Introduction

Students transferring between postsecondary institutions are an integral piece of the higher education system in the United States (Cheslock 2005). Approximately 3.6 million students enrolled in postsecondary institutions in 2008 and 37.2% of these students attended a different institution at least once over a period 6 years (Shapiro et al. 2015). Existing studies tended to focus on community college students and their transfer to baccalaureate-granting institutions (e.g., Melguizo and Dowd 2009; Melguizo et al. 2011; Townsend and Wilson 2006; Wang 2009). According to the Chronicle of Higher Education (Gonzalez 2012), among students enrolled in 4-year public institutions, about 50% of them transferred from their 4-year institutions to 2-year public institutions.

While Shapiro et al. (2015) found that “among those who started at 4-year institutions, those who started at 4-year public institutions had the highest transfer rate (36.5%) followed by those who started at 4-year private nonprofit institutions (34.3%) and private for-profit institutions (22.9%)” (p. 8), very little is known about students who were initially enrolled in 4-year institutions and transferred to another institution beyond descriptive statistics.

Li (2010) noted that these students have not received much attention for a variety of reasons. At an institutional level, it made some sense to label students who departed from their original institution of enrollment as “drop-outs” when analyzing the reasons for departure from that specific institution. However, Hossler et al. (2012) succinctly described the issues and consequences of this orientation to researching student transfer:

Students who leave are often counted as lost to attrition. The consequence of this is that we lack the complete story of where students came from, and what happens when they leave. In other words, when studies follow institutions as opposed to students, they can talk about where students start but not where they go. The statewide student unit record databases available in many states have the capability to address this issue. However, such tracking currently is limited to only those students who move within state boundaries and only within public institutions. (p. 11)

Research on students whose motivation was to switch institutions and not to leave higher education altogether was overlooked when transfer was examined organizationally rather than at an individual level (Porter 2003). Students transferred for a variety of reasons including having experienced academic difficulties at the first institution (McCormick 2003; McCormick et al. 2009), the opportunity to complete coursework at a decreased cost at another institution (Goldrick-Rab and Pfeffer 2009), to increase their fit at another institution elsewhere (Hossler et al. 2012), and/or relocate to a program that better aligns with their interests (Hossler et al. 2012; McCormick et al. 2009).

To understand more fully the student transfer phenomenon, it was critical to examine the various factors that influenced student transfer while also incorporating the time-varying nature of influence of those very factors. Factors that affected student transfer may increase, decrease, or remain constant over the course of a student’s tenure at an institution. It is reasonable to assume that the factors that influenced transfer during a student’s freshman year may be very different than the factors that influenced transfer in a student’s junior year. Additionally, a factor’s influence on transfer may wane over the course of a student’s enrollment at an institution. For example, departure related to college GPA may be relatively low in a student’s freshman year as the prospect of increasing the GPA remained viable. However, a student in their junior year may come to terms with the fact that they would not be able to gain admittance to a particular program they were interested in and decide to transfer because of their inability to increase their GPA due to time constraints.

Heeding the call of Hossler et al. (2012) and others, this study aimed to interrogate the student transfer phenomenon to provide insight on the factors that preempted students’ decisions to depart 4-year institutions and transfer into another institution of higher education (Kearney et al. 1995; Porter 2003). By using the nationally representative postsecondary education data, the current study longitudinally investigated transfer students originally enrolled in 4-year institutions and factors associated with their choice to transfer from those 4-year institutions. Coupled with event history modeling, this study investigated the timing of transfer-out behavior at 4-year institutions over time.

