Modelling IS Student Retention in Taiwan: Extending Tinto and Bean’s Model with Self-Efficacy

Abstract

The purpose of this study is to explore an Information System (IS) student retention model by modifying the integrated model of Tinto and Bean. The proposed model extends the integrated model by adding a construct named self-efficacy which is a psychological factor borrowed from self-efficacy theory.

Data was collected from six private institutions in Taiwan and participants included students studying in the IS discipline in 2009. Structural equation modelling was utilised to validate the proposed model. Although the measures used in assessing the fit of the model obtained reflected the overall strength of the hypothesized model, the present study was not entirely supportive of Tinto’s or Bean’s models.

Based on the significance of the factors in affecting the retention of IS students, the following intervention programs are suggested: (1) promoting self-efficacy programs, (2) provision of career consultancies and services, and (3) improvement of the teaching of IS core courses.

Keywords:

• Self-efficacy
• Information System Education
• Student Retention
• Tinto’s Model
• Higher Education

1. Introduction

The high attrition rate of higher education is a serious issue, and educators need to know what motivates students to stay in or drop out from their courses. In the UK, the rate of non-completion of degrees was 13% in 1982 ∼ 1983 and in 1997 ∼ 1998, the rate was 17% (The Education and Employment Committee, 2001). In Australia, one third of all students fail to graduate and about half of those who drop out in their first year (Department of Education, 2000). In the USA, despite the efforts provided by the federal government and some states of the United States for improving student retention, graduation rates still declined from 58% to 52% in the 1980s and 1990s (M. Scott, et al, 2006). Statistics indicate that in the USA more than 40 percent of all tertiary education entrants leave the institute they enrolled without obtaining a degree, and 75% of these dropout students leave in the first two years of tertiary education (DeShields, et al, 2005). In Taiwan, the dropout rate increased from 9.1% in 2004 to 18.6% in 2008. Moreover, the attrition problem was more serious in private than public institutes (Ministry of Education, 2010). In this study, we focus on finding the most relevant factors and provide information to school administrators of private institutes to facilitate their decision-making.

Student retention has been investigated for various student populations. Different study majors result in various student populations. Students’ study majors have been found to be an important factor on student retention in the higher education sector (Watson, et al, 2004). Research on student retention has been carried out in the nursing major (McConnell, et al, 2004), business major (DeShields, et al, 2005), and computer science major (Biggers, et al, 2008), but not in the Information System (IS) major. In this study, IS is used to refer to academic programmes that teach the framework of information technology and the technology for providing solutions for business.

The IS academic discipline has faced a decline in student enrolments when the job market softened (George, et al, 2005) as a result of the popping up of the Internet bubble, and the resolution of the Y2K problem. Enrolments in IS-related academic programs have declined seriously in the USA since the late 1990s (Granger, et al, 2007). As the number of students enrolling in IS has declined, retaining students in these curricula becomes increasingly important.

Studies of student retention in higher education have witnessed a marked increase over the last two decades. There have been growing interests in the construction of models and theories of student departure to explain the complex interactions of factors that affect student persistence or dropout (Mannan, 2007). Research on student retention has investigated the factors affecting student retention as well as validated the effect of these factors on various student populations. Although the research topic has been studied for decades, it is still very difficult to sort out the influence of the different variables which movements are correlated over time (Aksenova, et al, 2006).

As most of these studies have been performed in western countries, the findings of these studies may not necessarily be applicable to countries with different cultures and educational systems such as Taiwan in East Asia. The distinction between cultures that emphasize individualistic and collectivistic values received most attention in cross-culture psychology (Schwattz, 1994). East Asia is considered as the prototypic representative of collectivist culture, while the Western culture is representative of the individualistic one. The major difference between these two types of cultures is the level of in-group loyalty and identity (Yamaguchi, 1994). The objective of this study is to create an IS student retention model that provides an understanding of the factors that influence IS student retention in Asia.

2. Literature Review

Student retention has been an issue facing higher education for more than 70 years (J.M. Braxton, 2000). Numerous studies focused on predictors and models examining the phenomenon of student attrition. Some research has also investigated the effect of student population on student retention and the results found that it is influenced by the type of student population, such as students enrolled in different types of institutions, in different study majors/disciplines, and with different cultures.

