Effects of Student Predisposing Characteristics on Student Success |
VOL. 5, No. 1, 5-19
Abstract
The question of why some students successfully study through distance education and others do not is becoming increasingly important as distance education moves from a marginal to an integral role in the provision of post-secondary education. This paper first advances a multivariate framework for examining this issue. It then explores the predictive capability of students' "predisposing characteristics" in regard to their chances of successfully completing their first Athabasca University distance education course. Using Discriminant Analysis nine predisposing characteristics were found to be significantly related to success. The paper concludes with an assessment of the applicability of this approach to provide the basis for a comprehensive model for understanding success and persistence in distance education.
Résumé
Le fait que certains étudiants réussissent à poursuivre leurs études grâce à l'enseignement à distance alors que d'autres n'y arrivent pas est une question qui prend de plus en plus d'importance étant donné l'évolution de l'enseignement à distance qui passe d'un rôle marginal à un rôle plus complet dans les besoins de l'éducation post-secondaire. Dans un premier temps, on se propose d'analyser le sujet en le présentant dans un cadre à variables multiples. On explore ensuite la capacité prédictive des "caractéristiques de prédisposition" des étudiants par rapport à leurs chances de passer avec succès leur premier cours d'enseignement à distance de l'Université Athabasca. D'après l'analyse discriminante, on a trouvé une relation significative entre neuf caractéristiques de prédisposition et la réussite au cours. L'article se termine avec une discussion sur l'application de cette approche susceptible de constituer la base d'un modèle complet pouvant servir à la compréhension de la réussite et de la persistance dans l'enseignement à distance.
Introduction
Openness and accessibility, the hallmarks of many distance teaching institutions, all too often seem to be associated with significantly lower rates of sucessful completion of courses and programs of study than campus-based institutions. Questions related to why some students succeed and others fail (however "success" and "failure" are defined) are of both theoretical and practical importance, as distance education moves from a marginal to an integral role in overall educational provision. Over the past decade there have been a number of studies examining student success and persistence in distance education. Some authors, e.g. Woodley and MacIntosh (1980) and Pythian and Clements (1982), found that students who withdraw most commonly cite job and domestic reasons for their decision, although Woodley and Parlett (1983) suggest self-delusion may bias such after-the-fact reporting. In contrast, Kennedy and Powell (1976) and Brindley (1987) expressed doubts about the explanatory value of changes in life circumstances as causes of course withdrawal except in combination with other factors related to withdrawal. Others (e.g., Schwittman, 1982) have concentrated on motivation as a critical predictor of success. Woodley and Parlett (1983) found that sociodemographic factors such as previous educational level, gender, age, and occupation are associated with persistence. Along the same lines, Rekkedal (1982) and Taylor (1986) reported that student success was associated with such factors as assignment turnaround time, the nature of student-tutor inter-action, and course quality.
Recently, attempts have been made to link together some of these various perspectives on student success and persistence behavior using multivariate approaches. Sweet (1986), using Tinto's (1975) conceptual framework adapted to distance education, reported that such factors as goal satisfaction, institutional commitment, and tutor contact contributed significantly to success. Siqueira and Lynch (1986), using a broader spectrum of variables, found that student satisfaction with the course, frequency of visits to student drop-in centres, socio-economic status, and perceptions of course materials were significant in explaining persistence. Chacon-Duque (1985) reported that persistence was affected by such factors as quality of course materials, variety of media, and planned student support, while education and age were not so related. Sung (1986), assessing program and environmental based student perceptions along with entry motivation and educational preparation, found that availability of time was the best predictor of retention, although adequacy of course materials and support services were also important predictors. In contrast to other studies (e.g., Schwittman, 1982), Sung reported motivation to be an insignificant predictor.
Two characteristics of the current status of research into student success and persistence in distance education are evident from reviewing the literature. First, the subject area is highly complex and multi- dimensional. Second, research seems to be going in many directions at once and the results reported are often seemingly contradictory. Even sophisticated multivariate studies have been hampered by the use of a limited range of measures and a lack of standardized measures, and the use of single items to measure broad concepts. This is not surprising, given the early stage of theory development and empirical research in distance education. However, if our understanding of student persistence is to develop further, it is no longer enough to take some of the factors leading to course withdrawal into account while ignoring others. Theorists (e.g., Coldeway, 1982, 1986; Calvert, 1986; and Garrison, 1987) stress the need for a comprehensive approach taking into account all the experiences of distance learners as well as the unique aspects of the distance learning environment.