Transfer Students

Despite calls for additional studies to identify and research factors that influence students to transfer, research in this realm was sparse at best (Crisp and Nora 2010; Kearney et al. 1995; Porter 2003). The literature that existed tended to focus on community college students who transferred to baccalaureate-granting institutions (e.g., Ishitani and McKitrick 2010; Kirk-Kuwaye and Kirk-Kuwaye 2007; Melguizo and Dowd 2009; Melguizo et al. 2011; Townsend and Wilson 2006; Wang 2009). This overwhelming focus on vertical transfers, who accounted for less than 50% of transfer students at many 4-year institutions, creates a body of literature that neglects a sizable portion of students that start at 4-year institutions and, for some reason, transfer to another institution (Kirk-Kuwaye and Kirk-Kuwaye 2007). Literature detailing horizontal transfers, students transferring from a 4-year institution to another 4-year institution, or reverse transfers, students transferring from a 4-year institution to a 2-year institution, was almost nonexistent although this demographic represented a sizable proportion of students (Alpern 2000; Kirk-Kuwaye and Kirk-Kuwaye 2007). “The motives behind horizontal transfer, by contrast, are far more varied, including unsatisfactory academic performance, academic, personal, or social dissatisfaction, financial difficulty, and pursuit of programs unavailable at the first institution, to name a few” (McCormick et al. 2009, p. 1). Li (2010), using a nationally representative data set to study degree completion of transfers, discovered that students who “break their enrollment are 70% less likely to complete their degrees within 6 years than students who stay” (p. 233).

Given the substantial share of undergraduates at baccalaureate-granting 4-year institutions who transfer, the researchers deemed it critical to analyze the factors that preempted students’ decisions to transfer and determine which students were most likely to ultimately transfer by year (Laanan 2001; McCormick et al. 2009; Townsend and Wilson 2006). Alpern (2000) suggested using “longitudinal studies that analyze national databases that track enrollment patterns of persisters and dropouts… (to) provide researchers with important data that identify influential factors reported by community college transfer students” and horizontal transfer students about their reasons for transferring” (p. 22).

Conceptual Framework and Factors Affecting Persistence in College

Tinto’s (1987) student integration model served as the conceptual framework for the current study. Tinto’s theory was widely-accepted, lauded, and ubiquitous in the study of college student departure (Braxton 2000). Tinto (1987) attributed a student’s decision to remain at or depart from an institution to their social and academic environments within his or her institution. Tinto’s model suggested that students enter a postsecondary institution with a level of commitment to the institution that, in turn, affected how integrated they become within the institution. Integration was separated into two, distinct types: academic and social. The level of integration that a student achieved, in regard to each type, directly affected their decision to stay or depart from their institution. A number of researchers (Astin 1993; Pascarella and Terenzini 2005; Tinto 1987) have discovered that students who bond quickly and well at their institutions were more likely to succeed than those who do not.

Academic Integration

Lundquist et al. (2002) defined academic integration as a student’s general level of engagement in academic opportunities and activities, and they discovered that faculty support, accessibility to students, and quick responses to student questions were positively associated with students’ decisions to remain in school. Earlier, Stage (1988) studied first-year retention and found that the father’s educational attainment had a positive effect on the student’s academic integration that, in turn, had a significant influence on a student’s persistence to second year.

More recently, Swecker et al. (2013) discovered that academic integration, measured by the number of advising sessions, was significantly associated with student persistence. College GPA was a unique indicator of academic integration as it was simultaneously related to academic motivation and persistence (Allen et al. 2008). Tinto (1975) noted that “grade performance becomes… both a reflection of the person’s ability and of the institution’s preferences for particular styles of academic behavior” (p. 104).

Social Integration

Social integration, a student’s relationship to their peers and faculty members, was another essential component of Tinto’s model. Tinto hypothesized that students who were socially integrated were more likely to develop a commitment to and to experience a sense of belonging at their institution (Morley 2003). According to Cabrera et al. (1993), financial aid bolstered the levels of social integration that underlined the student’s decision to persist. Berger and Braxton (1998) investigated the effect of social integration on persistence at a highly selective private research institution. They found that organizational attributes such as fairness in academic and social rules and regulations significantly influenced the level of social integration. Woosley and Miller (2009) suggested social integration may play a key role in promoting an institutional commitment in students and, as a result, in influencing students to persist. Recently, Ishitani and Reid (2015) explored first-to-second year persistence at 4-year public and private institutions. They discovered social integration had a significant and positive effect on retaining students.

