The Attack of the Killer Courses: How Course Taking Patterns Affect Retention
Jaclyn Cameron
Research Analyst
DePaul University
jcamero4@depaul.edu
Abstract - Previous research has shown that student preparedness and performance in specific
academic subjects, namely mathematics, can affect student persistence. This research attempts to
go beyond math performance and identify specific courses and subjects that are detrimental to
academic success. Additionally, this research investigates the course taking patterns of both
traditional freshmen and community college transfers in an effort to distinguish unique problem
courses or subjects within each group. Since first year GPA is one of the best predictors of first
year retention, if institutions can better understand what goes into first year GPAs, students can be
better advised on how to navigate through known academic pitfalls. Killer courses are courses
which have a high proportion of students earning grades below C- and/or withdrawing. This
research will identify first year killer courses for entering freshmen and community college
students. Performance in, and number of, killer courses taken will be used to investigate impact on
first year retention. Failing, retaking or withdrawing from courses stalls time to degree and can
have a severe impact on financial aid. Thus, the intent of this research is to better inform the first
year academic experience and how it relates to persistence.
Introduction
Throughout the years retention research has produced a plethora of information about factors that predict
student persistence. One such predictor is that of college academic performance or GPA. Measured either in the
first quarter or the first year, GPA is one of the most influential factors for first year retention (see Adelman, 1999;
NCES, 2001). To many of us in the field, this is intuitive, as most institutions have academic performance criteria in
place that will block student enrollment if the performance standard is not met. This policy in and of itself will
influence GPA as a predictor of retention, or rather, attrition. The other piece of college academic performance is
the accumulation of credit hours. Adelman (1999; 2006) found that continuous enrollment and adequate progress
(which typically means an average of full time enrollment) to be a significant contributor to predicting retention and
degree completion. Other than GPA and taking ample amounts of course credit hours, however, what do we know
in detail about the first year academic performance of students?
For several years now ACT has been doing research about the pre-collegiate academic preparation of
students and how it relates to student retention. Overall, they have found that under-preparedness in math, English,
and science relates to student success and performance in the first year (see www.act.org/research for various
reports). Until now, however, there has been little published research about how previous academic preparedness
translates to college academic performance specifically. Anecdotally, we know that the subjects of math, English,
and science are also known challenge courses in the college curriculum. But what are the specific courses that are
contributing to a high GPA or low GPA for students?
A quick Google search revealed that this urban university was not the only institution using the term “killer
course” (see KSU, Univ. of Hawaii, and NC State). The fairly common definition of a killer course is any course in
which high proportions of students earn non-passing grades or withdraw after the drop date. The implication of
killer courses is that the potential for earning a poor grade is higher than other courses, thus lowering a student’s
cumulative GPA and resulting in no earned credit (if the student fails or withdraws from the course). Most
information pertaining to killer courses is associated with the ways in which students and colleges are implementing
strategies to reduce the failure rate. Such strategies include supplemental instruction, learning communities, parallel
courses, and educational centers such as writing and math labs. The quick scan of the killer course lists that were
provided by the few institutions did show math and science at the forefront of courses. However, depending on the
support, services, and curriculum provided at the institution, a killer course list may look drastically different from
one college to another. Logical reasoning would lead one to believe that if killer courses reduce GPA and can
prevent the earning of credit, then, by default, they would be related to first year retention. Unfortunately, my
search failed to find research to support this assumption.
The purpose of this project is to provide methodological research on the subject of killer courses and
retention. Two populations of students will be investigated in this study; new freshmen and community college
transfer students. These are student groups of interest because they are students with presumably the least
experience in higher education. Thus, their first year in the new 4 year institution is the baseline for continuance
within the institution.