In this section we review the literature of student retention to discuss the most influential models of student retention. They are Tinto’s (1975), Bean’s (1980), and Cabrera et al,’s (1993) models. In particular, Tinto’s model is still the most famous model of student retention. The framework of our proposed conceptual model is based on these models. Another factor of self-efficacy derived from psychological theory is discussed and is integrated into our proposed conceptual model.

2.1 Student Retention Models

One of the oldest student retention models is Tinto’s (1975) Student Integration Model in which he proposed an interaction model that laid the theoretical foundations for research on student attrition. The main concept of Tinto’s model is the level of a student’s integration into the social and academic systems of the college, which determines persistence or dropout. Numerous research have applied Tinto’s theory to student retention. For example, the retention self-study framework (Woodard, et al, 2001) drew heavily from the research of Tinto’s and Bean’s (1980, 1983) models and another model incorporated Tinto’s (1975) model with achievement behaviours (Ethington, 1990). A theoretical model of university student attrition expanded the theory of Tinto (1975) with the development of the institutional integration scales which were designed to measure academic and social integration, commitment to goals of achieving the degree, and institutional commitment (Pascarella, 1980). Several researchers also have used Tinto models to assess the impact of various factors on student retention in higher education institutes (Belch, et al, 2001; Berger, et al, 1998; Braunstein, et al, 2001; J. M. Braxton, et al, 2000; J. M. Braxton, et al, 1997). Tinto’s work has been cited more than 775 times (J. M. Braxton, et al, 2004).

Another influential model is Bean’s model (1980) which was derived from theories of organizational turnover and planned behaviour. Student attrition is viewed as similar to turnover in business organizations. A complex interaction of internal and external variables influences the direction of the student’s intentions and ultimately the decision to leave or persist. The model recognized that factors external to the institution can play a major role in affecting student decisions. Individual higher education student attrition is viewed as resulting from the following variables: student background variables, organizational variables, academic integration, social integration, environment variables, attitudes, grade point average (GPA), institutional fit, institutional commitments/ loyalty, and intention to leave or persist.

Finally, an integrated model which combined both Tinto’s and Bean’s models, was developed (Cabrera, et al, 1993) to offer an integrated framework for understanding the higher education persistence process. The two models claimed that persistence was affected by the integration between the students and the institution. Overall, the results of the integrated model support the propositions claimed in both models. For example, the relationships among academic and social integration constructs, as well as those among commitment constructs, are consistent in both models. Furthermore, support was also found for external factors influencing the academic integration, as well as the effect of encouragement from others on institutional commitment (Cabrera, et al, 1993). Thus, results indicated that when the two theories are merged into one integrated model a more comprehensive understanding of the complex interplay among individual, environmental, and institutional factors are achieved. Also, students’ intention to re-enrol was found to be highly predictive of student retention. This present study uses the integrated model as a framework to examine IS student retention at private higher education institutions.

2.2 Psychological Perspectives on Student Retention

When an individual enters an institution with psychological attributes based on particular experiences, abilities, and self-assessments, these will affect the individual academic outcomes. A psychological model proposed by Bean & Eaton (2001) indicated that psychological processes refreshed the traditional retention models and academic and social integration are taken as outcomes of these psychological processes. Among these psychological factors in that model (Bean, et al, 2001), one important factor is self-efficacy assessments (Bean, et al, 2001). Self-efficacy theory has been thoroughly researched by Bandura (1997) to explore the effect of self-efficacy on student retention. Central to self-efficacy theory, is the concept of self-efficacy which helps determine what activities individuals will pursue, the effort they expend in pursing those activities, and how long they will persist in the face of obstacles. Students with high self-efficacy can always find better strategies to complete the goals/degree, and respond more positively to negative outcomes than people with low self-efficacy (Seijts, et al, 2004). When individuals have self-confidence on task achievement, they develop higher levels of persistence and higher goals for task achievement (Bean, et al, 2001), rather then dropout. In addition, students with higher self-efficacy beliefs reported less physical and psychological distress and higher levels of achievement (Close, et al, 2008). Another research using structural equation modelling to assess the relative importance of self-efficacy and stress in predicting academic performance outcomes, identified self-efficacy to be a more robust and consistent predictor than academic stress (Zajacova, et al, 2005). Strong self-efficacy results in strong persistence. Another study examined a path model of Latino higher education students on student persistence intention and health and the results found that self-efficacy was directly related to persistence intention, and also to social integration (Torres, et al, 2001). These relationships are adapted to the conceptual model of the present study.

