The Community College Penalty? Examining the Bachelor’s Completion Rates of Community College Transfer Students as a Function of Time

Abstract

Objective: Recent studies have shown that community college transfer students are just as likely to graduate with a bachelor’s degree as students who directly enroll in a 4-year institution. However, these studies do not typically examine whether there is a penalty for community college students in terms of the length of time it takes to complete a bachelor’s degree. In this study, we seek to determine whether there are differences between community college transfer students and direct 4-year college entrants regarding the likelihood of bachelor’s degree completion as a function of time. Method: Propensity score matching with a posttreatment adjustment was used to create observationally equivalent groups of community college transfer students and rising 4-year college juniors. Propensity scores were calculated using a multilevel model with students nested within high schools to account for pretreatment contextual differences. Descriptive survival analysis was applied to ascertain whether differences existed in the cumulative rate of bachelor’s degree completion throughout a 7-year tracking period. Results: Several significant prematch differences between the two groups—community college transfers and 4-year college juniors—were established. It was then demonstrated how the matching process created adequate balance between the two groups on all observed covariates. Although community college students experienced an initial penalty with respect to degree completion until 125% of normal time, the penalty was no longer evident 6 years after initial enrollment in college. Contributions: Given this finding, we discuss the need for policies that help community college transfer students achieve timely bachelor’s degree completion.

Keywords

community college transfer students community college penalty propensity score matching event history analysis survival analysis

Recent scholarship explores whether bachelor’s degree seeking students are penalized for initial community college enrollment (Dietrich & Lichtenberger, 2015; Glass & Harrington, 2002; Lee, Mackie-Lewis, & Marks, 1993; Melguizo & Dowd, 2009; Melguizo, Kienzel, & Alfonso, 2011; Monaghan & Attewell, 2015). Several studies have established that there is no “community college penalty,” which refers to the assumption that students who enroll at community colleges and transfer to a bachelor’s granting institution are less likely to complete a bachelor’s degree (Koranteng, 2015). In other words, there is no difference in the likelihood of bachelor’s completion between community college transfer students and students who directly enroll at 4-year colleges. The aforementioned studies, however, did not explicitly explore whether these vertical transfer students, or community college students who transferred to 4-year colleges, were penalized by having lower cumulative rates of bachelor’s degree completion over time; these studies focused on dichotomous measures of degree completion as the main dependent variable and utilized time horizons for tracking bachelor’s degree completion of various lengths. Theoretically, students who plan to complete a bachelor’s degree and begin at a community college would experience a reduction in overall tuition, given that community colleges charge lower tuition than 4-year colleges. However, any cost savings that would have occurred could be negated the longer students take to complete their bachelor’s degree, assuming they maintain at least part-time enrollment, given ongoing tuition charges.

Time to Degree and the Community College Penalty

There has been a dearth of research that has explicitly studied time to bachelor’s degree completion among all community college transfer students. The studies reviewed in this section have either explicitly examined time solely among bachelor’s degree completers or considered time implicitly through describing the tracking period—or time horizon—used when comparing bachelor’s completion rates between vertical transfer students and native 4-year students.

Research focusing on bachelor’s degree completers has shown a clear difference in median time to bachelor’s degree completion between community college transfer students and their counterparts that directly enrolled at 4-year colleges (Bound, Lovenheim, & Turner, 2012; National Center for Education Statistics [NCES], 2011). According to NCES (2011), the median time to a bachelor’s degree for students initially enrolling at a public 4-year institution was 55 months, compared with 63 months for students initially enrolling at a community college, and that difference could be described as the community college penalty.

Bound et al. (2012) observed time to bachelor’s degree completion was longest for community college transfer students in comparison with students directly enrolling at all types of 4-year institutions (based on both sector and selectivity). This held true with a cohort of high school graduates from both the 1970s and the 1990s. They also found that lower proportions of community college transfer students had completed their bachelor’s degree at all points throughout the 8-year time horizon of the study: 4, 5, 6, and 7 years after entry. These differences also suggest a community college penalty. While Bound et al. (2012) controlled for observed, preexisting differences among community college students and students who directly enroll in 4-year institutions, such as academic preparation, these researchers did not conduct a quasi-experimental study using propensity score matching (PSM) methods. A quasi-experiment using PSM would have enabled a more direct comparison of community college transfer students and students who initially enrolled in 4-year institutions.