Methodology
The first step of this analysis identified three cohort years of entering community college transfer students
and first time, full time, degree seeking freshmen students. The student cohort rosters were aggregated into one
data set for transfers and one for freshmen. Next, all first year course enrollments, grades, credit hours, and
descriptions were gathered for each student, creating a duplicated student record database. Courses that were
dropped and/or were not lecture style courses were excluded. A crosstab query placed each unique course in a row
and tabulated the number of each letter grade (ranging from A to F and including null values, withdraws,
incompletes, and all other grade categories) earned in the course. Each row was totaled across to create four new
columns; total enrollment for the course (total across all grade columns), total number of grades D, D-, F, or FX
(defined as an unofficial withdrawal and counted as failing), total number of W grades, and total number of Ds, Fs,
and Ws (sum of previous two columns). A percent was calculated to represent the percent of grades earned that
were below C- (all Ds, Fs, and Ws).
To classify courses as killer courses, two criteria had to be met. First, the course had to have a minimum
enrollment of 150 students for freshmen and 100 students for transfers. The criterion differed for freshmen and
transfer courses as the population of community college transfer students is substantially smaller overall than the
freshmen population. The minimums represented an average of 50 and 33 student enrollments, respectively, in the
course per year. Second, the percent of D, F, and W grades had to exceed 10%. Courses that met both criteria were
identified, and then placed in hierarchical order based on the percent of D, F, and W grades earned. Commonly,
only the top ten courses are presented as killer courses, however, for this analysis, all courses which met the criteria
were classified as killer courses.
The list of killer courses for each group of students was next integrated into the original rosters, such that
each instance of a killer course was flagged in the duplicated data set. For each student, the total credit hours for all
killer courses taken were summed, along with the total credit hours of all courses that were not flagged as killer
courses. Additionally, the numeric grade points for all killer courses and non-killer courses were calculated for each
student. The numeric grade points were transformed into grade point averages for each student for both killer
courses and non-killer courses alike. Demographic characteristics were lastly added for each student. The resulting
data sets were unique student record rosters including credit hours of killer courses taken, credit hours of non-killer
course hours taken, total credit hours taken in the first year, grade point average of killer courses, grade point
average of non-killer courses, first year retention (coded as retained or not retained to the second year), and race
coded into underrepresented minority (African American/Black, Hispanic/Latino, and Native American/Alaskan
American) or not. The community college transfer data set also included a variable to distinguish between suburban
community college transfers and city community college transfers as well as the total credit hours transferred in.
Results
Killer Courses
The freshmen data set resulted in 7,226 unique students and 74,892 lecture style course enrollments.
Community college students totaled 3,059, with 22,913 lecture style course enrollments. The freshmen killer course
list, shown in Table 1 below, comprises 22 unique courses primarily in the disciplines of math, history, and biology.
Overall, 32.6% of all lecture course enrollments were killer course enrollments. The average course enrollment was
626 enrollments, with a median of 358 enrollments and the average DFW percent for the killer courses overall was
16%, with a median of 15%.
Table 1. Freshmen Killer Course List
Subject* Crse # Description Total Enroll % DFW
Biology 102 GENERAL BIOLOGY II 281 25.3%
Anthropology 102 CULTURAL ANTHROPOLOGY 1373 24.3%
Art 102 PRINCIPLES OF ART HISTORY 1357 24.3%
History 281 UNITED STATES 1800-1900 749 22.7%
History 219 WORLD HISTORY II 1283 21.8%
Biology 103 GENERAL BIOLOGY III 181 20.4%
Anthropology 103 ARCHAEOLOGY 307 19.9%
Biology 101 GENERAL BIOLOGY I 271 17.3%
History 210 MEDIEVAL PEOPLE: 400-1400 198 15.7%
Political Sci. 150 POLITICAL SYSTMS OF WORLD 1267 15.5%
History 280 UNITED STATES TO 1800 704 15.2%
Math 125 BUSINESS CALCULUS I 1059 14.2%
Math 140 DISCRETE MATHEMATICS I 854 14.1%
History 211 WESTERN EUROPE 213 13.6%
Comptr Sci. 130 INTERNET & WEB 198 11.6%
Economics 106 PRNCPLS. MACROECONOMICS 301 11.3%
Math 110 ALGEBRA FOR APPLICATIONS 409 11.2%
Math 101 INTRO TO COLLEGE ALGEBRA 1275 11.2%
Math 126 BUSINESS CALCULUS II 662 10.9%
Math 131 TRIG / PRECALCULUS 295 10.8%
Economics 105 PRNCPLS. MICROECONOMICS 289 10.7%
Math 151 CALCULUS II 251 10.4%
*Note: Some subject areas have been changed for clarity (e.g. Business Math = Math).