While a number of studies examined the effect of academic and social integration on student persistence, there is a dearth of information about the effects academic and social integration have on students’ transfer-out decisions from 4-year institutions. In part, this may be due to the proclivity of retention studies not to censor the varying types of departure. Instead, transfer students were typically aggregated along with students with other types of departures. Therefore, it was necessary to rely primarily on empirical findings found in existing retention studies to begin to understand and interrogate transfer-out behavior. However, some evidence suggested that academic and social integration played an integral role in students’ decision to transfer. For example, in the earlier report by Peng (1977), about 32% of students transferred from their 4-year institutions because they would like to attend school where they felt more like they belonged. Transfer students also departed institutions due to a change in their interests or their former school not having their desired programs or coursework (Li 2010).

Additional Relevant Theories

Astin’s (1975) theory of student engagement suggested that student departure be examined with consideration given to student’s background characteristics, academic ability, institutional environment, and other factors that influence the college experience. Bean and Metzner’s model (1985) built upon previous work by incorporating academic and other outcomes. They provided a revised model that conceptualized the interactions between background characteristics, academic variables, environmental variables, academic outcomes, intent to leave, and, ultimately, the decision to depart or remain. The work of Nora (1990) and Porter (1991) demonstrated the underlying importance of including financial aid variables in studies involving student persistence. Pascarella and Terenzini (2005) also called for the use of financial aid in studying persistence and recommended that race, gender, peer influence, and institutional attributes be included in future analyses. Relevant factors included in this study included race/ethnicity, parental education, student achievement, and student financial aid.

Other Student-Level Factors

The American higher education system has been characterized by the vast disparities and inequality of opportunity between racial and ethnic groups (Chen and DesJardins 2010). Students from racial and ethnic minority groups were at an increased risk for dropping out of college (Braxton et al. 1988; Carter 2006; D’Lima et al. 2014). African American, Hispanic, and American Indian students were found to be much more likely to dropout than their Asian American and white counterparts (Murtaugh et al. 1999). Ishitani (2006) found that these racial differences remained among first-generation students. Despite sharing first-generation status, African American and Hispanic students lagged in degree completion when compared to their white counterparts. Research on race and college transfer has primarily been concerned with the experiences/perceptions of students following a vertical transfer (Sólorzano et al. 2005; Wawrzynski and Sedlacek 2003), degree attainment post-vertical transfer (Wang 2009), and the academic/social integration of students following a vertical transfer (D’Amico et al. 2014; Jackson and Laanan 2015). However, Soares and Watson’s (2016) work using the National Longitudinal Survey of Freshman, a sample of 3924 participants at 28 selective universities, found that race did not play a significant role in a student’s decision to transfer.

In addition to race/ethnicity, varying levels of parental educational achievement have been explored to better understand how students’ backgrounds affect persistence. Parental education has been found to be a strong predictor of student persistence. First-generation college students, defined as students who had neither parent graduate from college, were more likely to drop out during their freshman year (Ishitani 2003), and, ultimately, were less likely to earn a college degree (Chen 2005). First-year, first-generation students had a 71% higher risk of dropping out than students with both parents who had college degrees (Ishitani 2003) and had significantly lower academic aspirations (McCarron and Inkelas 2006). Taken together, the cumulative effects of these factors affected first-generation students’ persistence to a 4-year degree. In their study of traditional-aged students who started college at 4-year postsecondary institution, Goldrick-Rab and Pfeffer (2009) found that students with parents who had above a bachelor’s degree were much less likely to transfer from a 4-year institution to a 2-year institution. They noted, “While nearly one-fourth of the children of parents who did not finish high school left their initial 4-year school to reverse transfer to a community college, we observed that pattern among less than 7% of the students whose parents had professional or postgraduate degrees” (Goldrick-Rab and Pfeffer 2009, p. 111). Conversely, Soares and Watson (2016) noted that parental education was not a significant predictor of transfer in their study.