Finally, the time-based outcomes measured in the studies by Bound et al. (2012) as well as NCES (2011) were conditional upon degree completion, so the studies did not take into consideration potential differences in the likelihood of degree completion between the community college group and direct 4-year college entrants. This disadvantage is overcome using event history or survival analysis in which cumulative completion rates for all vertical transfers and native juniors are compared at all points throughout the study.

In contrast to the explicit study of time in Bound et al. (2012) and NCES (2011), other researchers have examined time implicitly, through the acknowledgment of tracking periods used in studies that examine bachelor’s degree completion rates. Taken together, these studies have shown that there is considerable variation in the time horizon for tracking potential differences in bachelor’s degree completion rates. For example, Dietrich and Lichtenberger (2015) looked for potential differences at the end of a 7-year tracking period (175% of normal time). Melguizo and Dowd (2009) and Melguizo et al. (2011) examined potential difference in bachelor’s degree completion for 8.5 years after college entry, which equates to 212.5% of normal time. There have been two related studies that have used a time horizon of only 6 years and have found no significant penalty for community college transfer students (Lee et al., 1993; Monaghan & Attewell, 2015). The findings by Lee et al. (1993) and Monaghan and Attewell (2015) were in contrast to Melguizo and Dowd (2009) who argued that using traditional measures of bachelor’s degree completion, namely within 150% of normal time or 6 years, tends to exhibit a penalty for community college students because that time horizon is not long enough to adequately track bachelor’s degree completion for that group.

Lee et al. (1993) found that community college transfer students and direct entrants to 4-year colleges were just as likely to have completed a bachelor’s degree within 6 years of graduating high school, or 150% of normal time, particularly after controlling for student background and institutional characteristics. Monaghan and Attewell (2015) tracked bachelor’s degree completion for six academic years and found no statistically significant difference between community college students who transferred and a similar group of direct 4-year entrants. The sample used by Monaghan and Attewell (2015) was specific to less selective institutions, which current research has reported are the most likely destinations for community college transfer students (Brand, Pfeffer, & Goldrick-Rab, 2014; Lichtenberger & Dietrich, 2013). While their focus on less selective institutions may be applicable for many community college transfer students, these results may not apply to community college students who attend more selective institutions. Further, all of the studies focusing on community college transfer students and observationally equivalent direct entrants compare bachelor’s completion rates only at the end of the study (Dietrich & Lichtenberger, 2015; Glass & Harrington, 2002; Lee et al., 1993; Melguizo & Dowd, 2009; Melguizo et al., 2011; Monaghan & Attewell, 2015).

While these studies have implied that completion rates may differ based on the time horizon, we argue that the likelihood of completion must be directly considered as a function of time. Furthermore, unlike previous studies that studied time explicitly with bachelor’s degree completers only, we propose to examine all community college transfer students and a similarly comprised comparison group of students who directly enrolled in a 4-year institution. Constructing observationally equivalent groups of community college students and students who directly enroll in 4-year institutions would allow for a direct comparison of cumulative bachelor’s degree completion rates between these two groups. Moreover, studying time directly using survival analysis will enable an examination of whether students initially enrolling in community college suffer a penalty in bachelor’s degree completion, when the penalty exists, and for how long. The need to account for time has also been suggested in the work of DesJardins, Ahlburg, and McCall (2002), Goldrick-Rab (2006), and Goldrick-Rab and Pfeffer (2009). Furthermore, the timing of bachelor’s degree completion has moved to the forefront in current policy, primarily due to its direct relationship with reducing cost, both real and opportunity, and decreasing debt burden (Johnson, 2011; Weldon, 2013).