Research on modelling student retention has not explored the relationship between Tinto’s model and self-efficacy. In the application of Tinto’s model, research has not yet included the construct of self-efficacy as a predictor of student retention. In this present study, our conceptual model is based on Tinto’s and Bean’s models and we added the factor of self-efficacy to examine a student retention model in the context of a specific student population, namely students studying the IS discipline.

2.3 Effect of Student Population on Student Retention

Tinto’s model has been examined for different student populations, such as students at four-year higher education institutes (Pascarella, et al, 1986) and non-traditional students attending two-year institutions (Nora, 1987; Nora, et al, 1990). However, these results did not totally support Tinto’s arguments. Tinto’s model suggested that integration is crucial for positive academic outcomes. On the contrary, a study on academic performance for first year Australian university students identified that students’ integration to an institution had an adverse affect on academic achievement which was not in agreement with Tinto’s model (McKenzie, et al, 2001). Another study on persistence found that social integration had no significant effect on persistence for full time students (Munro, 1981). In the case of Mexican American university students, neither academic integration nor social integration affected retention rates significantly. Instead, institutional/goal commitments affected student retention significantly more than academic and social integration (Nora, 1987). Even when applying Tinto’s model to different student populations, the contributing factors to student retention were not consistent. In this present study, the objective is to investigate the factors affecting IS students who study in private institutions of technology in Taiwan as these students have less resources than the other types of institutions. The effects of two attributes of student cohort, namely study majors and institution types, are described next.

A number of studies investigated the effect of study majors on student retention. One study (St. John, et al, 2004) showed that African-American students re-enrolled in second year of higher education institutions studying high-demand major fields such as business, health, engineering and computer science are more likely to persist than those in other major fields. Another study (C. Scott, et al, 1996) investigated the differences of dissatisfaction as a reason of dropout between science/technology, art/humanities and business/law students and found a higher level of dropouts among students enrolled in non-traditional subjects (e.g. economics, business and law). Students with science and engineering majors are also more confident in their ability to successfully complete the academic requirements to earn higher grades and are more persistent in their majors (Lent, et al, 1984). Thus, there is evidence that majors of study have significant effects on student retention.

The other attribute that affects student retention is institutional type. According to Tinto’s study, the development of models or methods specific to types of educational institutes is required (1982). Public institutions were found to graduate a larger percentage of students than private ones (M. Scott, et al., 2006). As indicated by Tinto, type of institution may vary the persistence process. In a study employing path analysis methodology, it was found that different patterns result for different types of institutional settings (Pascarella, et al, 1983). Both Astin and Oseguera (2002) and Mortenson (1997; 1998) report on research designed to determine institutional and student characteristics that lead to higher retention and graduation. Astin and Oseguera used regression analysis and pointed out that institution types (public, private, college, university) have an impact on student persistence. Thus, types of institution may have various effects on academic outcomes. The objective of this present study is to investigate significant factors affecting student retention for private institutions.

Even when applying Tinto’s theory to different student populations, research results in finding different contributing factors. Both study major and institutional type are specific attributes of student population tested in this present study. The aim of this study is to investigate a specific student retention model for IS major students studying in private institute of technology in Taiwan. Identifying the factors that contribute to students’ intention to persist can improve the targeting of interventions and support services for students at risk of dropout.

3. Methodology

Based on a review of the literature, we derived a conceptual model, shown in Figure1, which extends the integrated model (Cabrera, et al, 1993) with the self-efficacy factor. Our model consists of eight factors which are: academic integration, social integration, encouragement from others, commitment to the institution, goal commitment (commitment to goals of achieving the degree), financial attitude, self-efficacy and intent to persist. All of these eight factors are measured using questionnaire items which have been used and validated in previous research.

Data analysis was conducted into two parts; the first part consisted of verifying the reliability and validity of the survey instrument while in the second part of the analysis the structural equation modelling (SEM) technique was used to examine the structural relationship among these eight factors. SEM is an appropriate technique to use in this study because it is a multivariate analysis technique that can deal with multiple relationships simultaneously and assess relationships comprehensively. Student retention is a complex behaviour. There are multiple relationships among these factors which affect students’ decision to dropout or not. SEM may bring better understanding to these complex interrelationships among the proposed factors.

