International Review of Research in Open and Distributed Learning A Predictive Study of Student Satisfaction in Online Education Programs

This paper is intended to investigate the degree to which interaction and other predictors contribute to student satisfaction in online learning settings


are unable to regul
te learning efficiently are unlikely to be satisfied (Artino, 2007;Puzziferro, 2008) This study investigated factors (i.e., Interaction, Internet self-efficacy, self-regulation) associated with student satisfaction in fully online learning settings.


Literature Review Student Satisfaction

Evaluation is important in distance education and it consists of different dimensions in alignment with the goals of a course or program (Olmstead, 2007).Course grades are often used as an indicator of student achievement in online instruction (Barnard, Paton, & Lan, 2008;Reinhart & Schneider, 2001;Noel-Levitz, 2011).But affective factors can be as important as cognitive factors in explaining and predicting student learning in online settings (Biner et al., 1997).Among the attitudinal constructs, student satisfaction, referring to student perceptions of learning experiences and perceived value of a course, may be particularly worthy of investigation.Student satisfaction is related to several outcome variables such as persist

ce (Allen & S
aman, 2008), retention (Debourgh, 1999;Koseke, & Koseke, 1991), course quality (Moore & Kearsley, 1996), and student success (Keller, 1983;Pike, 1993;Noel-Levitz, 2011).High Vol 14 | No 1

March/13 18 satisfaction leads to lower attrition rates, higher persistence in learning, and higher motivation in pursuing additional online courses (Allen & Seaman, 2008;Biner, Welsh, Barone, Summers, & Dean, 1997;Keller, 1987;Koseke, & Koseke, 1991).Higher education institutions consider student satisfaction as one of the major elements in determining the quality of online programs in today's markets (Yukselturk & Yildirim, 2008).Online learner perspectives provide valuable information on the areas that matter to students and help institutions gain a better understanding of their strengths a d challenges in provision of online programs (Noel-Levitz, 2011).With data on student satisfaction, course designers, educators, and administrators can identify areas where improvement is needed (Reinhart & Schneider, 2001).


Predictors of Student Satisfaction

Previous studies have determined factors that influence student satisfaction in distance learning environments (Artino, 2007;Bolliger & Martindale, 2004;Reinhart & Schneider, 2001;Sahin, 2007).The framework of this study was proposed based on the interaction model developed by Moore (1989) with the addition of potential variables including Internet self-efficacy and self-regulated learning.


Interaction.

Interaction has been deemed one of the most important components in distance education due to the isolation of instructors and learners (Moore & Kearsley, 1996).Moore (1989) proposed an interaction framework including learner-learner interaction, learner-instructor interaction, and learner-content interaction.

Learner-learner interaction refers to two-way reciprocal communication between or among learners who exchange information, knowled

regarding course con
ent, with or without the presence of an instructor (Moore & Kearsley, 1996).

Learner-instructor interaction consists of two-way communication between the instructor of a course and learners.Learner-content interaction is a process of individual learners elaborating and reflecting on the subject matter or the course content.In contrast with learner-instructor and learner-learner interaction, only one person -the learner -is directly involved in learner-content interaction (Moore & Kearsley, 1996).

Other authors have used Moore's interaction model as a basis for new interaction models.Hillman, Willis, and Gunawardena (1994) proposed learner-interface interaction to recognize the importance of technology interfaces in interaction with content, instructors, or learners.Anderson and Garrison (1995) expanded Moore's interaction model by adding three new types of interaction: instructor-instructor, instructor-content, and content-cont nt interaction.

Previous research has indicated the positive influence of interaction on student satisfaction in distance education (Bray et al., 2008;Burnett, 2001 Northrup, Lee & Burgess, 2002;Thurmond & Wambach, 2004).Of the three types of interaction, learner-learner interaction and learner-instructor interaction were investigated more often than learner-content interaction.Learner-learner interaction and learner-instructor interaction seem to be more related to and predictive of student satisfaction than learner-content interaction in most studies of online learning (Bolliger & Martindale, 2004;Jung, Choi, Lim & Leem, 2002;Rodriguez Robles, 2006;Thurmond, 2003).Battalio (2007) concluded that learner-instructor interaction was the most required interaction in his summary from several online studies.However, the f

dings are inconclusive.Some studies
indicated that the amount of interaction that learners have with the content is most important to student satisfaction in web-based learning, in comparison with learner-learner interaction and learner-instructor interaction (Chejlyk, 2006;Keeler, 2006).