In this quasi-experimental study, we used PSM along with a posttreatment adjustment to isolate the effect associated with what we considered the treatment—taking the community college to 4-year institution transfer pathway. We then used descriptive survival analysis, also known as event history analysis, to examine potential time-based differences between community college transfer students and observationally equivalent rising 4-year college juniors in bachelor’s degree completion. This analysis utilized a 7-year tracking period, which is up to 5 years posttransfer for the community college group.

Methodology

The major methodological components of this study included a PSM procedure with posttreatment adjustment, detailed later in this section, as well as descriptive survival analysis that used bachelor’s degree completion as the event of interest. PSM was selected because this method helps to create observationally equivalent treatment and comparison groups (Rosenbaum & Rubin, 1985). As a form of survival analysis, descriptive and inferential statistics were used to pinpoint exactly when the community college penalty began and when it was potentially eliminated.

Data

Our treatment group, the group of community college transfer students, originated from the Illinois high school graduating class of 2003. The initial study sample included 2,154 community college student transfers and 21,426 rising juniors. The final matched sample consisted of 2,117 community college transfers and 2,117 rising juniors. The community college transfer group comprised all individuals from the cohort who enrolled full-time at a community college the fall semester of 2003 and maintained full-time enrollment (fall and spring semesters) for 2 years prior to transferring to a 4-year institution in the fall semester of 2005. We opted to focus on full-time enrollees as that was the enrollment pattern of nearly all of the direct entrants to 4-year colleges. Full-time enrollment was identified using the enrollment status variable within the National Student Clearinghouse (NSC) student tracker and defined as 12 or more semester hours. Unfortunately, we did not have access to total credit hour accumulation.

Further, our counterfactual was to determine what would have happened to this group of community college transfer students if they had instead directly enrolled at a 4-year college. The larger group of potential comparison group members, rising 4-year college juniors, included direct entrants to 4-year colleges who maintained a similar pattern of enrollment at their respective 4-year institution and were still enrolled the fall semester of 2005. Members of the treatment group and comparison group were not allowed to vertically transfer during the first 2 years of the tracking period, but may have enrolled in summer school.

The data used in the study were made available to the researchers through data-sharing agreements with ACT and the Illinois Board of Higher Education. The main source of student-level background and academic performance information was the ACT, which at the time was administered to all high school juniors in the state of Illinois. College enrollment and degree completion was tracked using information obtained from the NSC. The information from ACT and the NSC was supplemented with high school information from the Illinois Interactive Report Card and the college-specific information from the Integrated Postsecondary Education Data System (IPEDS) and Barron’s.

PSM Methodology

A key step in PSM is the calculation of the propensity score, which is the probability of being in the treatment group, given a set of background characteristics (Oakes, 2010). Rosenbaum and Rubin (1983) gave the formal definition of a propensity score as the equation, p(X) = Pr(Z = 1 | X). In this study, Z is the binary variable that indicates whether a student is assigned to the treatment group (Z = 1 = “community college transfer”) versus the control group (Z = 0 = “native junior”) and X is a vector of background covariates. We discuss how we accounted for the nested nature of these data later in this methodology section.

To identify covariates used to predict the likelihood of community college transfer, we utilized the following general categories from a conceptual model proposed by Wang (2009): demographic background, academic performance, academic resources, psychological resources, and environmental factors at the high school and student level. To these general categories, we added college preference variables to help better establish the probability of being a community college transfer student. While the model by Wang (2009) suggests variables pertinent to postsecondary outcomes in general, such as persistence and baccalaureate attainment, additional variables are needed to help predict initial college enrollment, in this case, the likelihood of community college enrollment. College preference variables were added to variables identified by Wang (2009) because they are associated with the type of school that a student initially enrolls. Each of these categories is discussed in the following section and shown in Figure 1.

Figure 1.

Conceptual model used to create propensity scores.