The purpose of this present study is three-fold. First, a conceptual framework of student success and persistence in distance education will be advanced that will allow for a systematic empirical investigation of the phenomenon using multivariate statistical techniques. Second, the results of an investigation into the effects of one set of factors - predisposing characteristics - on student success will be presented. Finally, the implications of these findings will be discussed in the context of the proposed framework and in terms of what they suggest for further research into success and persistence.
An Empirical Model of Student Success and Persistence
On the basis of previous studies, one can classify the factors contributing to success and retention in distance education into three general categories. The first set consists of those characteristics students bring to the educational process at the time of entry, such as educational preparation, socioeconomic and demographic status, and motivational and perseverance attributes. These predisposing characteristics are either fixed or slowly changing throughout the duration of a student's involvement with a distance education institution and, as such, exert a relatively constant influence on students' chances of success. The second category consists of changes in life circumstances that disrupt or in some way alter the goals, expectations, and commitment with which students begin their distance education studies. Such life changes as personal illness, relocation, altered employment status, and family problems occur quickly and often unexpectedly. The third category contains factors that can be termed institutional, that is, under the control of the educational provider. These include quality and difficulty of instructional materials, access to and quality of tutorial support, and the administrative and other support service provided. The relationship between these three sets of factors is portrayed in Figure 1.
These three influences differ in their explanatory power and in the types of influence they exert on success and retention. Predisposing characteristics are antecedent predictors: that is, they are present before and during student involvement in distance education. The framework suggests that life changes and institutional factors do not, in most cases, act as direct causes of student dropout. While they do influence the probability of student success and persistence, they appear to do so primarily in interaction with predisposing characteristics. According to the model, they would have little independent explanatory power, since they affect the student body differentially based on predisposing characteristics. For example, the level of literacy (a predisposing characteristic) among students would likely be a more important predictor of success for distance teaching organizations that rely on print-based instruction than for organizations using other modes of instruction.
Ideally, a comprehensive explanation of success and persistence in distance education would involve the systematic development and testing of valid and reliable measures within the context of such a framework. This is a formidable task, both in terms of data collection and in statistical modelling. Fortunately, the analytical framework above does not require the concurrent investigation of all three sets of factors for systematic progress toward a comprehensive explanation of student behavior. The framework places predisposing characteristics first, both in terms of time and explanatory value: it is only when these characteristics are known and controlled for that the effects of the other two factors can be determined. It is for this reason that the emphasis of this phase of the research is on the effects of predisposing characteristics on student success.
Methodology
The Setting
Athabasca University (AU) teaches about 11,000 per year who account for approximately 16,000 course registrations. AU has an open admissions policy requiring only that students be at least 18 years of age and be residents of Canada. Students can enrol and register in courses throughout the year and set their own pace of study in a course. Most students study in a "home study" mode - that is they receive a print-based correspondence package and can contact a telephone tutor weekly.
Sample Size and Response Rate
Three hundred and one newly-enrolled students living in the Edmonton area were interviewed face- to-face before they started work on their first AU course. The cooperation rate was 94% and the overall response rate was 72%. Analysis of demographic data (e.g., gender, age, and general educational background) and completion status of the final sample with Edmonton area students in general showed the sample to be representative.
Defining Success
The framework displayed in Figure 1 refers both to success and the related concept of persistence - which at AU would mean students completing their intended programs of studies. Unfortunately, we are not yet able to report on the relationship of predisposing characteristics on persistence because of the long period of time many students take to register for their second and subsequent courses. (A time lapse of 18 months or more is not uncommon.) For the purposes of this study we defined success as whether newly enrolled students passed their first course at Athabasca University (AU). This is admittedly a very restricted definition of success in distance education. Certainly, one could argue that as long as students get what they want out of their course they "succeed" even if they do not submit a single assignment. However, a restricted definition was chosen for two main reasons. The first reason was related to the constraints imposed by student behavior at AU. About 40% of students successfully complete their first course. The mean pass mark is 81%. In addition, few students (3%) actually fail their first course. Most of those who do not pass withdraw without completing any assignments. This predominant bimodal distribution of completion behavior (i.e., good pass versus withdrawal) meant that defining success in a broader context (e.g., including proportion of work completed in a single course, final grades received, and progress through a program of studies, and so on), while theoretically possible, would have required a very large sample for meaningful statistical analysis.