Student’s precollege academic achievement on the ACT or SAT has been shown to have a measurable effect on student retention from their first year to their second year (Johnson, 2006). DeBerard et al. (2004) echoed this idea when they stated that, “universities which are more selective in terms of high school GPA and SAT should expect greater achievement and retention among their freshman” (p. 73). However, the significance of the ACT and SAT appear to wane as students’ progress past their early years and college achievement begins to take precedence. Campus-based studies have found that college GPA has a direct effect on student persistence, while high school GPA and other measures of ability do not (Nora and Cabrera 1996; Okun et al. 1996). Students’ academic capability has been shown to be a significant factor in students’ decisions to withdraw from an institution (Li 2010). Measures of academic achievement from students’ first-year, GPA and adequate progress in accumulation of credits, have been shown to be negatively correlated with reverse transfer. Additionally, college GPA has been shown to mediate the effect that parental education has on reverse transfer (Goldrick-Rab and Pfeffer 2009). Soares and Watson (2016) found that freshman GPA was a significant predictor of transfer among their sample.

The constantly rising costs associated with college attendance make financially related issues increasingly important to a student’s decision to withdraw or persist. Most studies have identified an inverse relationship between tuition/fees and student persistence (Paulsen and St John 1997; St. John et al. 1994, 1996), but the effects have varied depending on institutional type or student demographics. Chen and DesJardins (2010) found that the effects of financial aid on student withdrawal behavior varied across racial groups. They discovered that minorities have a higher risk of dropping out among non-grant recipients, but that their risk of dropping out is mitigated when they receive larger Pell grants. Herzog (2005) noted that retention from first year to second year was improved when a student received any type of financial aid package. In regards to transfer, financial restraints have been shown to be an insignificant predictor of a student’s propensity to transfer (Goldrick-Rab and Pfeffer 2009; Soares and Watson 2016).

Furthermore, family income has been shown to influence student departure. Ishitani (2003), in his analysis of the effects of factors on the departure of first-generation students, discovered that students from low-income families had an increased chance of departure during their first and second year of enrollment. Goldrick-Rab and Pfeffer (2009) interpreted the relationship between family income and transfer quite differently. They noted that:

Lateral transfer students appear to be a relatively elite set, since their levels of household income and parental occupational status are higher than average. Their motivations for changing colleges may be based on expressions of personal preference, possibly striving to move to a “better” school, but are clearly not connected to inadequate academic preparation in high school or poor performance in college (p. 115).

In summary, a number of factors have been shown to significantly influence student persistence and transfer at institutions of higher education. The factors that have been identified in the extant literature provide a jumping-off point for studies to develop models that cover relevant and necessary factors that contribute to students’ decisions to depart institutions, in general, and, specific to this study, transfer to another institution.

Data and Methodology

Study Data

This study used data drawn from the 2004–2009 Beginning Postsecondary Student (BPS: 04/09) dataset, which includes data collected longitudinally over a period of 6 years. The total sample size in the BPS: 04/09 was 16,680. A group of students who matriculated at either public or private 4-year institutions as first-time, beginning students was obtained from the BPS: 04/09 for the current study. First, missing values in continuous variables were deleted from the study data. Missing values in categorical variables were combined together in categories. This selection resulted in 7570 students who had intact information in the dataset.

Transfer was operationally defined as students who transferred from their initial 4-year institutions and never returned to their initial institutions during the survey observation period. Table 1 presents the number of transfers by academic year. Out of 7570 students in the study data, 1660 students were classified as transfer students. Out of 1660 transfer students in the data, over 49% of them transferred to other institutions during their first year, followed by 26% by the end of the second year. Not surprisingly, the number of transfers considerably decreased after the third year. Thus, students who transferred during their fourth year and years after were aggregated in one group. Moreover, there were 200 students who were still enrolled in their initial institutions at the end of the survey observation period. Since these students did not experience any type of departure but they were still enrolled in their institutions, they were right censored in the study data.

Table 1 Transfer counts over academic years
Full size table

Study Variables

The dependent variable in this study was whether students transferred or not in each academic year. The dependent variable was measured in a dichotomous manner. It is important to note again that transfer students in the study data never returned to their origin institutions before the end of survey observation period. Table 2 exhibits explanatory variables included in our study. Approximately 56% of the sample was female. About 72% of the students in the data were Caucasian, 9.2% Black, 8.6% Hispanic/Latino, 6.0% Asian and 4.6% multi-racial students or students with unknown ethnic backgrounds.