Demographics

To account for demographic differences between the community college transfer students and the comparison group of rising 4-year college juniors, we included the following covariates in the model used to calculate propensity scores: gender, ethnicity/race, and family income. The research findings for gender and transfer from a 2-year to 4-year institutions have indicated conflicting results (Dietrich & Lichtenberger, 2015). For example, the rate of transfer was fairly similar for male and female community college entrants among Illinois high school graduates (Smalley, Lichtenberger, & Brown, 2010). The finding by Smalley et al. (2010) was similar to Dougherty and Kienzl (2006) who also found no statistically significant differences by gender in vertical transfer. However, the earlier studies by Surette (2001) and Bailey, Jenkins, and Leinbach (2005) reported that females were less likely to vertically transfer to a 4-year institution than males.

In summarizing research on transfer rates from community college to 4-year institutions, Dietrich and Lichtenberger (2015) observed that race was also a key determinant in two of three studies on community college transfer. African Americans were significantly less likely to earn an associate degree or transfer to a 4-year institution than their White peers (Bailey et al., 2005). Similarly, Smalley et al. (2010) found that Asian and White community college students transferred to 4-year colleges at higher rates than either African American or Latino/a students. However, Dougherty and Kienzl (2006) reported that race was not a statistically significant determinant of vertical transfer. The discrepancy between these studies may be partially explained by the differences in the cohorts. Unlike Smalley et al. (2010), and Bailey et al. (2005), who studied millennial samples, Dougherty and Kienzl (2006) focused on Generation X students. Family income has been a salient predictor of vertical transfer to a 4-year institution. Prior research has consistently concluded that higher family income levels correlated with increased transfer rates to 4-year programs among community college students (Dougherty & Kienzl, 2006; Goldrick-Rab, 2010; Goldrick-Rab & Pfeffer, 2009).

Academic Performance

We included the following academic performance variables in the model used to calculate the propensity scores: high school grade point average (GPA) and scores on the four subject matter tests that comprise the ACT (English, mathematics, science, and reading). Academic performance variables such as high school GPA and standardized test scores have been shown to affect postsecondary enrollment (Cohen & Brawer, 2008; Cross, 1971; Smalley et al., 2010). As mentioned by Dietrich and Lichtenberger (2015), literature has established that students initially enrolling in community colleges are less ready for college than their peers who initially enroll in bachelor’s degree programs (Cohen & Brawer, 2008; Cross, 1971; Smalley et al., 2010). College readiness, in turn, may later affect college transfer rates.

Academic Resources

Wang (2009) explained that academic resources refer to the intensity of the high school curriculum. To assess academic resources in our model, we considered program type, which was measured by a categorical variable in which students report whether they are enrolled in a college preparatory program, a career and technical education (CTE) program, or enrolled in the general curriculum in addition to dual-credit status, measured as the number of semesters in which a student enrolled in dual-credit courses while in high school.

With regard to enrolling in CTE courses in high school and later college enrollment, research has shown conflicting results (Dietrich & Lichtenberger, 2015). Gemici (2011) observed that students who complete a CTE track earn a regular high school diploma at a significantly higher rate than students who complete a general high school curriculum. High school completion does not necessarily translate to postsecondary enrollment for CTE students. Some researchers have reported that increased CTE participation results in lower postsecondary enrollment rates (Levesque et al., 2008) while others have found that CTE students who participate in dual enrollment programs were more likely to enroll in either community colleges or 4-year institutions (Karp & Hughes, 2008; Reese, 2008). Thus, there is evidence that supports that CTE program participation may indirectly affect the probability of being a community college transfer student.

Dual-credit participation has been revealed to positively affect postsecondary outcomes. In general, students enrolled in community college sponsored dual-credit programs are more likely to enroll in postsecondary programs (Karp, Calcagno, Hughes, Jeong, & Bailey, 2007; Lichtenberger, Witt, Blankenberger, & Franklin, 2014). We argue dual-credit status could have an impact on enrollment patterns, such as taking the community college transfer pathway as opposed to directly enrolling at a 4-year college.

