 335 335 341 341 341
Instate Pearson
Correlation 0.055 0.004 0.000 0.083 -0.055 .486(**) 1 -0.080
Sig. (2-tailed) 0.315 0.939 0.995 0.127 0.314 0.000 0.143
N 341 341 339 335 335 341 341 341

15

High Pearson
School Correlation .177(**) .299(**) .332(**) -0.061 -0.064 0.011 -0.080 1
GPA Sig. (2-tailed) 0.001 0.000 0.000 0.268 0.245 0.845 0.143
N 341 341 339 335 335 341 341 341
**. Correlation is significant at the 0.01 level (2-tailed).
*. Correlation is significant at the 0.05 level (2-tailed).

References

Daughtery, T.K. & Lane, E.J. (1999). A longitudinal study of academic and social
predictors of college attrition. Social Behavior and Personality, 27 (4) 355-362.

DeBerard, M.S., Spielmans, Glen I., Julka, D.L (2004). Predictors of academic
achievement and retention among college freshmen: a longitudinal study. College
Student Journal (March 2004).

Dillard University 2011 IPEDS Financial Aid Survey.

Essig, C. (2010, October 12). Local colleges’ retention rates way above average.
Galesburg.com. Retrieved from http:www.galesburg.com/newsnow/

Fugar, C. V. (1998). Student retention, progression and academic performance at Dillard
University. Unpublished.

Jaynes, W.H. (2002). The relationship between the consumption of various drugs by
adolescent and their academic achievement. American Journal of Drug and Alcohol
Abuse, 28 (1), 15-35.

16

Lu, L. (1994) University transition: major and minor stressors, personality characteristics
and mental health. Psychological Medicine, 24, 81-87.

McGrath, M. & Braunstien, A. (1997). The prediction of freshmen attrition, An
examination of the importance of certain demographic, academic, financial, and social
factors. College Student Journal, 31 (3), 396-408.

Musgrave-Marquart, D., Bromley, S.P., Dalley, M.B. (1997). Personality, academic,
attribution, and substance use as predictors of academic achievement in college students.
Journal of Social Behavior and Personality, 12 (2), 501-511.

National Center for Educational Statistics (2009).

Porter, O.F. (1990). Undergraduate completion and persistence at four-year colleges and
universities: Detailed Findings, Washington, DC: National Institute of Independent
Colleges and Universities.

Scannell, J. (2011). The role of financial aid and retention. Retrieved from
http://www.universitybusiness.com/

Seidman, A. (2007) Minority student retention. Amityville N.Y.: Baywood Publishing
Co., Inc.

17

DU Report Retention A Dillard Specific Regression Model

More Related Content

Similar to DU Report Retention A Dillard Specific Regression Model (20)

More from Dillard University Library (20)

Recently uploaded (20)

DU Report Retention A Dillard Specific Regression Model