Internet self-efficacy.

Expanded from the self-efficacy theory in psychology (Bandura, 1977), researchers in education have indicated that

ficacy belief
positively influence achievement and persistence related to specific instructional tasks (Hodges, 2008;Pintrich & De Groot, 1990).Internet self-efficacy refers to the belief in one's capability to organize and execute Internet-related actions required to accomplish assigned tasks (Eastin & LaRose, 2000).There a e two reasons to include Internet self-efficacy as a predictor of online student satisfaction.First, online learning relies on Internet delivery through which various types of activities take place such as group discussions, collaborative projects, co munication with instructor or classmates, and so on (Roach & Lemasters, 2006).Technical problems while using the Internet may cause student frustration and dissatisfaction (Choy, McNickle, & Clayton, 2002).It seems important for online learners to possess high Internet self-efficacy to complete required tasks for an online course delivered through the Internet.

Secondly, Internet self-efficacy, as one of the three self-eff cacy constructs in web-based instruction, is less addressed than academic self-efficacy or computer self-efficacy.The impact of Internet self-efficacy on student satisfaction is scarce and inconclusive.For example, Eastin and LaRose (2000) indicated that Internet self-efficacy is positively correlated with expected outcomes including entertainment, social, and informational outcomes.Chu and Chu (2010)  Self-regulated learning.

Self-regulation, originally from psychology, was first defined by Bandura (1988).The central ideas underlying self-regulation are motivation and learning strategies that students utilize to achieve their learning goals.The scope of self-regulation has been expanded to studies in education areas (Pintrich, Smith, Garcia, & McKeachie, 1993).

Self-regulated learning refers to the degree to which students metacognitively, motivationally, and behaviorally participate in their own learning (Zimmerman, 1989).

Metacognitive processes involve learners' ability to plan, schedule, and evaluate their learning progress.Motivational processes indicate that learners are self-motivated and willing to take responsibility for their successes or failures.Behavior refers to the characteristics of the strategies that students utilize to opt

ize learning (Zimmerman
Martinez-Pons, 1986, 1988).

The importance of self-regulation in student performance is evident in traditional faceto-face learning settings (Pintrich & De Groot, 1990;Zimmerman & Schunk, 1989).

Unlike traditional classroom instruction, online learning is student-centered and much self-directed effort is required for success (Artino, 2007).Although most of the studies have indicated that the ability to self-monitor and self-evaluate at different learning stages is positively related to student performance or achievement, there is very limited research pertaining to the association between self-regulation and student satisfaction.

For example, Artino (2007) found motivational components of self-regulation are positively related to student satisfaction from a sample of participants attending military online training in the U.S. Navy.Puzziferro (2008) studied students who took online courses from a liberal arts community college and indicated that meta-cognitive self-regulation is positively correlated with student satisfaction at a significant leve .

This study focuses on metacognitive self-regulation because metacognitive processes are considered to be the most critical in self-regulation (Brockett & Hiemstra, 1991;Corno, 1986;Corno & Mandinach, 1983;Lee, Kim, & Grabowski, 2010).


Purpose and Research Questions

Based upon the literature review, interaction is critical in distance education; however, the literature is inconclusive regarding which type of interaction is most important in predicting student satisfaction in online settings (Bray, Aoki, & Dlugosh, 2008;Chejlyk, 2006;Keeler, 2006;Rodriguez Robles, 2006).In light of the characteristics of online learning, we assumed that Internet self-efficacy and self-regulation are important in online learning and included them as predictors.Moreover, the scarcity of research leads to ambiguity pertaining to the effect of self-efficacy and self-regulation on student satisfaction (Artino, 2007;Chu & Chu, 20 0;Puzziferro, 2008;Rodriguez Robles, 2006).Hence, the purpose of this study was to investigate the extent to which the potential variables -interaction, Internet self-efficacy, and self-regulation -predict are needed to make confident assumptions about any observed relationships (Stevens, 2002).