Psychological Attributes: Degree Aspirations

Studies focusing on various aspects of higher education have demonstrated the importance of accounting for student degree aspirations as influential to postsecondary achievement (Beal & Crockett, 2010; Boxer, Goldstein, DeLorenzo, Savoy, & Mercado, 2011; Eagan et al., 2013; Ou & Reynolds, 2008; Wang, 2013). Conceptualizing college transfer as one kind of postsecondary goal, we argue that student degree aspirations are associated with the likelihood of taking the community college transfer pathway. We hypothesized that students whose degree aspirations were technical in nature or stopped upon associate degree completion were perhaps more likely to directly enroll at a community college after high school graduation as opposed to a bachelor’s granting institution.

Environmental Factors

Wang (2009) as well as earlier research from Bean (1990) and Bean and Metzner (1985) noted that environmental factors may play a role in postsecondary outcomes, such as the likelihood of transferring from community college to a 4-year institution, relative to directly entering a 4-year college. In this study, we examined environmental factors at two levels: high-school-level context and student-level context.

High school context

Sociological studies have established the need to account for high school context (Alwin & Otto, 1977; Nelson, 1972; Roderick, Coca, & Nagaoka, 2011; Rowan-Kenyon, Perna, & Swan, 2011; Smalley et al., 2010). To account for such contextual effects, we included the racial composition of the given school, the percentage of low-income students, the percentage of students with limited English proficiency, and the school’s aggregate performance on the ACT.

Student-level context

At the student level we considered expectation to work, financial aid variables, and student living arrangements. Dadgar (2012) reported that holding a job was associated with a lower college GPA. In turn, lower grades may affect the ability to transfer vertically to a 4-year institution. Research has also shown that financial aid affects college enrollment outcomes (Avery & Hoxby, 2004; Bettinger, 2004; DesJardins, Ahlburg, & McCall, 2006). To account for the effects of financial aid, the following variables were included in the model used to create the propensity scores: financial aid expectation, tuition preference, and the importance of cost in selecting a college—all three factors are arguably related to one’s cost sensitivity. de Araujo and Murray (2010) reasoned that student academic performance improves when students live on campus. Moreover, community college students are likely to live off campus or at home as few community colleges offer on-campus housing (College Board, n.d.; Sheehy, 2015). As such, housing preference is a logical predictor of being a community college transfer student, as opposed to a direct 4-year college entrant. It should be noted that in the current study we used preferred living arrangement, a pretreatment variable, and not their actual living arrangement.

College Preference Variables

We considered three variables related to college choice or preference prior to high school graduation: location, distance, and community college preference. Location is one variable that affects the school in which one initially enrolls (Baryla & Dotterweich, 2001; Fletcher, 2012; Kolesnikova, 2010; Mattern, Wyatt, & Shaw, 2013; Tuckman, 1970). In a study of Texas community college students, campus location was one of the top reasons for selecting a particular community college (Barreno & Traut, 2012). To assess location, we examined whether students preferred to go to an in-state or out-of-state school. Mattern et al. (2013) reported that distance between a student’s home and the college of initial enrollment was positively related to transfer. In other words, the farther the college of initial enrollment to the student’s home, the more likely they were to transfer. They also stated that students who transfer were most likely to transfer to an institution that was closer to their home. Finally, students who prefer to attend community college are more likely to enroll at a community college; it is important to account for this factor when predicting the likelihood of being a community college transfer student.

Propensity Score Calculation With Nested Data: Complete Pooling Versus Partial Pooling

In educational research, calculation of propensity scores is often conducted using binary logistic regression models containing student-level variables exclusively. Griswold, Localio, and Mulrow (2010) referred to propensity score modeling that does not account for hierarchical or nested data structures as complete pooling. With complete pooling, treatment group members can be matched to any comparison group member, regardless of the potentially hierarchical structure of the data, such as students nested within schools. Propensity scores in hierarchical data structures have not been studied extensively (Thoemmes & West, 2011); yet, complete pooling of study group members in the development of the model does not fully take into consideration the wide variation across school contexts. In many instances, the nature of the data used in PSM studies is nested, particularly in educational research where students are often nested within schools (Arpino & Mealli, 2011).

