A student generated expressive writing program and the principles of finance course.

Willey, Thomas ; Willey, Liane Holliday

INTRODUCTION

Beginning finance courses, comprised largely of business majors, are typically challenged by a lack of background in the field, a complex and analytical subject matter and often a lack of interest and motivation (particularly for those who are not finance majors). To further complicate matters, beginning principles classes are frequently taught in large lecture rooms and to large numbers of students, thus decreasing the possibility of discussion, student/teacher interaction, and consequently, learning, interest and motivation. In an effort to diminish the effects of these variables, the authors of this study sought to make improvements in instructional strategies. Specifically, the authors empirically investigated the effects of a student generated expressive writing program on overall performance in a Principles of Finance class.

Writing in the classroom is not new, in fact, students have probably always taken written notes as a way to record teacher directed content information. But in recent years, research in the field of comprehension has found that the act of writing to learn information can be far more sophisticated than simply jotting down what the teacher has said. In a student generated expressive writing program learners are taught and encouraged to write out any questions they might have, concepts they need to review, and any additional supportive examples or elaborations they can think of. When they do these types of writings, they are more apt to apply critical thinking skills and a higher order of understanding to their general knowledge base. Apparently, this is the case because student generated expressive writing: 1) focuses thought; 2) makes thoughts available for introspection and further analysis; 3) translates mental images into more concrete pictures; and 4) motivates inter-communication when the teacher responds to the students writing (see Britton, et. al. (1975), Emig (1977) and Draper (1982)). While the first three factors serve as an intrapersonal study strategy, wherein students work and rework their notes and their confusing and clarifying thoughts as they relate to those notes, the last factor serves to build an interpersonal relationship between the instructor and the students. All of these factors work in combination to build a schema of better understanding and hopefully more motivation to do well in the course, because the student no longer feels isolated.

Though research in writing-to-learn has found its' way across most curricula, (see Fulwiler (1980) and Willey (1989)), it is still very new in quantitative courses. A review of literature found little theoretical and no empirical research related to finance courses and writing-to-learn factors. Dahlquist (1995) suggests properly designed and implemented writing projects can increase the communication skills and the overall learning of finance students. Templeton (1996) presented an overview of the potential benefits and costs of using writing exercises in a finance course. Ciccotello and Green (1997) investigated the impact written case analysis had on undergraduate and graduate finance students. Using survey results, the two authors found undergraduates viewed case writing less favorably than graduate students, citing that the process was more time consuming and frustrated than expected. However, the majority of the empirical research has been done studying the effects of writing in statistics and mathematics.

Specifically, Watson (1980) found that students who regularly wrote in "learning logs" or journals showed improved ability in quantitative problem solving. And Smith, Miller and Robertson (1991) found students who completed regular writing assignments in a statistics class showed improvement in their learning when compared to a non-writing control group. In light of these positive findings, it seems important to continue the investigation of the effects of writing on students' achievement in a Principles of Finance course. This article empirically examines the performance on exams of a student generated expressive writing program versus a non-writing control class.

METHODOLOGY

Fifty-seven undergraduate students enrolled in two Principles of Finance classes at a regional midwestern college composed the sample groups of this study. There were thirty-three students in the experimental group and twenty-four in the control group. The same instructor taught both sections during normal daytime class meetings. Following the completion of each chapter, the students in the experimental group were directed to write answers to the following questions:

1 Pretend you are teaching this class. Prepare an explanation of the
 major concepts from this chapter.

2 What concepts are unclear to you?

3 What concepts would you like to review?

The papers were then collected and reviewed by the instructor, who then made brief, but appropriate comments in one or more of the following domains: good job, needs improvement, seek further tutoring assistance. The papers were returned by the following class period. Also, the instructor took the opportunity to record major classroom errors in comprehension and adjust subsequent lectures to correct or review missed or confused areas. To facilitate completion of this work, the students were told that they would receive participation points totaling 10% of the overall grade, if they completed each writing assignment. The students in the control group received traditional lecture-based instruction with 10% of their total grade derived from problem sets.

HYPOTHESES

Six principal hypotheses were tested:

H01 There is no difference between the mean G.P.A. for the experimental and control groups.

The first hypothesis tests whether the difference in the mean G.P.A.s is statistically different from each other. A statistically significant difference in these values will imply that the samples are not similar, and the results could be biased due to non-homogeneous groups.

H02: There is no interaction effect between writing assignment and student quality on exam scores.