There were more female respondents (77%) than ales (23%).Most respondents were married (64.9 %).The Internet self-efficacy scale was developed by Eastin and LaRose (2000) to measure one's confidence in the abilit to be successful in performing certain tasks using

Internet-based technology.This measurement is a 7-point Likert scale with 8 items that ranged from 1 (very unlikely) to 7 (very likely).This scale was found to be reliable and internally consistent with a Cronbach's coefficient alpha value at .93, based on a population of 171 undergraduate students at a university.

The self- egulated learning scale was adopted from the metacognitive self-regulation subscale in the Motivated Strategies for Learning Questionnaire (MSLQ) developed by Pintrich et al. (1993).The scale is a 7-point Likert scale with 12 items ranging from 1

(not at all true of me) to 7 (very true of me).It assesses the extent to which learners used planning, monitoring, and regulating strategies during the learning process.It is reliable and valid with a coefficient alpha at .79.

The measure of interaction and satisfaction was modified from an existing instrument developed by the authors (2009) in a blended learni g environment.It is a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).Three stages were involved in the instrument development process.

• Stage 1: Slight modifications such as wording changes were made to assure the suitability of items given the context of this study was within a fully online learning setting.

• Stage 2: To ensure the content validity of the instrument, a content validity survey was conducted.Six experts, including university professors and professionals with either research expertise or teaching experiences in online learning, were invited to review the questions.Reviewers were asked to rate each item and

etermine whether the item was a
equate for these specific domains on a basis of three choices: essential, useful but not essential, and neither essential nor useful.Content validity ratio (CVR) was calculated based on the ratings from these six experts.The threshold of CVR value to maintain an item for a case of six reviewers is 0.99 (Cohen & Swerdlik, 2004).Considering the small number of reviewers, we decided to combine "essential" and "useful but not essential" into one option for CVR calculation.Items measuring similar concepts or with a CVR value lower than 0.99 were either removed or combined with other items.Wording changes were made based on the suggestions of reviewers.

Vol 14 | No 1 March/13 24

• Stage 3: The slightly revised items and combined items were sent back to the reviewers for a second-round rating to ensure they were adequate and necessary.

After item elimination and revision, there were eight items in the learner-learner interaction subscale, six items in the learner-instructor interaction subscale, and four items in the learner-content interaction subscale (Table 2).The Cronbach's coefficient alpha values, calculated based on the sample of this tudy, indicated the developed instruments are reliable (Table 2).


Findings

Descriptive Analyses of Variables Table 3 depicts t e average scores on the subscales.It seems that the amount of interaction among students is lower than that of interaction between earners and instructor or content.Most of the students had a slightly high level of Internet selfefficacy given that the average score was larger than the mid-point score of 4. The average score of self-regulation was n

tral, implying
that students may be able to apply some of the self-regulation skills properly in their learning, but not in a very refined way.Overall, students were moderately satisfied with their learning experiences in an online course.


Correlation Analyses

The Pearson correlation coefficients among the scales are presented in Table 4.The three types of interactions were all positively related to satisfaction.Students' responses on learner-content interaction (r = .664,p < .01)and learner-instructor interaction (r = .542,p < .01)are relatively highly correlated with student satisfaction in comparison with learner-learner interaction (r = .246,p < .05).It seems that when the interactions of students with their fellow students, instructors, or content increased, the level of Vol 14 | No 1 March/13 26 satisfaction was enhanced.Internet self-efficacy was also positively related to satisfaction (r = .437,p < .01).Students who had higher self-eff cacy in performing

Internet actions tended to be more satisfied with the course.There was no signif

ant relationship
between self-regulated learning and satisfaction (r = -.004,p > .05).As shown in Table 5, there was no multicollinearity for the predictors with tolerances larger than .10 and VIFs smaller than 10.Learner-instructor interaction (t(102) = 3.28, p < .05),learner-content interaction (t(102) = 6.15, p < .001),and Internet self-efficacy (t(102) = 3.82, p < .001)were significant predictors in explaining student atisfaction.

Learner-learner interaction (t(102) = -0.26,p > .05)and self-regulated learning (t(102) = -1.12,p > .05)did not significantly contribute to the prediction f student satisfaction.