To help account for high school context, we used what is referred to as partial pooling by using a hierarchical generalized linear model (HGLM) with students nested within schools. In partial pooling, students compared across treatment groups do not have to belong to exactly the same site or cluster. Students, rather, only need to belong to a similar cluster based on characteristics measured at the level of the cluster (Griswold et al., 2010). Our clustering procedures are addressed below. The HGLM was conducted by SPSS 22 to calculate the propensity scores. The following is the HGLM model for the probability of being a community college transfer student.

The equation used at Level 1 to predict the log odds of a student i in school j can be written as follows:

\begin{array}{l} η_{i j} = \log (π_{i j} / π_{C i j}) = β_{0 j} + β_{1} e t h n i c i t y_{i j} + β_{2} r a c e_{i j} + β_{3} g e n d e r_{i j} + β_{4} f a m i l y i n c o m e_{i j} + \\ β_{5} G P A_{i j} + β_{6} A C T_{i j} + β_{7} h s p r o g t y p e_{i j} + β_{8} d u a l c r e d i t_{i j} + β_{9} d e g r e e a s p i r a t i o n s_{i j} + \\ β_{10} e x p w o r k_{i j} + β_{11} e x p a i d_{i j} + β_{12} t u i t i o n p r e f e r e n c e_{i j} + β_{13} c o s t i m p o r t a n c e_{i j} + \\ β_{14} l i v i n g a r r a n g e m e n t_{i j} + β_{15} c o l l e g e l o c a t i o n_{i j} + β_{16} d i s t a n c e f r o m h o m e_{i j} + \\ β_{17} c o l l e g e p r e f e r e n c e_{i j} \end{array}

In this equation, $η_{i j}$ refers to the log odds of being a community college transfer student transfer versus the reference category, being a 4-year college entrant who persisted to the junior level.

The Level 2 intercept model showing variables at the high school level is as follows:

\begin{array}{l} β_{0 j} = γ_{00} + γ_{01} s c h o o l r a c e + γ_{02} p e r c e n t l o w i n c o m e_{j} + \\ γ_{03} p e r c e n t L E P_{j} + γ_{04} s c h o o l A C T_{j} + u_{0 j} . \end{array}

Combining the Level 1 and Level 2 intercept model results in the following:

\begin{array}{l} η_{i j} = γ_{00} + γ_{01} s c h o o l r a c e_{j} + γ_{02} p e r c e n t l o w i n c o m e_{j} + γ_{03} p e r c e n t L E P_{j} + \\ γ_{04} s c h o o l A C T_{j} + γ_{10} e t h n i c i t y_{i j} + γ_{20} r a c e_{i j} + γ_{30} g e n d e r_{i j} + γ_{40} f a m i l y i n c o m e_{i j} + \\ γ_{50} G P A_{i j} + γ_{60} A C T_{i j} + γ_{70 hs} p r o g t y p e_{j} + γ_{80} d u a l c r e d i t_{i j} + γ_{90} d e g r e e a s p i r a t i o n s_{i j} + \\ γ_{100} e x p w o r k_{i j} + γ_{110} e x p a i d_{i j} + γ_{120} t u i t i o n p r e f e r e n c e_{i j} + γ_{130} c o s t i m p o r t a n c e_{i j} + \\ γ_{140} l i v i n g a r r a n g e m e n t s_{i j} γ_{150} c o l l e g e l o c a t i o n_{i j} + γ_{160} d i s t a n c e f r o m h o m e_{i j} + \\ γ_{170} c o l l e g e p r e f e r e n c e_{i j} + u_{0 j} \end{array}

This combined model predicts the log odds of being a community college transfer student versus the reference category ( $η_{i j}$ ).

Posttreatment Adjustment

We argue that institutional selectivity may affect subsequent bachelor’s degree completion. We required the selectivity of the receiving 4-year institution for each community college transfer student to be the same as that of the comparison group member to whom they were matched. There has been support in the literature for accounting for the effects of selectivity, such as Barron’s (2003), when examining bachelor degree completion among direct entrants to 4-year colleges (Lichtenberger & Dietrich, 2012) and among community college transfer students (Melguizo & Dowd, 2009; Monaghan & Attewell, 2015).