Four different course sequences: General Biology 101, 102, and 103, U.S. History 280 and 281, Business Math 125
and 126, and Economics 105 and 106, made up five of the top eleven killer courses. This may be evidence that
students who begin the sequence performing poorly, will continue through the sequence earning similar grades.
Also high on the list were courses from Art and Anthropology departments. Some might consider this an
unexpected result; however, one possible explanation would be that, depending on the high school curriculum taken
by the students in the courses, this may be the first exposure to these disciplines. With no previous experience in the
subject matter, students may be challenged to develop successful learning strategies for understanding the novel
information.
The list of killer courses for community college transfer students is exhibited below (Table 2). Twenty
courses made the list, primarily in math and math related commerce areas. Killer course enrollments made up
22.7% of all lecture style course enrollments overall. Average course size of the killer courses was 247 enrollments,
with a median of 221, and the average DFW percent was 18%, with a median of 17.4%.
Table 2. Community College Transfer Killer Course List.
Subject Crse # Description Total Enroll % DFW
Accounting 101 INTRO TO ACCOUNTING I 210 28.6%
Math 125 BUSINESS CALCULUS I 358 24.0%
Remedial 204 BASIC APPLIED ALGEBRA 106 23.6%
Math 130 COLLEGE ALGEBRA & PRECALCULUS 282 23.0%
Comptr Sci. 240 INTRO TO DESKTOP DATABASES 105 21.9%
Economics 106 PRINCIPLES MACROECONOMICS 152 21.1%
Finance 310 FINANCIAL MANAGEMENT I 221 19.9%
Economics 105 PRINCIPLES MICROECONOMICS 248 19.8%
Math 101 INTRO TO COLLEGE ALGEBRA 226 19.0%
Continued on next page
Table 2. Continued
Art 102 PRINCIPLES OF ART HISTORY 132 18.2%
Math 126 BUSINESS CALCULUS II 299 17.4%
Economics 315 INTRO TO MONEY & BANKING 123 17.1%
Math 142 BUSINESS STATISTICS 231 16.9%
History 211 WESTERN EUROPEAN 146 16.4%
Political Sci. 120 AMERICAN POLITICAL SYSTEM 142 16.2%
Math 120 QUANTITATIVE REASONING 251 13.5%
English 104 COMPOSITN & RHETORIC II 207 13.5%
Accounting 102 INTRO TO ACCOUNTING II 235 12.8%
Philosophy 100 PHILOSOPHY AND ITS ISSUES 611 12.8%
Dual Listed* 228 BUSINESS,ETHICS & SOCIETY 796 10.9%
History 217 MODERN EUROPE: 1789-PRESNT 113 10.6%
*Note: Denotes a cross listed course. Course offered in subject areas of Religion, Management, and Philosophy.
One observation to make from this list is that, although all class levels of community college students are included
in the analysis, 15 of the 20 courses (including the remedial course, although it is listed as a 200 level course) are
100 level courses, considered freshmen level courses. Several math and math related courses appear on the list,
including accounting, algebra, finance, economics, and calculus. This may suggest that these students had less math
preparation in their previous education experience, which may have contributed to enrollment in community college
prior to a baccalaureate level institution.
It is important to note the similarities and differences between the freshmen and transfer students lists.
Math is the primary subject found on both list. The transfer list, however, consists of more math courses and higher
DFW rates in math courses than the freshmen list. Seven unique courses result on both lists; however, large
differences exist in the DFW rates for each course between freshmen and transfers (see Table 3). Art History is the
only course where freshmen students earn proportionately lower grades than transfer students.
