Factors Influencing Students’ Decision to Drop Out of Online Courses in Brazil

In recent years, e-learning has been the fastest growing educational form in students' numbers, and this industry's market revenue (Lee, Choi, &Kim, 2013). Despite this growth, concern about the significantly higher student dropout rate of students in online courses as compared with conventional learning environments has increased. Brazil has also registered a significant increase in the number of students interested in this type of education, but the dropout rate is a considerable concern to institutions. This study’s objective was to identify the relevant variables behind online students’ dropout decision in Brazil. After a literature review that determined the ten most recurrent and relevant variables, we heard professional e-learning experts. They indicated, from their standpoint, what the most pertinent variables influencing dropout would be. Based on this, we conducted a quantitative survey with e-learning students, considering the factors indicated in the literature on this subject and educational professionals’ indications. This study's contribution was to verify that the quality support is extraordinarily relevant and has a high correlation with students' perception of Usefulness, the quality of Course Content, and ease of System Use.


Introduction
The rise of e-learning offerings and demand in Brazil has increased institutions' concern about student dropout rates. This study aims to identify factors influencing students' decision to drop out of online courses in Brazil.
The analysis proposes an e-learning model based on the literature review and a discussion with e-learning monitors, executives, and experts, who indicated the dropout rate's primary influencers. A quantitative survey occurred with e-learning students, using the variables indicated by the experts and the literature.
This study's theoretical contribution is combining theories from different research fields, including learning psychology, TAM and System, and Information (IS) failure. The proposed model addresses technical aspects, including the Information system success model (DeLone & McLean, 2003), the Technology Acceptance Model (Davis, 1980), and students' psychological aspects. It also contemplates the cultural dimension (Hoststade, 2001b) intrinsic to the learning style. In the model, service quality refers to teacher and tutor support to students, and it has a moderating effect on the perception of the system, information quality, and Perceived Usefulness.

Dropout Definition
The definition of dropout can have different meanings. A student dropout (or non-completer) can be any student leaving his or her institution of initial enrollment (Spady, 1971) or those deciding to withdraw from e-learning courses (Haydarov, Moxley, & Anderson, 2013;Levy, 2007). Levy (2007) defined dropout students as those who voluntarily withdraw, incurring financial penalties. Those opting to drop a course during the "add/drop period" are not considered "non-completers" because their tuition is fully refunded. Bean (1979) used the term attrition to refer to students who leave an institution without completing a given program. Attrition represents a reduction in the number of students remaining enrolled during the reported calendar year. The attrition rate includes dropouts who are either students withdrawing themselves from the program or those dismissed from the university's program.
The attrition rate includes students who may be in good academic standing but did not enroll in the reporting year (Haydarov, Moxley, & Anderson, 2013).
For this study, the concept of dropout students or non-completers (Levy, 2007) refers to all students not completing a particular course or leaving voluntarily without reaching the goals of an e-learning course, for whatever reason. Dropout is the difference between the number of students enrolling in a course during the base calendar year and those completing the course. This study focuses on course dropout and not program dropout and considers only the students' voluntary withdrawal decision. Forced dropout, decided by the institution, due to low grades or failure, was not recognized as a student-motivated dropout decision.

Goal Commitment
Eskreis-Winkler et al. (2014) studied the effect of goal commitment or passion and perseverance for long-term goals on the retention of students in high school. The Academic goal commitment refers to the ability to work independently, maintain motivation, sustain focus on personal and academic goals, and maintain motivation even when they have other life commitments despite conflicting commitments (Holder, 2007). A student's academic goal commitment is a significant and robust predictor of an intention not to drop out. Students with high academic goal commitment were less likely to drop out of high school (Strom & Savage, 2014).
It is a direct influencer of dropout decisions (Tinto, 1975;Pascarella & Terenzini, 1980), and a significant and robust predictor of an intention not to drop out. Students must be motivated to attend an e-learning course. Goal Commitment can have different motivational facets. Student motivation can be Intrinsic when involved with the activity; its enjoyment affects the students' study approach. Extrinsic factors referring to the search for external approval of recognition and their level of commitment to attaining some specific goal are also motivational elements (Lepper, 1988). The intrinsic motivation is particularly important for non-traditional students, most of them adults, because they consider the relevance of skills and subject matter to their career and life goals (Kember, 1989).