Another way to conceive of selectivity is as a type of posttreatment adjustment. Posttreatment adjustments are needed in quasi-experimental studies because of changes in study populations that occur over time (Lemons, Fuchs, Gilbert, & Fuchs, 2014). In the case of the current study, we argue that the effects of the treatment, or taking the community college transfer pathway, might be altered by an increased likelihood of enrolling at competitive and less competitive 4-year institutions, as opposed to more competitive colleges (as based on Barron’s). Enrolling at relatively less selective institutions, in turn, could serve to lower students’ likelihood of bachelor’s degree completion. However, as the transfer destination is determined after the treatment, including selectivity in the PSM model is not appropriate. The adjustment must be made posttreatment, in this case, after taking the community college transfer pathway. We used the Barron’s (2003) college selectivity categories because that was the information available to these students as they transitioned from high school to college. Using posttreatment adjustments has been theoretically supported by Flores and Flores-Lagunes (2009) and Frangakis and Rubin (2002).

PSM

We used the propensity scores generated from the multilevel model described above, with students nested within schools, to find the nearest neighbor. We also matched with replacement, allowing comparison group members to be matched to more than one treatment group member, in an effort to address some of the drawbacks associated with removing comparison group members after they are first matched to a treatment group member (Dehejia & Wahba, 2002; Rosenbaum, 1995). To ensure that the matches maintained a similar profile, we used a caliper of .25 SD based on the propensity scores, as suggested by Rosenbaum and Rubin (1985).

Finally, to keep as many cases as possible, particularly with the requirement of matching on institutional selectivity, we used a dummy variable adjustment or missing data indicator method for all cases with missing data on one of the control variables (Cohen & Cohen, 1975, 1985; Stuart, 2010). While this technique has been revealed to be invalid for handling missing data in regression studies (see Allison, 2009; Greenland & Finkle, 1995), it has been recommended as an appropriate and effective method in the context of propensity scores (Stuart, 2010).

While we used HGLM in SPSS to calculate the propensity scores, relational database development software called Filemaker Advanced was employed to complete the match. Relational databases store specifically defined tables from which data sets can be accessed and linked in multiple ways without altering the original data sets. Using Filemaker Advanced, we created a table with the treatment group members and a separate table for comparison group members and created a relationship between the two. A row in the treatment database table was potentially associated with one or more rows in the comparison table, which reflected our matching process. We used this relationship to identify the nearest neighbor within the aforementioned caliper (.25 SD) that was also enrolled at a 4-year college with the same level of selectivity as the receiving institution for the given community college transfer. Once the matched pairs were identified, a flat file was created with information on each of the matched treatment group members and their respective match from the comparison group. In the end, this flat file was used in all of the postmatch analyses.

Results

Prematch Comparison

As displayed in Tables 1 through 3, there were several significantly large standardized differences between the community college transfer students and the larger group of potential comparison group members prior to our matching procedures. Standardized differences greater than .10 are indicative of imbalance and any difference greater than .20 is considered significantly large (Cohen, 1977; Normand et al., 2001; Rosenbaum & Rubin, 1985).

Table 1.

Demographic Background, Academic Performance, Academic Resources, and Psychological Attributes.