Table 3. Killer Courses for both Freshmen and Transfers.
Description Freshmen: % DFW Transfer: %DFW Difference
PRINCIPLES OF ART HISTORY 24.3% 18.2% 6.1%
PRINCIPLES MICROECONOMICS 10.7% 19.8% 9.1%
PRINCIPLES MACROECONOMICS 11.3% 21.1% 9.8%
WESTERN EUROPEAN 13.6% 16.4% 2.8%
INTRO TO COLLEGE ALGEBRA 11.2% 19.0% 8.8%
BUSINESS CALCULUS I 14.2% 24.0% 9.8%
BUSINESS CALCULUS II 10.9% 17.4% 6.5%
One observable difference is the overall type of courses that resulted for the community college students
compared to the freshmen students. The freshmen list consists of many courses from the college of Liberal Arts &
Sciences that fulfill liberal studies (or general education) requirements, such as Biology, History, and Anthropology.
In contrast, the transfer list includes few courses that would be considered part of the liberal studies curriculum, and
instead is comprised of courses outside the college of Liberal Arts & Sciences. Several courses on the list are entry
level courses taken to fulfill the requirements of a major. A probable explanation for this result is that the transfer
students have transferred in many of the needed liberal studies requirements, thus have moved on to courses within
the intended major. Another difference between the two groups relates to the sequenced courses on the lists. The
sequenced courses on the freshmen list are predominantly organized on the list with the later courses in the sequence
having higher proportions of D, F, and W grades than the earlier courses. On the transfer list, however, the
sequenced courses do not appear on the list in such fluid sequences (e.g. not all math courses within the sequence
appear on the list). Additionally, the courses that are part of a sequence are typically ordered on the list with the
early courses in the sequence have higher DFW rates than the later courses in the sequence. This may be perceived
as a function of transfer students having more experience with courses in higher education, thus are more able to
adapt to new subjects and learning strategies. Likewise, this result may be a function of the different class levels of
transfer students, as more upper level transfer students may be enrolled in the later level courses may have a
tendency to perform better than freshmen level transfers, influencing the proportion of D, F, and W grades.
Predicting Retention
To determine how killer courses relate to first year retention, two separate questions were asked of the data.
The first question asked if the amount of killer courses taken affects retention. The second question explores if
performance in killer courses predicts retention. Separate logistic regressions were run for each group of students on
both research questions. The first set of regressions predicted first year retention (retained or not retained), using
measures of the percent of killer courses taken out of all courses taken within the year, the total number of course
hours taken, total credit hours of non-killer courses taken and race (underrepresented minority or not). The
proportion of non- killer courses taken out of all courses taken was not included in the models due to its high
correlation with the absolute number of courses taken. The second research question predicted retention for only
students who took at least one killer course, thus a killer courses GPA could be calculated. All the variables in the
first regressions were included in the second regressions, in addition to killer course GPA, non-killer course GPA
and the interaction term of total killer courses taken and killer course GPA. The purpose of the interaction term was
to investigate if the amount of killer courses taken impacted performance in those courses, and how, in turn, that
relationship related to retention. Both community college transfer regressions also included the additional control
variables of credit hours transferred in and location of community college (within the city and not). Previous
research with the community college transfer population has shown that students who transfer in more credit hours
are more likely to be retained, and students from the city community colleges are less likely to be retained. The
community college student list was limited to only students who transferred in the fall or preceding summer terms,
so to capture a full year of course taking history.
Regression 1. Predicting retention using course taking activity.