The squared semipartial correlation (Table 5) informs the uniqueness of a predictor, which is the amount of variance that cannot be explained by other variables entered in the equation.Learner-content interaction explained the largest amount of unique variance (16.1 %) in satisfaction compared to other the four predictors.Internet selfefficacy and learner-instructor interaction followed with 6.2 % and 4.5 % of the unique variance in student satisfaction respectively.

Vol 14 | No 1 March/13 28


ANOVA Analyses

One-way ANOVA analyses were performe to explore the effects of student background variables on the three types of interaction, Internet self-efficacy, and self-regulated learning.Levene's Test of Equality was performed in advance to ensure the assumption of equal v riance was met.Age and marital status were found to have no significant influence on any of the predictor variables.Post hoc tests were used to compare all groups of participants with each other.

As indicated in Tables 6 and 7, gender and class level had a significant effect on learnerlearner interactio .Female students (M = 2.99, SD = 1.07) had significantly more learner-learner interaction than male students (M = 2.44, SD = 1.28),F(1, 106) = 4.48, p < 0.05, η2 = .04.Students n undergraduate-level courses (M = 3.47, SD = 1.10) were found to have significantly less interaction with their classmates, as opposed to those in graduate-level courses (M = 2.65, SD = 1.09),F(2, 105) = 4.93, p < 0.01, η2 = .08.


One-Way ANOVA of Class Level on Predictor Variables

The number of hours spent online per week was significantly associated with Internet self-efficacy and self-regulated learning; the magnitude of this effect was medium (see Table 8).Students who spent less than 5 hours online each week (M = 5.78, SD = 1.11) had a significantly higher level of Internet self-efficacy than those who spent more than 20 hours online (M = 4.27, SD = 1.73),F(4, 103) = 2.48, p < 0.05, η2 = .08.Students who spent 11-15 hours online per week for coursework (M = 4.36, SD = 0.67) were significantly more self-regulated as opposed to those who spent less than 5 hours (M = 3.70, SD = 0.93), F(4, 103) = 2.72, p < 0.05, η2 = .09.


Undergraduate


M SD

Under/graduate
M SD Graduate M SD F(2,

Discussion

The purpose of this study was to analyze the effect of important predictor variables on student satisfaction in online settings.The model of this study was b ilt upon Moore's interaction model with the addition of two variables, Internet self-efficacy and selfregulated learning, which were assumed to be important in this study.The results confirmed the importance of interaction (Bray et al., 2008;Chejlyk, 2006;Keeler, 2006;Rodriguez Robles, 2006).All three types of interaction were significantly correlated with student satisfaction.Among the three types of interaction, learner-instructor interaction and learner-content interaction significantly contributed to student satisfaction while learner-learner interaction was a poor predictor of student satisfaction.

Learner-content interaction was the strongest predictor of student satisfaction, which confirms the findings of Chejlyk (2006) and Keeler (2006).Learner-instructor interaction followed as the second strongest predictor that signi

cantly contributed to
student satisfaction.This result suggested that the design of online content may be the most important contributor to student satisfaction.Online learners may spend most of Vol 14 | No 1 March/13 31 their time reading and digesting content through thinking, elaboration, and reflection.

Organization of content, document layout, and ease of accessing online content may influence learners' interaction with course content.Inclusion of media tools or interactive videos (Anderson, 2003;Havice, Davis, Foxx, & Havice, 2010) may stimulate learners' motivation to learn and in turn increase student interaction with course content.

To increase learner-content interaction, it is necessary to understand the attributes of technolo ies that support interaction and instructional design that fits specific learning contexts (Anderson, 2003;Wagner, 1994).Compared to other settings (e.g., synchronous, hybrid), learner-content interaction has greater influence on learning outcomes in asynchronous settings (Bernard, Abrami, Borokhovski, Wade, Tamim, Surkes, & Bethel, 2009).Inclusion of tasks that involve collaboration and searching online resources may help enhance learners' interaction with content.For instance, problem-based learning would encourage online earners to apply their information search skills to resolve authentic problems, which in turn increases learners' interaction with the content as well as their problem solving skills ( n & Reigeluth, 2008).