Learning Style
In the literature review, the learning style was considered relevant to students' decision to persist or drop out (Rovai, 2003;Willging & Johnson, 2004;Holder, 2007;Hart, 2012;Lee, Choi, & Kim, 2013;Moore, 2014). Learning Style supports learners' self-discipline, self-motivation, and responsibility (Kerka, 1996apud Rovai 2003). Aparicio et al. (2017) have proven that student energy and vigor inherent to their learning style can affect individual performance and e-learning success. The study gathered the student's psychological characteristics of Grit, defined by Duckworth et al. (2007) as the intrinsic personal motivation to persist and pursue long-term objectives and information system variables. Rovai's Composite Persistence Model (2003) highlights the direct effect of internal factors such as learning style, goal commitment, and self-esteem on the decision to persist. Holtbrügge and Mohr's (2010) findings showed that individuals' cultural values affect their learning style preferences. In the same way, Guild (1994) states that culture can facilitate the learning process.
As the online offering format tends to be globalized, a different cultural background can affect students' satisfaction and decision to persist or drop out of the course. Kizilcec and Halawa (2015) compared dropout in online learning in different geographic areas. They found substantial differences in their level and grade performance among students in two different region groups (Latin America, Asia, and Africa) than the second group of countries (North America, Oceania, and Europe). Online students in the first group were only half as likely to persist, and they presented substantially lower grades than those in the second group.
As the research took place in Brazil, we considered it relevant to include this variable in the model, considering the possible links between the culture and learning style.
Culture can influence people's values, motivational factors, personal goals, and determination to reach them (Teperino et al., 2006;Kizilcec & Halawa, 2015;Bandura, 2001). People's behavior is a consequence of social system influence sharply defined by culture. It can predict human behavior, influence their judgment regarding their capabilities, and focuses on goals (Bandura, 2001).
The growth in e-learning business across borders, attracting students from different countries and cultures has brought some concern on which format would best serve the different learning styles. The search for a one-size-fits-all model is a challenge; however, the convergence of education systems seems to be more necessary to participate in the globalization of educational institutions, especially in the e-learning format, which is not affected by the country's borders.
As culture can influence learning style, it is relevant to consider cultural aspects and Brazilian culture's specificities.

Support Quality
The main variables of e-learning are human beings and tutors, and they still have an essential role in engaging and motivating students, even on a technological platform (Nerguizina et al., 2010). Support quality may be related to practical assistance with technology and computer (Hart, 2012) and to pedagogical support influencing how learning and teaching should be delivered considering interaction, collaboration, and communication between learners and teachers in the e-learning environment (Masoumi & Lindström, 2012).
Although technical support does not predict persistence, dropout students present a lower satisfaction rate with support services (Ivankova & Stick, 2007). Students may have different levels of computer skills, and tutorial support may be helpful. As in the online environment, there is isolation from faculty; this issue could be addressed, ensuring personal tutors' support. In the e-learning environment, support quality may encourage students to use the system because it provides willingness and attention to address students' difficulty (Aparicio et al., 2017).
As students may face different problems in online platforms, the effectiveness of the support can affect student perception of quality, which is critical for academic success and retaining students. The majority of students reported support quality as one of the most critical factors in persisting (Bunn, 2004).
TAM Model applied toE-Learning Davis' (1989) Technology Acceptance Model (TAM) had the goal to explain the determinants of computer acceptance and their influence on user behavior. TAM confirms that perceived Usefulness is of primary relevance for computer acceptance behavior. Perceived Usefulness had a significant effect on the intention to use the computer. The TAM-Technology Acceptance Model (Davis, 1989) refers to identifying inhibitors to technology usage intention, an essential element in the e-learning environment.