The predictors were added to the regression model in two blocks; the control variables of race, transfer
hours (entered for transfer students only), and location of previous institution (entered for transfer students only)
were first entered, followed by total credit hours taken within the year, total killer course credit hours taken, and
total non-killer course credit hours taken. The analysis for freshmen resulted in an overall fit of the model was
significant with a Chi-Square value of 1399.6 (df = 4, N =6753, p < .05). The Nagelkerke pseudo R2 suggests that
32% of the variance in retention is explained by the model (-2 Log likelihood = 4625.77). The model correctly
predicted 88.4% overall, however, that result was driven by the 98.8% of retained students correctly predicted,
compared to only 35.4% of the non-retained students correctly predicted. The table below illustrates the significant
predictors in the model. Race was the only non-significant predictor. The number of non-killer course credit hours
was significant, whereas an increase in non-killer credit hours taken improved the prediction of cases into retained
Table 4. Variable in the Equation: Freshmen
Variables in the Equation
95.0% C.I.for EXP(B)
B S.E. Wald df Sig. Exp(B) Lower Upper
a Underrepresented Minority (1 = yes) -.007 .093 .005 1 .941 .993 .828 1.192
Non-Killer Course Hours .077 .007 129.213 1 .000 1.080 1.066 1.095
Killer Course Hours .079 .009 78.173 1 .000 1.082 1.063 1.101
Total Hours .080 .006 165.324 1 .000 1.084 1.070 1.097
Constant -4.939 .243 414.795 1 .000 .007
a.
or not retained categories. Killer course hours and total course credit hours were similarly significant predictors, as
positive increases in either variable improved the chance of retention. The significant variables result with low odds
ratios, however, and only increase the odds of being retained by 1.08 units for each unit change in the variables.
The transfer model was significant overall (Χ2 = 399.11; df = 6, N = 1755, p < .05) and accounted for 35%
of the variance (Nagelkerke pseudo R2 = .350; -2 Log likelihood = 1131.47). The percent of correctly classified
cases rose from 84.2% to 86.2 % overall with the addition of the variables in the model. This result was driven by
the high percent of correctly classified retained students (96.1%), compared to only 33.2% of correctly classified
non-retained students. The following table (Table 5) displays the predictors with relevant statistics. Using a
stringent criterion of alpha < .05, race was the only non-significant predictor in the model. However, race was a
significant predictor in the previous step of the model, indicating that it was a valid control variable to use. Location
of the previous institution did not result as significant in the previous step of the model, however, it did result as a
Variables in Transfers.
Table 5. Variables in the Equation: Community College the Equation
95.0% C.I.for EXP(B)
B S.E. Wald df Sig. Exp (B) Lower Up per
a Transfer Hours .005 .002 5.345 1 .021 1.005 1.001 1.009
Location of T ransfer Inst. (1 = City ) -.403 .199 4.107 1 .043 .668 .452 .987
Underrepresented M inority (1 = Yes) .335 .172 3.791 1 .052 1.398 .998 1.960
Non-Killer Course Hours .115 .010 142.117 1 .000 1.122 1.101 1.143
Total Course Hours -.026 .011 6.234 1 .013 .974 .954 .994
Killer Course Hours .079 .012 44.652 1 .000 1.083 1.058 1.108
Constant -1.233 .229 29.077 1 .000 .291
a.
significant predictor in the overall model. Students who came from a school in the city reduced the odds of being
retained by .668. Transfer hours were significant in both steps of the model, such that more transfer hours increased
the odds of persistence. The significance (or lack of) of the control variables in the final model suggests that there
are mediated relationships between the variables of race, transfer hours, and location of transfer institution with
academic performance. The variable that most prominently predicted retention was the number of non-killer course
credit hours taken within the year. In effect, students who took more non-killer course hours were 1.12 times more
likely to be retained. The number of killer course credit hours follows in the order of most to least influential
variables, as a student who took more killer course hours was 1.08 times more likely to persist. Total hours taken
the first year was the third variable in the model, such that the more hours taken by the end of the year, the odds of
being retained decrease by .974 per unit increase.
Regression 2. Predicting retention using course taking activity and academic performance.