Data Reduction and Development of Scales and Indices
The instrument yielded dichotomous, categorical (mostly Likert scale measures) and continuous data. Due to the large number of items measured (64) in the instrument, the potential number of variables to be analyzed in a multivariate analysis were reduced. First, variables with a large proportion of missing data, in most cases contingency questions applying to sub-groups of the sample, were excluded. Secondly, single measure and composite items with highly skewed distributions were removed. Finally, the remaining variables were, in most cases, grouped to increase the "reliability" of the measures to be used in the multivariate analysis and to reduce the possibility of highly correlated variables being entered into the predictive equation. Items were grouped if conceptually they seemed to measure similar phenomena, and were found to be mathematically correlated. In addition, some grouped indices were developed using factor analysis. Single item measures were converted to standard scores to ensure that their relative contributions to multiple measures were equivalent before being grouped either additively or multiplicatively. Skewness co-efficients were then used to determine whether the distribution of the resulting scales and indices were normal. Although normality of individual measures does not guarantee multi-variate normality, it does provide some measure of confidence that the remaining multivariate distribution will be normal (Tasuoka, 1976).
The multivariate technique selected for the data analysis was Discriminant Analysis. Unlike other acceptable techniques (e.g., multiple regression), this analyzes the interactions among a number of "predictor" variables to arrive at a single, composite score that allows one to predict outcomes on a case by case basis. Finally, the information provided by the Discriminant functions can be used to classify future samples of individuals. Discriminant Analysis also permits both continuous and dichotomous variables to be used as discriminating variables (Gilbert, 1968).
Defining the Discriminating Variables
A range of variables were chosen to measure factors predisposing students toward success or withdrawal/failure in their first AU course. These variables, their composition, and the scale of measurement are shown in . Although this set of discriminating variables was developed in the light of existing literature on student success in distance education, not all factors were addressed because of difficulties in measurement in what was already a lengthy instrument (e.g., learning style preferences) and what was thought to be limited applicability to AU (e.g., degrees of institutional commitment among first time students, most of whom would take one or two courses before leaving).
Results
A stepwise discriminant analysis was performed to assess prediction of membership in two groups (pass group and fail/withdrawal group). A stepwise analysis, as opposed to direct simultaneous entry of all discriminating variables, was used because of the large number of variables (15) selected for analysis. Simple statistics, including means and standard deviations, for variables included in the stepwise discriminant analysis are provided in . Since no . priori prediction concerning the relative importance of the 15 variables to be included in the analysis was made, the stepwise procedure was used not only to reduce the number of discriminating variables, but also to provide the best linear combination of variables that were included in the analysis. Only those variables that made a significant contribution to the discriminant function were permitted entry, based on a simple F-to-enter criteria. Variables with the highest F-ratios were entered first.
Of the original 301 cases in the study, 58 were excluded from the analysis because of at least one missing discriminating variable. The remaining 243 students were distributed such that 153 were members of the fail/withdrawal group and 90 were members of the pass group.
Stepwise variable selection of variables to enter into the equation was made on the basis of minimizing the overall Wilke's Lambda (U statistic). While other stepwise procedures are available, Tabachnick and Fidell (1983) recommend that in the absence of contrary reasons, Wilke's Lambda is the procedure of choice. Eight variables were included in the final discriminant function (Chi square=62.13; 8 d.f.; p<.001). The standardized canonical discriminant function coefficients, shown in , suggest that the primary variables responsible for discriminating between students who pass and those students who fail/withdraw were persistence, marital status, need for success, need for support. Other variables that contributed to the significance of the function were students' literacy score, financial stability, study habits, gender and the students' rating of previous educational preparation.
Overall, based on the discriminant function, 68.7% of the students were classified correctly. provides a summary of the classification results. Variables that made no significant contribution to the discriminant function included current educational level, educational commitment, level of support, attitudes towards studying, number of children, and respondent's age.
Discussion
Profile of Successful AU Students
The variables which were included in the discriminant model can be used to construct a reasonably detailed profile of potentially successful and "at risk" AU students. There appear to be nine major criteria differentiating between successful and unsuccessful AU students. Students who rated themselves highly on various measures of persistence related to taking on new projects were more likely to succeed in their AU studies. Married students (including those who had a common law relationship) were more likely to succeed than single students. This could be a measure of a more general underlying variable of the existence of support structures.