Perceived Usefulness
Students perceive the Usefulness of online courses as the practical value or utility and relate it to the Intent to leave. This factor, previously indicated as an essential element to retain traditional students (Bean, 1979), was also confirmed as a critical variable in the online dropout decision (Davis, Bagozzi, & Warchaw, 1989). Perceived Usefulness strongly influences people's intentions to accept computer system usage. Students believe that education used advantageously will lead to self-development, employment, or better job performance (Davis, Bagozzi, & Warshaw, 1989;Bean, 1979;Davis, 1989).

System Failure
System design can be a big issue for e-learning if students do not get what they desire from the course or have to put more effort than expected, generating dissatisfaction (Chee-Wee Tan, Benbasat, & Centefelli, 2016). System quality, including responsiveness and reliability of response, can be inhibitors and have a negative and higher impact on technology usage intention (Cenfetelli & Schwarz, 2011;Willging & Johnson, 2004). There are statistically significant differences between completers and the dropout group (Holder, 2007). Some of them   An extensive questionnaire consisted of 37 questions focusing on control variables and independent constructs. Students rated each item on a five-point Likert scale ranging from "Strongly Disagree" to "Strongly Agree".
At the end of the questionnaire, three open questions asked the students to share their personal experience with the online course they dropped out.
Cronbach's Alpha used to measure internal consistency indicates how closely related sets of items are grouped. Except for the Academic Goal and Learning Style variables, the other variables presented a reliability coefficient of .70 or higher, which is considered "acceptable" in most social science research situations. The average amount of variance (AVE) to assess discriminant validity results were higher than 0.50, according to Fornell-Larcker (1981) criteria. The Coefficient of Determination (R2) analysis to measure the accuracy of the model was performed to measure the variance proportion of the dependent (dropout) variable. The rates found for the independent variables (Usefulness, information failure, and system failure) were close to 0.50, showing that the model's inputs can explain approximately half of the observed variation. Multicollinearity or intercorrelations were tested among predictors. The results show a lack of multicollinearity issues. All the variation inflation factors (VIF) suggested by Tabachnick and Fidell (2012) were less than 0.90.  Vol. 9, No. 5; pedagogical support influencing how learning and teaching should be delivered considering interaction, collaboration, and communication between learners and teachers in the e-learning environment (Masoumi & Lindström, 2012). Although technical support does not predict persistence, dropout students present a lower satisfaction rate with support services (Ivankova & Stick, 2007).

Conclusions
E-learning in the world became a fundamental methodology of learning. It gives people the chance to update their knowledge, develop an understanding of new things occurring abroad, and become more competitive.
In Brazil, it also became a phenomenon. There are more students registered in online classes than in mortar-and brick classes. However, for Brazilian society, the institutions, and the students, it is a new experience. There is still little comprehension of how it works, the level of difficulty, the best technological platform to offer, the depth of content, and its design more appropriate.
This study's significant contribution was the use of the construct support quality as a moderator variable affecting three constructs already identified in previous studies as relevant in students' course evaluation. This discussion on the dropout analysis suggested that support quality had a strong influence on system failure, information failure, and perceived Usefulness and not directly on dropout decision.
The online course environment's quality support is determinant to minimize difficulty that students face and evaluate the course quality. That conclusion reinforced Souza's study (2009). Her work concludes that factors such as the quality of interaction between students and teachers and information quality are the leading causes of dropping out.
Due to the little experience, dropout is a definite concern. The model "one fits all" is not adequate. Some people talk about artificial intelligence in learning that would offer a methodology appropriate for each student, allowing him to develop knowledge at his own pace.
Brazil has 210 million people and is the fifth extensive territory in the world. There is also a gap in education that makes e-learning a vital tool to attend the whole population all over the country.
Based on this study, it is recommended that institutions focus on the quality of their e-learning platform, the quality of the course content and design focusing on the Usefulness, competencies, and practice. It is also necessary to invest in support quality to help students overcome their eventual lack of goal commitment and weakness of their learning styles.