The final two regressions built directly upon the previous two regressions. Variables were entered into the
model in three blocks. The first block controlled for race, transfer hours (for transfer students), and transfer location
(for transfer students). The second block additionally controlled for total credit hours taken, killer course credit
hours taken, and non-killer course credit hours taken. The last block added the variables of interest of cumulative
GPA for killer courses taken, cumulative GPA for non-killer courses taken, and the interaction term of killer course
GPA and killer course hours taken. For these analyses, only students who took at least one killer course was
included, as only these students had a GPA for killer courses.
The overall model fit for freshmen students was significant (Χ2 = 851.41; df = 7, 4804, p < .05). A pseudo
2
R of .296 suggested that nearly 30% of the variance within retention was explained by the variables in the model (-
2 Log likelihood = 2967.80). Overall, 89.6% of the cases were correctly classified by the model, with 99% of
retained cases and 29.7% of the non-retained cases were classified correctly. This is an improvement from the
baseline classification before variables were entered of 86.4% correctly classified overall. Table 6 shows how the
variables contributed to the model. The first control variable, race, was significant in the final step, however, it
Table 6. Variables in the Equation: Freshmen with Killer Course Enrollments
Variables in the Equation
95.0% C.I.for EXP(B)
B S.E. Wald df Sig. Exp (B) Lower Up per
a Underrepresented M inority (1 = Yes) -.367 .119 9.512 1 .002 .693 .549 .875
Non-Killer Course Hours -.106 .022 22.578 1 .000 .900 .861 .940
Killer Course Hours -.120 .037 10.485 1 .001 .887 .824 .954
Total Course Hours .227 .021 116.412 1 .000 1.255 1.204 1.308
Killer Course GPA -.054 .101 .289 1 .591 .947 .778 1.154
Non-Killer Course GPA .738 .082 81.874 1 .000 2.091 1.782 2.454
Killer Course GPA by Killer Course Hours .014 .013 1.160 1 .282 1.014 .989 1.039
Constant -5.804 .426 185.450 1 .000 .003
a.
was not significant in the previous steps of the analysis. Again, race becomes a factor when other variables are
involved, and the model suggests that underrepresented minority students were less likely to be retained. Of the
second set of control variables, total course credit hours was the most predictive variable in the model overall. For
each unit increase in total credit hours, a student ass 1.25 times more likely to be retained. Total killer credit hours
and total non-killer hours were also significant, although their relationship with retention was negative. It is a slight
decrease, but the more killer course hours and non-killer course hours taken, the odds ratio used for correct
classification decreased by .89 and .90, respectively. Only non-killer course GPA from the third block resulted as a
significant predictor of retention, and was the second most influential variable in the model. Thus, one unit increase
in GPA contributed to a 2.09 unit increase in the odds ratio to be retained.
The transfer student model was also significant overall (Χ2 = 276.98; df = 9, N = 1282, p < .05).
Nagelkerke’s pseudo R2 = .40, indicating 40% of the variance within retention can be contributed to the variables in
the model (-2 Log likelihood = 573.11). The overall model predicted 91.4% of the cases correctly (98.2% of
retained cases and 32.6% of non-retained cases correctly classified), an improvement from the baseline model with
no variables of 89.7%. The following table (table 7) displays the results of the variables in the model. Transfer
Table 7. Variables in the Equation: Community College Transfers with Killer Course Enrollments.
Variables in the Equation
95.0% C.I.for EXP(B)
B S.E. Wald df Sig. Exp(B) Lower Upper
a
Transfer Hours .013 .004 11.839 1 .001 1.013 1.005 1.020
Loaction of Transfer Inst. (1 = City) -.605 .323 3.498 1 .061 .546 .290 1.029
Underrepresented Minority (1 = Yes) .147 .256 .328 1 .567 1.158 .701 1.913
Non-Killer Course Hours .110 .013 73.030 1 .000 1.116 1.088 1.144
Total Course Hours -.011 .014 .658 1 .417 .989 .962 1.016
Killer Course Hours .003 .038 .004 1 .947 1.003 .931 1.080
Non-Killer Course GPA .674 .147 21.158 1 .000 1.963 1.473 2.616
Killer Course GPA -.164 .197 .693 1 .405 .849 .577 1.248
Killer Course GPA by Killer Course Hours .044 .017 6.475 1 .011 1.045 1.010 1.081
Constant -3.206 .638 25.257 1 .000 .041
a.