Course design wo

d affect learne
s' interaction with the content and their instructor (Moore & Kearsley, 2005).Course design with low flexibility leads to reduction of learner-instructor interaction (Eom & Wen, 2006;Giossos, Koutsouba, Lionarakis, & Skavantzos, 2009).The more rigid a course, the less autonomy a learner has.A highlystructured course design may be preferred by some online learners as it provides basic course information (e.g., course objectives, teaching strategies, evaluation me hods) as well as specified guidelines along with each task or assignment (Lee & Rha, 2009).
Learner-learner interaction was not a significant predictor for student satisfaction, which is contrary to the studies from Jung et al. (2002) andRodriguez Robles (2006) where learner-learner interaction was found to be the strongest predictor in web-based learning environments.This finding appears to be reasonable since this study was conducted in

lly online
nvironments where students may not have many opportunities to interact with their classmates.In the study of Jung et al. (2002), collaborative activities were designed as part of the online courses, resulting in a significant influence of learner-learner interaction on student satisfaction.The amount of collaborative learning design in online settings seems to be an important factor that leads to the effect of learner-learner interaction on student satisfaction.In terms of social constructivist perspectives, collaborative projects and group assignments enable the processes of conversation, discussion, and negotiation among learners (Woo & Reeves, 2007).The online courses collected in this study were eight-weeks long and were offered in summer.This type of acceler ted online course may decrease instructors' willingness to involve a great amount of group activities given the time constraints.

The significant influence of Internet self-efficacy on student satisfaction was supported in this study, which is contrary to previous research where Internet self-efficacy was not Vol 14 | No 1 March/13 32 found to be a critical factor of student satisfaction (Rodriguez Robles, 2006).Given the results, it may be helpful for institutions to provide appropriate training regarding Internet skills to improve stu ents' Internet self-efficacy before online courses are implemented.Research has shown that students with a higher level of Internet selfefficacy have better information searching skills (Thompson, Meriac, & Cope, 2002) that may increase their confidence in utilizing the Internet as well as trouble shooting during learning.

Contrary to the tudy of Puzziferro (2006), self-regulation was neither correlated with nor predictive of student satisfaction.The capability of managing learning pace did not seem to be a critical variable to student satisfaction.However, there were not many studies investigating the effect of self-regulation on student satisfaction in web-based settings.More research may be needed to verify the effect of self-regulation on student satisfaction.

This study also examined the effect of student background variables on three types of interaction, Internet self-efficacy, and self-regulation.Female students were found to engage in more learner-learner interaction than male students.Students in graduatelevel courses had more interaction with their classmates than those in undergraduate courses.Students who registered for advanced courses such as tho e offered in graduate school may come from similar professional backgrounds and have similar interest in the content area.It may be easier for them to share and exchange ideas with their classmates or brainstorm on content-related topics, which in turn helps the acquisition of content knowledge.

The finding of the association between time spent online and Internet self-efficacy was interesting.It seemed to make sense that students who spent less than 5 hours online had higher Internet self-efficacy than those with more than 20 hours.Students who were more confident in using the Internet f r their coursework might have spent less time online; on the contrary, those who were not familiar with the Internet at all might need to spend more time going through the tasks required for the course.Limitations and Suggestions for Future Research

Given the small sample size of this study, future research is recommended to verify and generalize the findings among diverse students.Employing the measurement scales in future studies in alternative contexts may provide additional evidence of the validity of these instruments.In addition, other variables (e.g., class size, course structure, learning style, and so on) that may affect student satisfaction should be included and examined in web-based settings to expand the understanding of online learning (Lee & Rha, 2009;Zacharis, 2011).

The design of the courses in which students were enrolled was not assessed.This makes it possible that the nature of the course designs led to our finding that learner-learner interaction was not a significant predictor of student satisfaction.Future research should assess the design of online courses and use this as a moderating factor in the prediction of student satisfaction.Such research could shed further light on whether learner-learner interaction is a consistent predictor of student satisfaction.


Conclusions

This study indicated that the interaction framework with the inclusion of two predictors (i.e., Internet self-efficacy, self-regulation) was valid.Learner-instructor interaction, learner-content interaction, and Internet self-efficacy were significant predictors of student satisfaction in fully online learning settings, while learner-learner interaction and self-regulated learning did not predict student satisfaction.Lea ner-content interaction was the strongest predictor among those significant predictors of student satisfaction.The importance of interaction in online learning was confirmed.