Table 2 Descriptive summary for study sample
Full size table

According to the Federal TRIO program definition, students whose parents did not complete a bachelor’s degree were defined as first-generation students. Approximately 40% of students in the sample fell into this category. Thirty-three percent of the students in the data had parents who both had at least bachelor’s degrees. Quartile variables were used to measure effects of family incomes. Similarly, the sum of SAT verbal and math scores were disaggregated in quartiles in this study. ACT composite scores were converted to an estimated SAT score for those who only reported ACT composite scores in the BPS: 04/09.

Grade point average (GPA), academic integration, and social integration were continuous variables in this study that were measured twice during the observation period (in 2004 and 2006). The average GPA in 2004 was 3.01 with a standard deviation of 0.71, whereas it was 3.16 with a standard deviation of 0.51 in 2006. GPA estimation in 2006 was based on students who were still enrolled in their original institutions in 2006.

Yearly amounts of Pell grants and loans (including PLUS loans) were included to assess effects of financial aid on student’s transfer-out behavior from 4-year institutions. While averages of Pell grants were rather consistent over time ranging from $2758 from $2853, averages of loans increased from $3460 to $5436 as students advanced in their college careers. Students’ academic integration was a derived variable in the BPS: 04/09 dataset based on average responses on how often students participated in the following: study groups, social contact with faculty, meeting with an academic advisor, or talking with faculty about academic matters outside of class. In a similar vein, the social integration consisted of participations in the following areas: fine arts activities, intramural varsity sports, and school clubs. Average scores for academic and social integration variables were multiplied by 100 in the BPS: 04/09 data by default. For 2004 academic and social integration, average scores were 90.4 and 66.5 with standard deviations of 41.3 and 53.3, respectively. The average score and standard deviation for 2006 academic integration was 101.7 and 43.2, while the same statistics for social integration were found 75.5 and 54.2, respectively.

Students were asked their academic disciplines twice: in 2004 and in 2006 as a part of the BPS: 04/09 data collection. Academic disciplines to which students belonged at matriculation were incorporated in this study as explanatory variables. The BPS: 04/09 included 12 academic discipline groups and an undeclared major category. Some individual disciplines such as life, physical sciences, and math or vocational and other technical majors were combined together due to a smaller sample size in each group. For students’ 2006 academic disciplines, this study included the variable indicating if their academic discipline was the same as 2004, instead of including 12 academic disciplines.

This study also added basic characteristics of institutions where students initially enrolled, such as institutional control and Carnegie Classification. About 59% students chose public 4-year institutions as their initial postsecondary institutions. As for Carnegie Classification, over 45% of the sampled students enrolled in Doctoral institutions, followed by 33% in Master’s institutions and 16% in Baccalaureate institutions.

Methodology

The uniqueness of the current study was to capture longitudinal transfer behaviors. Researchers may take a snap shot of transferred students at the end of the second or third year to study transfer behaviors. This practice leads to the inclusion of aggregated numbers of students who left institutions at different times. However, we believe that students with different characteristics may transfer to other institutions at different times for different reasons. In order to address the longitudinal effects of student characters on transfer behaviors, this study employed an event history modeling technique, also known as survival analysis.

Event history modeling, designed to examine longitudinal phenomena, has a number of advantages over other analytical procedures. Student departure is comprised of different types of departure behavior. Transfer is considered as one of such types. Event history modeling censors different types of departure, such as dropout, transfer, stop-out and graduation. Attempting to incorporate different types of departure over time in binary logistic regression is rather challenging. One first needs to create data only including transfer status, and then further development of data for each academic year is required to encompass a longitudinal framework. Others may suggest multinomial logistic regression, but interpretation of coefficients in this procedure is convoluted due to multiple reference groups. Furthermore, multinomial logistic regression also needs separate data for each academic year. These separate datasets are created by removing subjects that already experienced any departure in previous years. Given the inefficiency of employing logistic regression approaches, event history is a superior and more efficient analytical procedure than single or multinomial logistic regression to examine longitudinal data.