hours was the only control variable in first step to remain a significant variable in the final model. The relationship
to retention is positive, as increased transfer hours increased the probability of being retained. The second set of
control variables contributed one variable to the overall model as significant. Non-killer credit hours taken was the
most important variable in the model, such that increased non-killer course credit hours first and foremost increase
the probability of persistence. Of the last block of variables entered both non-killer course GPA and the interaction
between killer course GPA and killer course credit hours taken were significant. Non-killer course GPA, the second
most important variable in the model, was positively related to retention, and one unit increase in GPA increased the
odds ratio by 1.96. The interaction variable was also a positive relationship and the odds ratio increased by 1.05.
The interaction between the two variables is graphed below. Generally, an increase in killer course hours produces a
decrease in GPA, although this is not an absolute progression.
Figure 1. Interaction between Killer Course Credit Hours and Killer Course GPA: Community College Transfers
Conclusions
Freshmen and community college transfer students are vastly different groups in many ways. This research
has uncovered one enrollment characteristic in which these students differ. Freshmen and transfers enter the
institution with different academic challenges, experiences, and pathways. Within the first year of enrollment,
freshmen students are much more likely to have academic difficulties in the general education curriculum,
specifically in areas of math, history, and biology. Conversely, community college transfer students are more likely
enter the institution with some general education requirements already fulfilled. Thus, they are ready to begin their
coursework in major specific areas. Interestingly, however, is the evidence that community college students have
more difficulty in the college entry level math courses than the freshmen students (as measured by the math courses
closer to the top of the list for community college transfer students, whereas the math courses are closer to the
bottom of the freshmen list). The similarities on the two lists helps to expand our general knowledge about student
performance within the first year, yet the differences in the lists focuses our attention towards the separate potential
academic pitfalls of each unique group of students.
Considering the first set of logistic regressions, this research indicates that taking killer courses does have a
slight relationship with first year retention. The first set of regressions, dealing with only academic activity in killer,
non-killer, and total course credit hours taken predicted that freshmen and community college transfer students
would be more likely to retain if more killer, non-killer, and total hours were taken. The interesting difference
between the two groups, however, was the order of importance in which the three variables resulted in the analysis.
For transfer students, non-killer course hours were most influential, followed by killer course hours and total course
hours. Freshmen student retention was most influenced by total credit hours first, followed by non-killer course
hours and killer course hours. A possible explanation for this difference may be due to the likelihood of community
college students to be part time and/or have external obligations. Taking more of the non-killer courses as
compared to the killer courses (which may be considered more difficult), will advance the student at a faster rate.
As this is not their first experience in higher education, transfer students may also maximize the course offerings
such that they take courses that will be counted towards requirements of a major. Referring back to the killer course
lists, the transfer list has many more courses that are discipline specific than freshmen, whose list is predominantly
general education type courses. In contrast, freshmen students may be taking courses that are interesting, yet only
accountable for the general education degree requirements. Considering this is their first time in higher education,
freshmen students may not know what to expect in terms of curriculum requirements, academic demands, or the
personal responsibility needed to succeed. The fact that total hours taken comes out first for the freshmen students
may suggest that it does not matter what type of course the student takes, if the student is challenged to maintain
course hours each term, then they will be less likely to persist. Thus, it may be that transfer students are better equip
to take courses that will maximize their investment in higher education and freshmen students may find that college
is not what they expected or were ready for, therefore regardless of the type of courses taken, will slow their
progress and not return for a second year.