This hypothesis tests for the existence of an interaction effect, which if present, causes the primary impact of the writing process to be unclear. If the null hypothesis is not rejected, the research can focus on the main treatment effect of the writing program.

H03-06: There is no difference between the overall performance on exam scores between the control and the experimental groups.

These four hypotheses will gauge the effectiveness, or lack of effectiveness, of the student generated expressive writing program as measured by the difference in exam scores.

Table 1 shows the summary statistics of the four exams for the experimental and control groups. The mean student beginning of the semester G.P.A., which is expected to be positively correlated with student performance (Chan, Shum and Wright (1997); Heck and Stout (1998)), showed the average G.P.A. for the experimental group to be 3.54% higher than that of the control group (2.926 versus 2.826). The degree of dispersion, as measured by the standard deviation of average G.P.A., between the two groups was within 1.76% of each other (0.578 and 0.568 for the experimental and control groups, respectively). The coefficient of variation (CV), a relative measure of dispersion, for the experimental group was 0.198, as compared to the CV for the control group of 0.210, which shows the variability of the G.P.A. for the two groups to be essentially the same. These summary results indicate an approximately identical student quality composition, as measured by G.P.A., in each group.

The mean exam scores were very similar for the four exams. Exam 1 covered the finance function, financial statements, cash flow and ratio analysis. Exam 2 included the time value of money and valuation. Exam 3 covered risk and return and capital budgeting. The topics on Exam 4 were short-term and long-term financing and the cost of capital. The differences between the experimental and control group means were +0.19 for Exam 1, -3.10 for Exam 2, +3.77 for Exam 3 and -0.49 for Exam 4. Further hypotheses testing (H03-06) will gauge whether these differences are statistically significant from each other.

The amount of absolute dispersion for three of the four exam scores was very small between the two groups; 1.27 for Exam 1, 0.74 for Exam 3, and 0.24 for Exam 4. Additionally, the CVs between the two groups for the above three exams are virtually identical. The experimental and control CVs for Exam 1 were 0.09 and 0.11, Exam 2 both groups were 0.20 and Exam 4 both groups had a CV of 0.17. However, Exam 2 scores for the experimental group showed more variability with a standard deviation of 15.66 and a CV of 0.23 as compared to 10.03 for the control group and CV of 0.14, an absolute value difference of 5.63.

RESULTS OF ANOVA TEST FOR G.P.A. (H01)

The purpose of the one-way analysis of variance (ANOVA) is to test for the equality of mean G.P.A. score between the experimental and the control group. The results of this test allow further testing of the remaining hypotheses. Since the variance ratio (F) as reported in Table 2 is 0.42 (p value of 0.52), the null hypothesis should not be rejected. This indicates that the variation in G.P.A. for both the experimental and control groups is not statistically different from each other. In fact, the variance ratio is close to one, meaning that the amount of variation between the experimental and control groups is approximately equal. The significance of this test is that if, for example, the mean G.P.A. for the control group had been higher than that of the experimental group, then interpreting H03-06 would be very uncertain or ambiguous. For example, the difference in performance on exam scores could have been a result of the higher mean G.P.A. of the control group.

TESTING FOR AN INTERACTION EFFECT (H02)

This hypothesis was examined using the multiple regression equations used to test for H03-06. Using these equations, a statistically significant result for the interaction variable (Dummy * G.P.A.) would indicate an interaction effect. Results of the four regression equations showed no existence of an interaction effect (Tables 3 through 6). The p values for the exams were, respectively, 0.443, 0.258, 0.302 and 0.772. Therefore, the null hypothesis (H02) was not rejected for any of the four exams. This allows a clearer interpretation of the main treatment effects of the writing program. If the null hypothesis was rejected, indicating a statistically significant interaction term, most of the interpretation would fall on the interaction effect, as opposed to the main treatment effect.

TESTING FOR THE EFFECT OF THE WRITING PROGRAM (H03-06)

Table 3 shows that, with the exception of G.P.A., none of the independent variables were statistically significant as predictors of test performance on the first exam. G.P.A. was significant at a p value of .001, meaning that teaching style, student expressive writing or the lecture only method of teaching, did not explain performance on first exam, only G.P.A. The positive sign on the G.P.A. coefficient in the equation shows that the higher (lower) the G.P.A, the better (worse) the exam score. The Adjusted R2 indicates that 24% (F value = 6.88 and p value of 0.001) of the variability of test scores can be explained by the model. In other words, 76% of the variation in scores on Exam 1 was due to other important, but unmeasured, factors. One possible variable for inclusion might be the time of day the class is taught.