Figure 1 Conceptual model of student retention

3.1 Data Collection

The population studied in this research was made up of IS students enrolled at six private institutions from southern Taiwan. These institutions were selected because of similarities in geographical locations and enrolment criteria. The student population consisted of about 1000 first-year and second-year IS students enrolled at these six institutions in 2009.

A questionnaire survey was administered both as an online survey and face-to-face interviews to obtain a sufficient number of respondents for the study. The participants were IS students in their first-year or second-year of study and a total of 404 responses were received from both the face-to-face and on-line surveys. The average age of the respondents was 21-year-old. Some responses from respondents who were neither first-year nor second-year students were deleted. Similarly, responses from students who were not studying an IS major or not enrolled at a private institution were discarded as well. Furthermore, responses with no internal consistency were dropped. For example, a response of “strongly agree” to the question “I plan to re-enrol at this university next semester (Fall 2009).” and “strongly agree” to question “I plan to transfer to another university.” is contradictory and hence is a candidate for deletion from the survey. After data cleaning, the final number of responses was 253.

3.2 Measurements

For measuring these eight constructs, measurement items were selected from several instruments developed by Bean (1980), Cabrera et al. (1993), Kraemer (1997), Nettles et al.(1985), Pascarella & Terenzine (1980), Williams & Coombs (1996), Mussat-Whitlow (2004) and CIRP (Cooperative Institutional Research Program). Academic integration is measured by informal contact with faculty and participation in classes while social integration is measured by peer group relations and faculty relations. In Tinto’s model, both factors are argued to affect an individual’s commitment to the specific institution and the goal of university study completion. The academic grade is measured in academic integration which was pointed out in Tinto’s model. Encouragement from others, as argued in Bean’s model, is measured by the social influence of parents and peers and has a strong direct effect on retention (Bank, et al., 1990). Other studies confirm that encouragement from others affects the decision to persist (Bean, et al., 1985). Commitment to the institution and the goal of graduation are also used in Tinto’s model and the present study. We also used financial attitude as included in Bean’s model since it has been found to affect student persistence (St. John, 2000). We used self-efficacy, as measured by an individual’s confidence in academic and social work, to extend the integrated Tinto and Bean’s model. The dependent variable is the student’s intention to leave was used as a predictor of the persistence behaviour of the student (Cabrera, et al., 1992).

A total of thirty-four items were employed to measure the eight constructs of our proposed student model. The symbolic questionnaire items for measuring each construct are:

Academic integration

• I have performed academically as well as I anticipated I would.

• I am satisfied with my academic experience.

Social integration

• Since coming to this university I have developed close personal relationships with other students.

• It has been easy for me to meet and make friends with other students at this university.

Encouragement from others

• My family approves of my attending this university.

• My family encourages me to continue attending this university.

Financial attitudes

• I am satisfied with the amount of financial support (grant, loans, family, jobs) I have received while attending this university.

• My financial situation is stable.

Institution commitment

• I am confident I have made the right decision in choosing to attend this university.

• I enjoy the study environment.

Goal commitment

• It is important for me to graduate from university.

• It would be helpful for my future career to obtain the certification.

Self-efficacy

• I can carry on conversations well with others.

• I can plan my academic work well.

Intention to persist

• I plan to graduate from this university.

• I plan to drop out of higher education.

3.3 Data analysis

When using multiple items to assess a construct, it is important to examine the reliability and validity of the items used for measuring the particular construct (Cork, et al, 1998). While reliability is concerned with the accuracy of the measuring instrument, validity is concerned with what the constructs attempt to measure. The survey instrument was first purified using item-to-total correlation and any item with a correlation of less than 0.3 was eliminated (Kumer, et al, 1995).

Next, the Cronbach’s alpha coefficient was used to examine the reliability of the measurement items. This coefficient gives an indication of the consistency or repeatability of the measured variables for representing the construct. Although an alpha value greater than 0.7 is the best indication of the reliability of the items for measuring the construct, a value greater than 0.5 is acceptable (Hair, et al, 2006). The threshold value of alpha used in this study was 0.5.