Additionally, gender and class level significantly influenced learner-learner interaction.

The effect of time spent per week online was substantially influential for Internet selfefficacy and self-regulation.

The practical implications of this study are that both instructors and course designers should pay attention to content design and organization given that learner-content interaction substantially contributes to student satisfaction.



online learning.The influence of student background variables on predictor variables was explored.In a dition, this study examined the unique contribution of key predictor variables in explaining the variation of student satisfaction scores.1. Are the three types of interaction, Internet self-efficacy, and self-regulated learning correlated with student satisfaction?2. Are the three types of interaction, Internet self-efficacy, and self-regulated learning significant predictors of satisfaction?3. What are the effects of student background variables (age, gender, marital status, class level, and time spent online per week) on the three types of interaction, Internet self-efficacy, and self-regulated learning?Method Sample The sample of this study consisted of undergraduate and graduate students enro led in summer-session online courses from the College of Education at a western university.The summer-session courses were 12 weeks long.With the permission of and assistance from the instructors who agreed to have their students participate in the study, an online survey link was distributed to 11 online courses (e.g., Adolescence, Cognitive Psychology, Stress Management, Internet Development) from the disciplines of psychology, special education, instructional technology, and physical education.Of the 291 enrolled students from 11 online courses, 111 completed the online survey, a return rate of 38%.The number of respondents in this study fulfilled the requirements of a multiple regression model with 5 independent variables in which at least 75 pa ticipants




The researcher contacted instructors about their willingness to include their online students in this survey.Different deliveries were utilized by interested instructors to distribute the survey link, including email, Blackboard announcements, Blackboard discussion threads, or some alternative means.The survey was hosted in SurveyMonkey.To increase the response rate, follow-up emails were sent to instructors as a reminder message.Data analyses were conducted using SPSS 16.0.Descriptive analyses were conducted to present the student basic information and the average score of predictor variables and

udent satisfaction.Correlation analysis was perf
rmed to understand the relationship between the three types of interactions and student satisfaction.Multiple regression analyses were performed to investigate whether five predictor variables significantly predict student satisfaction.ANOVA was performed to investigate the effect of student background variables on the five predictors.




The amount of time spent online per week also significantly influenced students' selfregulation level when comparing the students who spent less than 5 hours with those who spent 11-15 hours.Based on the esult, students spending 11-15 hours online per week were more self-regulated than those who spent less than 5 hours.It seemed that 11-15 hours were an adequate amount of time for students to properly manage their learning pace to complete an online course, as opposed to those spending less than 5 hours.Students who spent less than 5 hours online might have rushed through the course content and finished the required tasks without acquiring a deep understanding of the content, given that a more self-regulated

erson would
ave better skills in applying appropriate learning methods and managing the learning process to ensure better acquisition of knowledge.




Instructors should pay attention to students and provide feedback to students in a timely fashion or encourage students to ask questions through different mechanisms.Implementing a technology training orientation before online courses start may help increase students' confidence in performing Internet-related tasks required by the course and in turn enhance student satisfaction.Gender and class level seem to be good indicators of the amount of interaction amo g learners.Instructors are encouraged to design more collaborative activities in undergradu te courses to enhance learner-learner interaction.Time spent online may inform instructors about students' Internet se f-efficacy and self-regulation level.




Table1shows that most respondents were between the ages of 26 and 35 years old.There were only a few students over 46 years old.More than half of the respondents took undergraduate-level courses.Most respondents spent less than 15 hours per week on the online course.
Table 1 InstrumentationStudent Background Information The survey included questions on demographics, five predictor variables, and studentsatisfaction. Demographic questions covered gender, age, marital status, course level,Frequency and the hours spent online per week. This study adopted Internet self-efficacy and self-Percentregulation instruments created by prior researchers (see below). InstrumentGender development was involved for interaction and student satisfaction scales.Male2523%Female8377%Marital statusMarried6964%Single3936%Age18-253734%26-355450%36-451312%46-5544%Above 5600%Course levelUndergraduate level6661%Undergraduat