There are three methodological orientations in event history modeling; nonparametric (e.g., Kaplan–Meier estimation), semi-parametric known as Cox regression (including discrete-time hazard modeling), and parametric which includes a number of modeling techniques, such exponential, log-logistic, Weibull, etc. The selection of which parametric model should be used depends on the shape of time dependency over time.

The most widely used technique is Cox regression modeling, and some researchers in the area of educational research have employed discrete-time application of Cox regression (Chen and DesJardins 2008, 2010; DesJardins et al. 1999). Using Cox regression to analyze longitudinal retention data, an overall coefficient for each explanatory variable is estimated.

In this study, we employed an exponential modeling with period-specific effects as our modeling technique to unpack coefficients unique to each academic year. This particular modeling technique allows the effects of time-constant variables, such as gender or race to vary across academic years. In other words, being female is constant, while the effects of being female on transfer may differ from year to year. For instance, females may be less likely to transfer than males during the first year but not during the second year. Such possible varying effects by years would not be revealed when a researcher used aggregated transfer status data. Moreover, this particular modeling technique is more advantageous and efficient than a series of logistic regressions, because researchers do not need to develop separate data for each academic year to examine the time-varying effects of both time-constant and time-varying explanatory variables.

Prior to conducting the parametric analyses, data integrity was tested. Variable Inflation Factors were ranged from 1.025 to 2.211. Thus, there was no alarming issue regarding multicollinearity. Potential influential points among continuous variables, such as GPA, and academic and social integration, were examined using Cook’s and Mahalanobis distances. Findings of these two tests indicated there were no alarming outliers in the continuous variables included in this study.

Results

Table 3 presents event history modeling estimates for transfer students by academic year. Odds ratio, Δ%, and statistical significance were included in each year. Odds ratios greater than 1 indicated positive effects on transferring to other institutions, whereas odds ratios less than 1 were associated with reducing the likelihood of transfer. Δ% indicated the change in the probability of transfer. For instance, Δ% for female in the first year was − 0.277, translated as females being about 28% less likely than males to transfer in the first year.

Table 3 Event history modeling results
Full size table

This study discovered unique time-varying effects of certain variables on transfer behaviors. Females were found to be less likely than males to transfer in each academic year except for the third year, and the effect was most prominent later years in college. Females were about 47% less likely than males to leave their original 4-year institutions in the fourth year and after. In comparison to Caucasian students, Asian and Hispanic/Latino students were less likely to transfer. Asian students were about 38 and 56% less likely to transfer in their first and second years, while Hispanic/Latino students were 43 and 40% less likely to do so in the first 2 years in college.

Effects of family incomes were statistically significant mainly in later years in college in this study. An odds ratio associated with the low income quartile indicated that students in this group were twice more likely to transfer in the third year, followed by 46% more likely to do so in the first year. Students in the low-middle groups were 26 and 80% more likely transfer during the first and second years.

Negative effects of lower admission test scores extended over 3 years. Interestingly, the level of negative effect largely depended on which quartile a student belonged to and which academic year. The negative effect of the lowest quartile on transfer was most prominent in the third year, followed by the first year. The negative effect of the second quartile was strongest in the second year, followed by the first year. The second year was also found to be the highest risk period for students in the third quartile to transfer.

College GPAs were associated with reducing the likelihood of transfer in the first 2 years. Every one-point increase in the GPA was lowering the probability of transfer by 13 and 24% in the first and second years, respectively. Effects of social integration were statistically significant through all the academic years. The effect level of social integration was prominent in the third and fourth. For every 50-point increase in social integration scores, the likelihood of transferring to other institutions was reduced by approximately 11% in the first year and 10% in the second year. The effect of social integration was boosted for third and fourth year. Students were 45 and 36% less likely to transfer for every 50-point increase during the third and fourth years.

Increased amounts in Pell grant reduced the likelihood of transfer in the first and third years. Every $1.000 increase in Pell grant was associated with lowering the odds of transfer by 19% and 14% in the first and third years. Loan amounts were also found to have positive effects on reducing transfer rates over 3 years. Students were less likely to transfer in second, third and fourth years by 8, 12, and 43% for every $1000, respectively.