Construct validity is the extent to which the set of measured variables actually represents the theoretical construct being measured. Two measurements were used to examine construct validity. First, convergent validity was used to confirm that the measurement items used for the construct converge sufficiently and share a high proportion of variance in common. Second, discriminant validity was used to ensure that the construct is sufficiently distinct from other concepts. Factor analysis was used to perform both measurements.

Factor analysis is used to directly examine construct validity by factorial validity. Both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) are commonly used for factor analysis. EFA is used to group the items into factors while CFA is used to test whether the data collected fits a proposed model (Zajacova, et al, 2005). In addition, EFA is a posterior concept used to define the underlying constructs and CFA is a prior concept used to validate the scales for measuring specific constructs (Hair, et al, 2006). If the prior theory or hypotheses have been investigated by past research, CFA is recommended to examine construct validity rather than EFA. The survey instrumentation was drawn from a range of sources in the literature, and EFA was then used in the first step to assess the construct validity. CFA is then undertaken in the second step to verify the construct validity. Since the measurement items used in the survey instrument were derived from various work in reported in literature, EFA was used in the first phase to assess construct validity, while CFA was used in the second phase to verify construct validity.

In the EFA phase, factor loadings were used to estimate the convergent and discriminant validity of the instrument. Two criteria are suggested in the literature. First, items with factor loadings of less than 0.5 among all factors are to be deleted. Second, items with factor loadings of greater than 0.5 and appearing for more than one factor are also to be deleted.

In the CFA phase, convergent validity and discriminant validity were used to verify construct validity. Three criteria were used to assess convergent validity. First, factor loadings of all standardized items should be higher than 0.7. Second, the composite reliability (CR) should be higher than 0.6 (Diamantopoulos, et al., 2000; Joreskog, et al., 1989). Third, the average variance extracted (AVE) should be higher than 0.5 (Joreskog, et al, 1989). For discriminant validity, the square root of AVE of each construct should be higher than the correlation coefficients between the particular construct and any other construct (Chin, 1998).

After having established the reliability and validity of the data collected, SEM was used to explore the relationships among the proposed concepts for modelling IS student retention in Taiwanese private institutions. In this present study, eight constructs are expected to be related to each other and SEM was used to model the structural relationships. After obtaining the structural model, three offending estimates were used to test the goodness of fit of the model (Hair, et al, 2006). They are: (1) negative error variances or insignificant error variances for a construct, (2) standardized coefficients exceeding 1.0, and (3) very large standard errors associated with any estimated coefficient. In addition to these three criteria, other measures of fit recommended for comparative fit analysis in educational research (Schreiber, et al., 2006) were also used to verify the model. These include: (1) the Comparative Fit Index (CFI), (2) the Incremental Fit Index (IFI), (3) the Normed Fit Index (NFI), (4) the Tucker Lewis Index (TLI), (5) the Goodness of Fit Index (GFI), (6) the Root Mean Square Error of Approximation (RMSEA), and (7) the Root Mean Square Residual (RMR). These seven indices were used in the study to test the overall fit of the model to the data collected.

4. Analysis Results

While examining the item-to-total correlation, two items were deleted as the coefficients were below 0.3. All Cronbach’s alpha values exceeded 0.5, meaning that the reliability of the data was acceptable. In the EFA phase, the factor loading values of all items were greater than 0.5 and no items with a factor loading value of greater than 0.5 appeared twice for more than one factor. Thus, the survey instrument was deemed to be sufficiently valid in terms of construct validity. In the CFA phase, four items were deleted due to factor loadings of less than 0.7, and two items were deleted because their AVE values were less than 0.7. After examining the discriminant validity, another two items were again deleted due to their square root AVE values being less than the coefficients of other constructs.

After the data was cleaned and reliability and validity established, a path model was constructed and the previously discussed goodness-of-fit metrics were applied to check whether there was a good fit between the proposed model and the data collected. The CFI, IFI, GFI, NFI, TLI were all greater than 0.9, and the RMSEA, RMR were less than 0.05. The seven indices of model fitness all met the criteria of goodness-of-fit.