Results of Exam 2, as shown in Table 4, were similar to those of Exam 1. The G.P.A. variable was the only statistically significant factor in performance (p value = 0.086) and the explanatory power of the model declined slightly from 24% to 22%. The dummy variable for the writing process and the interaction effect variable were not statistically significant.

Table 5 shows that Exam 3 results deviated from the pattern of the first two exams. On this test, none of the independent variables were statistically significant. A possible explanation of the relative poor performance of the model (Adjusted R2 of the regression model of 6.7%), may be due to the content covered on the exams. Two of the traditionally more difficult subject areas in the course, the risk-return tradeoff and capital budgeting, were included on this exam and may have contributed to the lower results.

Table 6 shows the regression analysis results for Exam 4 reverting to the pattern of the first two exams. The p value for G.P.A. was .015, with the dummy and interaction variables showed no statistical significance (p values of .697 and .772, respectively). The independent variables explained 21.5% of the variation in the Fourth Exam score.

DISCUSSION

Results of this study were disappointing, but could be at least be partially attributed to a series of uncontrollable extrinsic variables. First, sample sizes were compromised when fewer students than expected enrolled in the two groups, and more students than expected dropped the control group course. As a result, there was a limit on the number of variables that could have been added to improve the model. This was a very unfortunate occurrence, because of the nature of the treatment and the difficult level of the course. The inclusion of additional independent variables such as: 1) ACT/SAT scores; 2) Gender; 3) Major field of study; 4) Indication of attitudes about writing; may have led to more promising results. Second, there existed an internal validity threat from the possibility of diffusion effects. Since the experimental group had their class at 9.30 a.m. and the control group had their class at 2.00 p.m., it was possible for diffusion of knowledge gained from writing assignments and of test questions to have taken place.

Despite the statistically insignificant results, research in this area should be continued. Intuitively, it would seem that any treatment that encourages high level thinking should be effective. There is a distinct possibility that a student expressive writing program should work, unless students are intrinsically opposed to writing and extrinsically motivated to feel it is simply too much extra work. Admittedly, there is little professors can do to overcome the former concern, but perhaps the latter thought could be reduced. Some possible strategies are the use of extra-credit points with the assigned writings or devoting a portion of class time to work with peers and/or the instructor in completing the assignment. If the above concerns are addressed in future research, we contend that the program has the potential to be an effective tool for increasing student learning.

REFERENCES

Bloom, B. (1980). The new direction in educational research: Alterable variables. Phi Delta Kappan, (February), 382-385.

Britton, J., T. Burgess, N. Martin, A. McLeod & H. Rosen (1975). The Development of Writing Abilities, London: Macmillan, 11-18.

Chan, K. C., C. Shum & D. J. Wright (1997). Class attendance and student performance in principles of finance. Financial Practice and Education, 7(2), 58-65.

Ciccotello, C. S. & S. G. Green (1997). Student-authored case studies in finance: Performance and observations. Journal of Financial Education, 23(Spring), 55-60.

Dahlquist, J. (1995). Writing assignments in finance: Development and evaluation. Financial Practice and Education, 5(1), 107-112.

Draper, V. (1982). Formative writing: Writing to assist learning in all subject areas. In Gerald Camp (Editor), Teaching writing: Essays from the bay area writing project, Berkeley, CA: Boynton/Cook, 148-183.

Emig, J. (1977). Writing as mode of learning. College Composition and Communication, 28, 122-128.

Freeman, M. & M. Murphy (1990). The write thing in the mathematical sciences. Mathematics and Computer Education, 24(2), 116-121.

Fulwiler, T. (1980). Journals across the disciplines. English Journal, 69(9), 14-19.

Ganguli, A.B. (1989). Integrating writing in developmental mathematics. College Teaching, 37(4), 140-142.

Geeslin, W. E. (1977). Using writing about math as a teaching technique. Mathematics Teacher, 70(2), 112-115.

Heck, J. L. & D. E. Stout (1998). Multiple-choice vs. open-ended exam problems: Evidence of their impact on student performance in introductory finance. Financial Practice and Education, 8(1), 83-101.

Howard, J. (1984). Recognizing writing as the key to learning. Education Week, (September 5), 12-13.