Study findings herein suggested that students in certain disciplines were less likely to transfer than those with undeclared majors. Students who were in Life, Physical Sciences and Math were about 31% less likely to transfer out in the first year, whereas students enrolled in Vocational/Technical programs were about 33% less likely to transfer in the same academic year. For the second year, the likelihood of transfer was reduced by about 55 and 46% when students were enrolled in Computer/Information science, while students enrolled in Health related programs were about 53% more likely to transfer. Students who changed majors were approximately 64% more likely to transfer in the third year.

Turning to institutional characteristics, students enrolled in Baccalaureate colleges and Special institutions were 30 and 48% less likely than students enrolled in Doctoral institutions to transfer in the second years. Students who enrolled in Master’s institution were about 31% less likely than students attending research institutions to transfer in their third year.

Limitations

There are several limitations pertaining to the current study to be discussed.

College GPA, academic, and social integration were surveyed in the first and third years. Although the levels of these variables were prone to vary from one year to another, the first-year college GPA and integration variables served to estimate their effects on transfer behavior for the first 2 years in college. Similarly, the third-year GPA, academic, and social integration were operationally employed as proxies to examine their effects on student transfer for the third year and beyond.

Generalizability of social integration was limited to involvements in fine arts activities, intramural varsity sports, and school clubs, whereas application of academic integration was restricted to social contact with faculty, meeting with an academic advisor, or talking with faculty about academic matters outside of class, and participation in study groups. Furthermore, it was unclear in the study data at which point in the first and third years students were asked to complete survey questionnaires gauging academic and social integration. It is reasonable to believe that students’ answers might differ whether they answered the survey at the beginning or at the end of academic years.

The inclusion of institutional admission selectivity was initially considered for this study; however, this variable was omitted due to its lack of consistency. This particular variable, which indicated admission selectivity in BPS: 04/09, was found inconsistent across students who attended the same institutions.

One may believe that investigation of types of transfer, such as reverse transfer is an imperative research topic. However, this inquiry is beyond the scope of this study.

At last, we suggest that readers interpret findings for the fourth year and years after with caution. Given that the number of transfers in the fourth year was small (n = 61), characteristics among these transfers might have been overestimated. Thus, we advise readers to treat findings associated with the fourth year and years after as noteworthy references rather than robust findings.

Discussion

Institutions may not be able to identify students who do not return to their campuses until registration is complete for the following semester. Transfer students are often masked by aggregated counts of dropouts. An institution is able to move a step further by tracking their leavers using the service offered by the National Student Clearinghouse. However, a national benchmark to illustrate behaviors of transferring students at four-year institutions longitudinally is absent.

Coupled with the BPS: 04/09, national longitudinal data, this study revealed how different student and institutional characteristics were associated with student transfer behavior over time at the national level.

Upper-level courses are known to be more costly than lower-level courses (Li, 2010). This may explain the reason that average loan amounts increased as students advanced in their college careers. However, this may also suggest that students’ institutional commitments are strengthened when students and their parents invest more money into their college education. Unlike students who withdraw from and never return to postsecondary institutions, students who persist and graduate view a college education as a form of upward social and economic mobility. Thus, they may not consider transferring to other institutions as cost-effective.

Moreover, this study also suggests that students who experienced academic misfit were prone to student transfer. Students who changed majors were ultimately more likely to transfer, evidence parallel to the findings of Li (2010). Li found that desired programs/courses offered at transfer destination institutions were the most frequents reason for students transferring out of 4-year institutions.

While study findings herein suggest that transfer-out behavior is mainly a function of admission test scores that students bring into college, a number of other explanatory variables confirmed their effects of reducing a chance of transfer. These explanatory variables include gender, race, college GPAs, and social integration. However, some of these variables are inherent and not able to be changed, such as gender and test scores. Instead of discussing how to increase the number of students with lower risk, institutional personnel may need to focus on factors applicable to their educational practice to reduce institutional transfer rates. For example, social integration unveiled its positive and longitudinal effect on lessening the number of transfers. Furthermore, the effect of social integration increased in later years in their native institutions. This finding demonstrated the importance of the social domain when it comes to discouraging student transfer.