When examining the parameter estimates for the proposed model, the modification indices revealed that a large reduction in the chi-square can be expected if the structure paths between the following relationships are freed: (1) encouragement and goal commitment, (2) encouragement and social integration, (3) goal commitment and institutional commitment, and (4) self-efficacy and academic integration. The modified model displays a better statistically significant chi-square value and the model and its structural coefficients are shown in Figure 2.

Figure 2 Structural student retention model

Hypothesized effects found to be significant are represented by solid lines while dotted lines represent non-significant hypothesized effects. These hypotheses were found to be strongly significant at the p=0.001, 0.01 and 0.05 significance levels, except for the five hypotheses that were not significant.

Intent to persist was found to be directly affected by self-efficacy (γ=0.199), goal commitment (β=0.359), and institutional commitment (β =0.701). Self-efficacy also had an indirect effect on intent to persist via goal commitment (0.430*0.359=0.154), social integration, and institutional commitment (0.674*0.223*0.201=0.030). Thus, the total effect of self-efficacy on intent to persist was 0.384 (0.199+0.154+0.030). The largest total effect on intent to persist was accounted by self-efficacy (0.384), followed by goal commitment (0.359), academic integration (0.225), institutional commitment (0.201), encouragement (0.134), financial attitudes (0.099), and social integration (0.045) as shown in Table 1.

Table 1 Standardised total effects on ‘intention to persist’

Construct	Total effect
Self-efficacy	0.384
Goal commitment	0.359
Academic integration	0.225
Institutional commitment	0.201
Encouragement from others	0.134
Financial attribute	0.099
Social integration	0.045

In turn, encouragement had an indirect effect on intent to persist via institution commitment (0.260*0.201), academic integration and goal commitment (0.360*0.307), and academic integration and goal commitment (0.360*0.307*0.359). The total effect of encouragement on intent to persist was 0.134. Similarly, the total effects of financial attitudes, academic integration, and social integration on intent to persist were 0.099, 0.225, and 0.045 respectively.

In another perspective, academic integration had a direct effect on social integration (0.145), institutional commitment (0.540), and goal commitment (0.307). Social integration had a direct effect on institutional commitment (0.225).

Contrary to expectations, self-efficacy had no effect on academic integration, social integration had no effect on goal commitment, and goal commitment had no effect on institutional commitment. In addition, the explained variance determined by the Square Multiple Correlation (R2) for each endogenous variable of the obtained model was as follows: intent to persist was 38.1%, goal commitment was 36.1%, institutional commitment was 70.1%, academic integration was 48.1%, and social integration was 53.6%.

Thus, the results partially support the proposed model. The structural relationships between academic and social integration factors, and between financial attitudes and academic integration, were consistent with Tinto’s and Bean’s theoretical frameworks. The results also showed that the external factor, encouragement from others, had direct effects on academic integration and commitment to the institution.

5. Discussion

The findings revealed that the three most important factors that affect students’ intention to persist are: (1) self-efficacy, (2) goal commitment and (3) academic integration, and these are discussed in turn below. In particular, self-efficacy is found to be strongly associated with a student’s decision to drop out.

5.1 Effect of Self-efficacy on Student Retention

Self-efficacy was found to have the most important effect on intent to persist and hence it is regarded as the most significant construct determining IS students’ intention to persist. Our finding of the link between higher self-efficacy and higher intention to persist is consistent with prior studies (Close, et al., 2008; Lent, et al., 1986; Moulton, et al., 1991). Although self-efficacy has great influence on an individual’s decision to withdraw from higher education studies, so far it has not been investigated within the framework of Tinto’s model. Thus, the main contribution of our work is the extension of Tinto’s model with the psychological self-efficacy construct.

In terms of implications for practice, this finding suggests that intervention programmes designed to enhance IS students’ intention to persist should promote self-efficacy. Four sources of self-efficacy that are essential to consider are: (1) personal performance accomplishments, that is the experience of successfully performing the behaviours in question (2) vicarious learning, (3) social persuasion such as encouragement and support from others, and (4) emotional arousal, such as anxiety and sweating (Bandura, 1989). Three of them are considered for enhancing self-efficacy in the study.

Regarding both personal performance accomplishments and social persuasion, emphasis on building self-beliefs through verbal persuasion methods (Pajares, 1996) is suggested as intervention. In the early stages of learning the IS discipline at private institutions in Taiwan, students encounter difficulties with learning new courses, such as programming and database concepts. Performance feedback informs students of progress in their studies, strengthens self-efficacy, and sustains motivation (Schunk, et al., 2001). Thus, positive effort feedback and ability feedback to students on programming homework may increase self-efficacy.

