Internet Learning Volume 6, Number 2, Fall 2017/Winter 2018 | Page 14

Better Together: How Blending Course Modalities Impacts Student Persistence
It is striking how consistent
these findings are, showing clearly
first, that online courses—in and of
themselves—do not have a negative
impact on student persistence. Second,
while students who take all of
their courses online show the lowest
historical persistence rate, those who
are mixing modalities show superior
persistence rates than either fully online
or fully on-ground. The differences
between groups are quite interesting
as well:
• For 4-Year institutions, students
who blend showed:

5.65% higher average persistence
than online students
1.08% higher average persistence
than on-ground students
• For 2-Year institutions, students
who blend showed:
19.23% higher average persistence
than online students

6.2% higher average persistence
than on-ground students
• For the Proprietary institutions,
students who blend showed:

9.7% higher average persistence
than online students
2.3% higher average persistence
than on-ground students
The pattern is clear within the
set of studied institutions—blending
students persist at a far higher rate than
fully online students and have modestly
higher persistence rates than fully onground
students.
This begs the question of why
this is so. In order to move beyond
persistence outcomes by course-taking
patterns and begin to investigate the
characteristics of these different student
groups, we can further leverage
the Civitas Learning predictive modeling
platform to identify the particular
mix of data variables for each student
course-taking population.
The most predictive of the several
hundred data elements in the specific
models for each institution can be surfaced.
These variables, both raw and derived,
represent different aspects of the
student’s data record: some are demographic
variables, some are enrollment
patterns such as how close to the start of
the term did the student enroll, financial
aid, etc. Comparing these variable
types—and their predictive rank—can
improve our insight into differences
between course-taking populations. By
way of illustration, Table 2 shows the
top-ranked data features for one of the
study institutions, looking across three
different course-taking populations. As
can be seen, while there is some variability
in the set of features, most of the
difference comes via their ranked position
in the list. For example, Change
in GPA is fourth on the list for the onground
population, but does not appear
in the list for the other two groups. Yet,
Age is on all three lists, but in varying
positions.
13