Social integration unveiled its positive and longitudinal effect on lessening the likelihood of transfer and should guide the thoughts and plans of decision-makers as it relates to retaining students. Furthermore, the effect of social integration increased as students continued persisting in their native institutions, which was consistent with Tinto’s (1987) student integration model. In fact, social integration’s influence on student’s decision not to transfer becomes stronger over time. It is important that social integration be conceptualized as a fluid construct rather than a rigid and static idea. Program, activities, and policies can be crafted to identify and target students who will be most affected by specific interventions aimed at increasing their social integration at their institutions. Also, one may recall that social integration measures were based on basic social activities such as fine arts activities, intramural varsity sports, and school clubs. Thus, it is reasonable to believe that the impact of social integration is expected to further reduce the rate of transfer students if a college/university incorporates activities and interventions relevant to that institution at appropriate times during students’ educational career.

Simulation of Longitudinal Transfer Behavior

One of the challenges that institutional practitioners face is how research findings can guide them in their institutional planning. Researchers typically discuss an effect and the significance of each variable from their cross-sectional studies as we have discussed so far in this study. In reality, negative and positive coefficients counteracted with each other that would make it difficult to depict an overall risk of student with a certain set of characteristics.

Furthermore, our study findings on time-varying effects encompassed a number of academic years, which resulted in multiple coefficient values for the same variable. Such findings created complicated profiles of at-risk students. In order to mitigate the complexity of interpretation and to improve applicability of study findings to institutional practitioners and policy makers, we developed simulated scenarios of longitudinal student transfer risks.

For simplicity reasons, let us choose male Hispanic/Latino students as examples for simulated scenarios. Figure 1 includes longitudinal transfer risk trends for two male, Hispanic/Latino, students (Students 1 and 2) who had a first-quartile admission test score with an average GPA of 2.00 each year. Student 1 (solid line) did not have any social integration scores while Student 2 had 50 points in social integration for each year. As a result of minimum social integration scores, the probability of transfer for Student 1 was 30% in his third year, whereas the transfer probability for Student 2 (dotted line) was 19% in the same year. However, Student 2 still shared a similar pattern of transfer risks over time. In addition to increasing Pell Grant by $1000, Fig. 2 presents Student 2 increased the level of social integration by 50 points each year from the previous year (i.e., 50 points for first year, 100 points for second year, and so on). The third year transfer risk was noticeably lowered as a result of increasing the levels of integration. The original transfer risk of over 19% in the third year was reduced to less than 6%. These illustrations support the importance of social integration in college success, in particular as prevention strategies to reduce the number of transfer students. Also, graphical displays of longitudinal transfer behaviors serve as an excellent aid in policy discussion and developing intervention programs.

Fig. 1
figure 1

Transfer risk probabilities for Hispanic male student with 1st quartile admission test score (Student 1) without integration scores and for Hispanic male student with 1st quartile admission test score with integration scores (Student 2)

Full size image
Fig. 2
figure 2

Transfer risk probabilities for Hispanic male student with 1st quartile admission test score without integration scores (Student 1) and Hispanic male student with 1st quartile admission test score with increased integration scores and $1000 increases in pell grants in (Student 2). Note Transfer probabilities for the fourth year are extremely low due to a very small number of transfer students in the fourth year

Full size image

Summary

Given the significant effects of social integration, a lack of social support within the college environment partially triggers students to decide to transfer to other institutions throughout college years. Furthermore, conventional wisdom may suggest that students with higher academic achievement are inclined to transfer to other institutions for more academic challenges. However, the findings in this study failed to support this notion, particularly students who had higher GPAs were less likely to transfer during the first 2 years in college. Our study also found that students with lower admission test scores were more likely to transfer during the first 2 years in college. Moreover, Horn and Kojaku (2001) reported that about 40% of students who were initially enrolled in 4-year institutions transferred to less selective 4-year institutions, followed by about 39% who transferred to less-than-4 year institutions. This may suggest that students transferred due to poor academic performance at their original institutions. Future research examining types of transfer including reverse and horizontal transfers may shed light on a better understanding of transfer-out behaviors at 4-year institutions.