Smith, C.H., D. M. Miller & A. M. Robertson (1991). Using writing assignments in teaching statistics: An empirical study. Mathematics and Computer Education, 25(1), 21-33.

Templeton, W. K. (1996). Beyond the numbers: Writing to learn finance. Journal of Financial Education, 22 (Spring), 56-60.

Watson, M. (1980). Writing has a place in a mathematics class. Mathematics Teacher, 73 (October), 518-519.

Willey, L.H. (1989). The effects of selected writing-to-learn approaches on high school students' attitudes and achievement. (Doctoral Dissertation, Mississippi State University, 1988). Dissertations Abstracts International, 49.

Thomas Willey, Grand Valley State University

Liane Holliday Willey, Grand Valley State University

Table 1. Summary Statistics of G.P.A. and Exam Scores

 G.P.A. Exam 1 Exam 2

Experimental
Mean: 2.93 80.61 68.48
Standard Deviation: 0.578 7.38 15.66
Coefficient of Variation: 0.198 0.09 0.23

Control
Mean: 2.83 80.42 71.58
Standard Deviation: 0.568 8.65 10.03
Coefficient of Variation: 0.201 0.11 0.14

 Exam 3 Exam 4 N

Experimental
Mean: 65.94 71.55 33
Standard Deviation: 13.24 12.21
Coefficient of Variation: 0.20 0.17

Control
Mean: 62.17 72.04 24
Standard Deviation: 12.50 11.97
Coefficient of Variation: 0.20 0.17

Table 2. Analysis of Variance: G.P.A. of Control and Experimental
Group. (n = 57)

Source DF SS MS F Probability

Factor 1 0.138 0.138 0.42 0.520
Error 55 18.130 0.330
Total 56 18.268

Table 3. Exam 1 Regression Model (n = 57)

Dependent Variable Constant Dummy G.P.A.

Exam 1 55.81 6.73 8.71
T-statistics (7.70) (0.70) (3.46)
p values 0.001 0.485 0.001 ***

Dependent Variable Interaction Adjusted [R.sup.2] F-Value

Exam 1 -2.53 24% 6.88
T-statistics (-0.77)
p values 0.443 0.001 ***

Dummy = 1 if experimental group, 0 = if control

G.P.A. = student's grade point average at the beginning of the
semester

Interaction = Dummy * G.P.A.

* Significant at 90% confidence level

** Significant at 95% confidence level

*** Significant at 99% confidence level

Table 4. Exam 2 Regression Model (n = 57)

Dependent Variable Constant Dummy G.P.A.

Exam 2 49.88 -22.96 7.68
T-statistics (3.95) (-1.38) (1.75)
p values 0.001 0.174 0.086 *

Dependent Variable Interaction Adjusted [R.sup.2] F-Value

Exam 2 6.53 22% 6.38
T-statistics (-1.14)
p values 0.258 0.001 ***

Dummy = 1 if experimental group, 0 = if control

G.P.A. = student's grade point average at the beginning of the semester
Interaction = Dummy * G.P.A.

* Significant at 90% confidence level

** Significant at 95% confidence level

*** Significant at 99% confidence level

Table 5. Exam 3 Regression Model (n = 57)

Dependent Variable Constant Dummy G.P.A.

Exam 3 54.63 -14.73 2.67
T-statistics (4.13) (-0.84) (0.58)
p values 0% 40% 0.564

 Adjusted
Dependent Variable Interaction [R.sup.2] F -Value

Exam 3 6.23 6.7% 2.34
T-statistics (1.04)
p values 0.302 0.084 *

Dummy = 1 if experimental group, 0 = if control

G.P.A. = student's grade point average at the beginning of the semester

Interaction = Dummy * G.P.A.

* Significant at 90% confidence level

** Significant at 95% confidence level

*** Significant at 99% confidence level

Table 6. Exam 4 Regression Model (n = 57)

Dependent Variable Constant Dummy G.P.A.

Exam 4 44.35 -5.81 9.80
T-statistics (3.94) (-0.39) (2.51)
p values 0.001 0.697 0.015 **

Dependent Variable Interaction Adjusted [R.sup.2] F-Value

Exam 4 1.48 21.5% 6.12
T-statistics (0.29)
p values 0.772 0.001***

Dummy = 1 if experimental group, 0 = if control

G.P.A. = student's grade point average at the beginning
of the semester

Interaction = Dummy * G.P.A.

* Significant at 90% confidence level

** Significant at 95% confidence level

*** Significant at 99% confidence level