The second set of analyses muddles the overall conclusions that could be made from the first set of
analyses. For the freshmen, the addition of the GPA variables did not change the significance of the course hours
taken variables, however, it did change the directionality of the relationship with retention into a negative. For the
community college transfers, the addition of the GPA variables deleted the significance of two out of the three
previously significant course hour predictors. Non-killer course GPA was the only GPA variable to be significant
for both groups. The transfer group, however, did have the interaction term result as significant, and again, this
might be explained by transfer students having less time to devote to studying, therefore the increase of difficult
courses may decrease performance in such courses, thus influencing or forcing students to leave.
Overall, this research is more likely a proxy for what is already known about first year performance and
student retention. More so than specific courses, the analyses performed may actually be capturing the overall
course performance of the students, such that students who take more courses and perform well in those courses are
more likely to be retained than students who are not taking as many courses and are performing less well. One
important finding of this study was the possible mediated relationships that may exist between demographic
variables and performance. Where one variable is thought to be an influential factor for retention on its own, this
research has hinted that it may be the interaction of characteristics that are better at determining the profile of
attrition and persistence.
This study is limited by the lack of use of several additional variables known to predict retention (i.e. ACT
scores, high school GPA, other measures of academic preparedness, student satisfaction, etc). The inclusion of such
variables in to the models may well supersede the significant predictors found in this research as variables of
importance. The models created in the analyses were much more useful in predicting retention, but were not
adequate for predicting attrition. The classification of non-retained cases would benefit from further inclusion of
variables. Additionally, the killer courses were included in the analysis as individual courses that made up the
group of killer courses. Future research will consider using aggregations of killer courses by department or likewise
creating a killer department list as opposed to a course list. This will broaden the scope of academic challenges
students face instead of using a narrow focus of a specific list of courses.
Knowing the courses at the institution with the highest percentage of D, F, and W grades earned can be
helpful to many aspects of an institution. Advisors can help students choose courses based on their academic
preparedness and inform students about possible challenges with taking multiple killer courses within a term. The
development of a killer course list for different groups is an important step in providing specific services to a
heterogeneous population of students. Killer course lists are also a way to indirectly identify areas of improvement
for student learning, whether that be to investigate supplemental instruction type support or different teaching
approaches. A grade point average is no more than a direct reflection of performance in courses. Just because killer
courses specifically do not predict first year retention, that does not mean that they do not have an effect on the
factors that do predict retention, such as first year cumulative GPA. Much more in depth research must be done to
determine the first year academic triggers that predict retention.
Bibliography
ACT (2004). National ACT Scores Up but Readiness Challenge Continues.
Autumn Activity Report. Retrieved at: http://www.act.org/activity/autumn2004/scores.html
ACT (2004). ACT College Readiness Benchmarks, Retention, and First-Year College GPA:
What’s the Connection? College Readiness Report. Retrieved at:
http://www.act.org/research/policymakers/pdf/2005-2.pdf
Adelman, C. (1999). Answers in the toolbox: Academic intensity, attendance patterns, and bachelor’s degree
attainment. Jessup, MD: U.S. Department of Education.
Adelman, C. (2006). The toolbox revisited: Paths to degree completion from high school through college.
Washington, D.C.: U.S. Department of Education, 2006.
Kennesaw State University, (2004). Hazardous conditions ahead: Roadways to academic success are riddled with
potholes, sinkholes, and sharp turns. Retrieved from:
http://vic.kennesaw.edu/documents/pdf/study/study_academic_challenges_differences_2004.pdf
North Carolina A&T State University, (2006) Web document retrieved from: http://qed.ncat.edu/ir&p/high-fail.htm
U.S. Department of Education. National Center for Education Statistics. High school academic curriculum
and the persistence path through college, NCES 2001–163, by Laura Horn and Lawrence K. Kojaku.
Project Officer: C. Dennis Carroll. Washington, DC: 2001.
University of Hawai’i at Manoa (2005). Distribution of grades: Couses with 100 students or more and “Killer”
courses. Retrieve at: http://www.hawaii.edu/iro/adhoc/gdkillerma06.pdf.