Students who rated the consequences of not passing as serious were more likely to pass their first course. Successful completers tended to rate their chances of succeeding in their studies higher than those who eventually withdrew from or failed their first course. Respondents who indicated that they needed support from others to complete difficult tasks and who said they found it important to discuss course work with other students were numbered among the unsuccessful group. This is not surprising, given that AU's home study mode of delivery places a high value on independence, a value which is positive for independent learners but may not serve more dependent learners. Year-round admissions, the absence of pacing, and the onus on students to initiate contact with tutors and other support services all emphasize the independence of students. Student literacy as measured by a Cloze test was also related to student success. Again this is not surprising given AU's open admissions policy and the fact that most learning has to be done using print-based materials. Two measures were used to create this variable: household income and perceived financial security. Respondents receiving the highest score on this combined measure were more likely to succeed than those with lower scores. It is interesting to note that household income alone was not associated with successful completion in bivariate analysis. Students who said they had a designated place for study, regular times for study and generously estimated the study time needed to successfully complete their course were more likely to pass. In addition, respondents who rated themselves as well organized in terms of time management skills and said they generally had the time to do what they intended to do were also likely to succeed. Similarly, students who rated the value of their formal and informal (that is, out of school) learning as high in terms of preparing them for university studies at AU tended to succeed. Interestingly, the level of previous educational experience, although measured in the study, did not enter the model as a significant predictive factor, while students' subjective ratings of their educational experience did. This suggests that formal educational qualifications may not be as accurate a measure of preparedness for distance education study as many would argue. (This is not to suggest that no correlation existed between previous educational background and success - it did - but that other measures, associated with previous educational background, were better predictors.) Female students were more likely to succeed than male students. However, in a previous analysis when we included the mean completion rates of the courses chosen by students in the model, gender ceased to be a significant predictor variable. This suggests that gender differences in completion rates are, at least partly, explainable by course choice.
Generalizability of the Results
The set of factors that predict student success and persistence among AU students would not necessarily apply to other populations of distance learning students and other institutions. Indeed, the analytical framework proposed in Figure 1 would lead one to expect the relative predictive power of predisposing characteristics and the characteristics themselves to vary according to institutional factors and even life changes (insofar as student populations with different sociodemographic make-ups may be differentially subject to such disruptive changes in life circumstances as, for example, divorce and unemployment). As stated previously, AU's system of the provision of tutorial and other forms of support places a premium on independence in learning as students do not meet in classes or tutorial sessions, progress through courses is unpaced and students are usually expected to take the initiative before receiving available forms of institutional support. Not surprisingly, such qualities as independence, organizational abilities, and the existence of outside support structures figure predominantly in predicting success. However, changing institutional factors to include paced instruction, opportunities for group learning, and "laid on" support provision, may well de-emphasize the importance of such predisposing characteristics and bring others to the fore. On the other hand, it is also reasonable to suppose there are certain aspects intrinsic to distance teaching and the adult populations who learn through such methods that may yield a set of "generic" predisposing characteristics in distance education applying across institutions and student populations. The answer to this question is empirical and can be tested through systematic comparative study.
Implications for Further Research
The discriminant model classified 69% of the respondents correctly in terms of their "success" on their first AU course. This compares with an expected 50% average that would have been achieved by random allocation.
Our analysis shows that a substantial amount of the variance of completion behavior is explained by predisposing characteristics. However, the explanatory value, though significant, does not account for all of the variance in completion behavior, thus indicating that predisposition is not, in effect, predestination. Therefore, the interaction of institutional factors and changes in life circumstances, with student predisposing characteristics should provide a clearer picture of student success, as proposed in the model shown in Figure 1.
The model becomes most useful if students, on entry, can be "risk stratified" - that is, if students can be determined as "at risk" of withdrawal/failure or predisposed toward success. Although the model as a whole has a successful classification rate of 69%, it seems to be more powerful in predicting withdrawal/failure: the model successfully classified 90% of the cases in the lowest quintile of discriminant scores. Although this finding is not conclusive, it opens up the potential for using the model to identify "at risk" students, particularly as the discriminant score assigned to each subject can be roughly interpreted as a "risk quotient."
The encouraging predictive ability of the preliminary predisposing characteristics model opens the way for research at two levels. At the first level, knowledge of students' predisposing characteristics (and by extension their "risk quotient") allows a distance teaching organization to target interventions to those most in need. This should result in more efficient interventions and will facilitate better measurement of the effectiveness of interventions in two senses. First, this approach will exclude those who neither need nor want additional or special service, thus reducing the need for large sample sizes to determine effectiveness. Second, the use of data generated by the model allows for quasi- experimental designs to control for the effects of important extraneous factors, thus permitting a more accurate assessment of the relationship between interventions and outcomes (i.e., improvements in success and retention rates).