Regarding vicarious learning, students improve their self-efficacy by watching others who are similar to themselves succeed in tasks. Team projects would be suggested to be designed for IS students. As team projects can be used as means to group students together for pursuing the same goals, they could learn by watching team members. Students forming groups in classes have also been found to have a significant positive effect on student retention (Johnson, 2000–1).

5.2 Effect of Goal Commitment on Student Retention

The intention of IS students to persist was also found to be very much affected by goal commitment. This is probably because IS/IT jobs are more attractive than other jobs and hence IS students are more committed to the goal of obtaining an IS/IT job. This finding suggests that interventions such as organising future career seminars would enhance the goal commitments of students. Information on workplace and job opportunities in the IS/IT field would be highly beneficial to students of this discipline.

Although the links between encouragement from others, social integration, and institutional commitment, on goal commitment were not significant, goal commitment still had a strong relationship with persistence. This is because IS students have strong beliefs about their future career and they are not affected by external factors.

5.3 Effect of Academic Integration on Student Retention

Academic integration was found to affect institutional commitment (0.540), goal commitment (0.307), and social integration (0.145). These relationships are consistent with Tinto’s theory. The high correlation between academic integration and institution commitment can be explained by the fact that interaction with faculty and good academic performance affect commitment to students’ goals and the institution they are studying at. When students get along with the teaching staff and they obtain good academic performance, they feel satisfied with the institution and are committed to their future goals. In Tinto’s model, academic and social integration are the two most important factors for the retention of higher education students.

Although students may withdraw from their courses for a variety of reasons, IS students who leave because of misunderstandings or negative experiences would be worthwhile to retain by offering them more information and seminars designed to provide a better understanding of the discipline. Teaching only elementary IS/IT concepts may increase the risk that students view IS negatively as a potential area of study. Using mini case studies in class to actively engage students in their own learning would help them discover principles that are important in the context of IS and they can learn to be responsible for their own learning (Mukherjee, 2000). If students have a better understanding of what they will learn for the four-year duration of their degree, they may be more likely to persist in their studies.

With regard to the practitioner’s perspective, the present study provides an integrated understanding of the factors that affect IS student retention in private higher education institutions in Taiwan. Intervention strategies must address those constructs that are found to be contributing predictors of intention to persist. Other disciplines may use this study as a starting point to investigate student retention in their specific disciplines.

To further validate the results obtained above, a qualitative approach using face-to-face interviews is recommended for future work. Conducting interviews provides an opportunity to explore the complex and persistent behaviours, attitudes, beliefs and reactions from the perspective of different respondents (Denzin, 2009).

6. Conclusions

A modified version of Tinto and Bean’s integrated model of student retention was tested on IS students in private higher education institutions in Taiwan. This study tested an extended version of Tinto’s model of student retention with self-efficacy to explain IS student retention in Taiwanese private higher education institutes. A questionnaire survey (administered by face-to-face interviews and an online server) was used to collect data from six private institutions located in southern Taiwan. Participants were students commencing their first year or second year of study in the IS discipline in private institutions in 2009. After data cleaning, 253 valid responses were obtained from first- and second-year IS students. SEM was used to examine the relationship among the factors in the proposed model, and significant relationships were found to affect students’ intention to persist.

The most important three factors affecting students’ intention to persistence were found to be self-efficacy, goal commitment, and academic integration. Self-efficacy was found to be strongly associated with students’ decision to dropout, followed by goal commitment and academic integration. However, contrary to expectations, self-efficacy did not predict academic integration among IS students to study in private institutions.

Based on these findings, intervention programs that take advantage of these factors were suggested. Strategically designing interventions to enhance students’ self-efficacy is especially crucial for IS students in private higher education institutions. Career consulting services and workforce seminars are recommended to keep students informed of the latest workplace demands. Providing more information on future career opportunities would be more likely to help students persist in their studies. Finally, from an academic teaching point of view, enlightening students’ concept of the IS discipline to prevent misunderstanding of the IS role (for example, learning to solve business problems rather than programming technology) would also make students more likely to continue their studies.