Practical Implications
As an academic contribution, this study gathered students and expert opinions in the same research considering the two different points of view. That combination helped to emphasize the importance of some factors to dropout decisions.
This study also innovated gathering theories from four different fields: Individual characteristics (locus of control, goal commitment, and learning style); Institution support (social integration and support quality); The impact of Information System Failures (Information and System Failure) that has been studied in an e-commerce website, and Technology Acceptance Model (Usefulness, Ease of Use and Subjective Norms).
Another significant contribution in this study was using the construct support quality as a moderator variable affecting three previously identified constructs in previous studies as relevant in students' course evaluation. The relationship between support quality to the perception of system failure, information quality, and Usefulness is relevant. It highlights the importance of the support quality to balance any possible inefficiency or student difficulty with the platform, content quality, and students' perception of the course's Usefulness for their lives.
Students demand frequent attention from professors and tutors. This support indirectly influences the dropout decision, influencing the perception of course practicality, quality of the system, and content.
It is not possible to state that this need for student support is due to Brazilian cultural characteristics. However, with the proliferation of online courses in Brazil, the observance of this need for support quality becomes relevant to minimize the dropout rate.

E-learning in Brazil is still a new business, and institutions
have not yet found the best format and density of content and design and the best tools to keep students motivated.
There are common factors affecting students' decision to persist or drop out, showing the relevance of those factors. Factors such as the course content and course design, the Usefulness of the content to student's job activities, the institutions' support on content, and technical issues are of great importance in students' evaluation.
As already stated by Nerguizina et al. (2010), even though e-learning occurs on technological tools, human beings and tutors have an essential role in engaging and motivating students. The findings confirm Rothkrantz (2017) theory that monitors and professors can impact students' achievement, motivating, and creating good emotions in them in an online environment.

Limitations
This need for support may be related to Brazilians' characteristics discussed earlier, along with the study. It would be necessary to test the hypothesis in another country with a different culture to generalize these findings.
Brazilians are collectivists and have a low level of long-term orientation, usually avoiding uncertainty, not taking risks. E-learning quality and validity is uncertain and can cause distrust. Besides that, working alone is not part of Brazilian culture. They need to interact, socialize. Brazilians have a reduced orientation for action and organizational planning, and they manage time inefficiently, one of the requisites of the online environment.
Interpersonal relationships with peers can help students to build self-esteem. However, e-learning can be a lonely experience, and integration in the academic community may become difficult.
Brazilian culture defends hierarchical differences based on teacher and peer support. Onsite classes are more likely to meet the needs of their learning style.
Although the relation between culture and learning style has been discussed, Learning Style did not correlate with dropout. The sample was not adequate to prove that there is a difference in learning styles in different countries. The study occurred in only one country and one institution. It would be valuable to test the model in classes where the number of international students is high. It would probably be possible in well-known institutions where the courses are in English, a more disseminated language.
This concern for learning style can be valuable for institutions, considering that e-learning is an intense globalization process.
The survey focused on students in short-term courses, eight weeks at the most, that today represent most of Brazil's online courses.
However, in 2018, the number of online undergraduate courses offered in Brazil accounted for 53% of the total undergraduate courses provided according to the census that registered a low completion rate of 37% (Censo, 2018). As the number of undergraduate courses online has already surpassed the number of undergraduate courses in person, a further study focusing on this type would be recommended.
Theoretically, as an undergraduate course requires the student to complete all the disciplines to acquire the diploma, it could be assumed that the dropout rate would be lower. However, the statistics have not proved this expectation.
Another suggestion for further studies would be to analyze the dropout rate, specifically in longer-term courses like the Graduate Specialization (MBA) courses, to test the validity of the correlations encountered. The demand for MBA courses is increasing, and they are of great interest to the institutions because they generate more revenue.