At the first level, our research to date affords us the opportunity to assess the effectiveness of individual interventions predicated on predisposing characteristics. However, such research does not accommodate interactions among predisposing characteristics, institutional factors, and life changes as portrayed in Figure 1. On the other hand, the results do provide empirical support for pursuing a second level of research - a systematic program of research into success and persistence within the above framework. At this second level, the framework allows for the analysis of interactions among predisposing characteristics, institutional factors, and life changes, to allow the university to adapt and integrate its program and service infrastructure to reflect the heterogeneous and changing needs of adults learning at a distance.
References
Brindley, J. E. (1987). Attrition and completion in distance education: The student's perspective. Unpublished M.A. thesis, University of British Columbia, B.C.
Calvert, J. (1986). Research in Canadian distance education. In I. Mugridge & D. Kaufman (Eds.), Distance education in Canada (pp. 94–110). London: Croom Helm.
Chacon-Duque, F. J. (1985). Building academic quality in distance higher education. . Monograph in higher education evaluation and policy. Pennsylvania, P.A.: Pennsylvania State University.
Coldeway, D. O. (1982). Learner characteristics and success. In I. Mugridge & D. Kaufman (Eds.), Distance education in Canada (pp. 81–93). London: Croom Helm.
Coldeway, D. O. (1986). Recent research in distance learning. In J. S. Daniel, M. A. Stroud, & J. R. Thompson, Learning at a distance: A world perspective (pp. 29–37). Edmonton: Athabasca University.
Garrison, D. R. (1987). Researching dropout in distance education. Distance Education, 8(1), 95–101.
Gilbert, E. S. (1968). On discrimination using qualitative variables. Journal of the American Statistical Association, 63, 1399–1412.
Kennedy, D., & Powell, R. (1976, November). Student progress and withdrawal in the Open University. Teaching at a distance, 7, 61–75.
Pythian, T., & Clements, M. (1982, Summer). Drop-out from third level math courses. Teaching at a Distance, 21, 33–45.
Rekkedal, T. (1982). The dropout problem and what to do bout it. In J. S. Daniel, M. A. Stroud, & J. R. Thomson (Eds.), Learning at a distance: A world perspective (pp. 118–122). Edmonton: Athabasca University.
Schwittman, D. (1982). Time and learning in distance study. Distance Education, 3(1), 141–156.
Siqueira de Freitas, K., & Lynch, P. (1986). Factors affecting student success at the National Open University of Venezuela. Distance Education, 7(2), 191–200.
Sung, N. (1986). Perceptions of program and learning environment as determinants of persistence and poststudy attitudes in adult distance learning. Unpublished D.Ed. dissertation, Pennsylvania State University, P.A.
Sweet, R. (1986). Student dropout in distance education: An application of Tinto's model. Distance Education, 7(1), 68–91.
Tabachnick, B. G., & Fidell, L. S. (1983). Using multivariate statistics. New York: Harper & Row Publications.
Tatsuoka, M. M. (1976). Discriminant analysis. Research issues 13: Data analysis strategies and design for substance abuse research. Washington, D.C.: U.S. Government Printing Office.
Taylor, J. C., et al. (1986). Student persistence in distance education: A cross-cultural, multi- institutional perspective. Distance Education, 7(1), 68–91. Tinto, V. (1975). Dropout from higher education: A theoretical synthesis of recent research. Review of Education Research, 45, 89–125.
Woodley, A., & McIntosh, N. (1980). The door stood open: An evaluation of the open university younger student pilot scheme. Falmer Press.
Woodley, A., & Parlett, M. (1983). Student drop-out. Teaching at a Distance, 24, 2–23.
Richard Powell is Head of Institutional Studies in the Centre for Distance Education at Athabasca University. Before joining Athabasca University he was a researcher in distance education at the Open University (U.K.) and a research manager with the Lesotho Distance Teaching Centre in Lesotho, southern Africa.
Christopher Conway is a Senior Research Analyst within the Centre for Distance Education at Athabasca University. Before coming to Athabasca University, he taught statistics and research methods at Ryerson Polytechnical Institute.
Lynda Ross is a Research Analyst at Athabasca University. Before coming to Athabasca University, she was a technical analyst at the Gerontology Research Centre at the University of Guelph.
PID: http://hdl.handle.net/10515/sy5wd3qf2
![Creative Commons License](http://i.creativecommons.org/l/by/3.0/80x15.png)
This work is licensed under a Creative Commons Attribution 3.0 License.
![]() |