	Unmatched sample			Matched sample with posttreatment adjustment
	Community college transfer	4-year rising junior		Community college transfer	4-year rising junior
Baseline covariates	“Treated” (n = 2,154)	“Untreated” (n = 21,426)	SD	“Treated” (n = 2,117)	“Untreated” (n = 2,117)	SD
Precollege characteristics
Demographic background
Race: White	81%	71%	0.26	80%	79%	0.03
Race: Latino	3%	5%	−0.10	3%	3%	0.00
Race: Asian	3%	7%	−0.18	3%	2%	0.06
Race: African American	3%	8%	−0.22	3%	3%	0.00
Race: Other	3%	3%	0.00	3%	3%	0.00
Race: Missing	8%	8%	0.00	9%	9%	0.00
Gender: Male	47%	43%	0.08	47%	46%	0.02
Income: High $80k+	15%	29%	−0.32	15%	12%	0.09
Income: Mid-high $50k-<$80k	23%	19%	0.10	23%	21%	0.05
Income: Mid-low $30k-<$50k	20%	14%	0.16	20%	24%	−0.10
Income: Low <$30k	11%	9%	0.07	11%	12%	−0.03
Income: Missing	30%	30%	0.00	31%	32%	−0.02
Academic performance
High School GPA: 3.5+	26%	45%	−0.37	26%	24%	−0.05
High School GPA: 3.0-3.4	26%	21%	0.12	26%	24%	−0.05
High School GPA: 2.5-2.9	14%	8%	0.20	14%	17%	0.08
High School GPA: <2.5	8%	3%	0.22	8%	9%	0.04
High School GPA: Missing	25%	23%	0.05	25%	27%	0.05
ACT math	21.1	24.9	−0.80	21.1	20.7	0.09
ACT English	20.7	24.6	−0.80	20.7	20.2	0.11
ACT reading	21.3	24.7	−0.63	21.4	21.0	0.08
ACT science	21.2	24.0	−0.67	21.3	20.8	0.13
Academic resources
High School Program: College prep	42%	57%	−0.28	42%	41%	−0.02
High School	10%	5%	0.19	10%	9%	−0.03
Program: CTE
High School	22%	15%	0.18	22%	23%	0.02
Program: General
High School	26%	24%	0.05	26%	28%	0.05
Program: Missing
Sems. dual-credit	0.48	0.29	0.24	0.46	0.45	0.01
community college
Sems. dual-credit 4-year	<0.01	0.02	−0.13	<0.00	<0.00	0.00
Psychological attributes
Degree aspirations: <2 year degree	1%	<1%	0.12	1%	1%	0.00
Degree aspirations: Associate	3%	1%	0.14	3%	3%	0.00
Degree aspirations: Bachelor’s	36%	23%	0.30	36%	35%	0.02
Degree aspirations: Master’s	17%	25%	−0.19	17%	16%	0.03
Degree aspirations: Professional	19%	29%	−0.23	19%	18%	0.03
Degree aspirations: Other	3%	2%	0.06	3%	3%	0.00
Degree aspirations: Missing	23%	21%	0.05	23%	25%	−0.05

Abbreviations: GPA = grade point average; CTE = career and technical education program.

Note. Due to rounding, totals do not always add up to 100%.

Table 2.

Environmental Factors.

	Unmatched sample			Matched sample with posttreatment adjustment
	Community college transfer	4-year rising junior		Community college transfer	4-year rising junior
Baseline covariates	“Treated” (n = 2,154)	“Untreated” (n = 21,426)	SD	“Treated” (n = 2,117)	“Untreated” (n = 2,117)	SD
Environmental factors
High school context
Percent White	85%	74%	0.51	85%	85%	0.00
Percent limited English proficiency	1%	2%	−0.18	2%	1%	0.04
Percent low income	16%	16%	0.00	16%	17%	−0.06
ACT average	20.4	21.2	−0.41	20.4	20.3	0.06
Student-level factors
Expected work hours: None	8%	19%	−0.31	8%	7%	0.04
Expected work hours: 1-10	19%	24%	−0.12	20%	17%	0.08
Expected work hours: 11-20	35%	27%	0.18	35%	37%	−0.04
Expected work hours: 21-30	13%	6%	0.24	12%	11%	0.03
Expected work hours: 31+	2%	1%	0.08	2%	3%	−0.06
Expected work hours: Missing	23%	23%	0.00	23%	25%	0.05
Aid expectation: Yes	63%	64%	−0.02	63%	63%	0.00
Aid expectation: No	14%	15%	−0.03	14%	13%	0.03
Aid expectation: Missing	22%	21%	0.02	23%	24%	−0.02
Tuition preference: <$1,000	1%	<1%	0.12	1%	1%	0.00
Tuition preference: $1,000-$1,999	2%	1%	0.08	2%	3%	−0.06
Tuition preference: $2,000-$2,999	4%	1%	0.20	4%	6%	−0.09
Tuition preference: $3,000-$3,999	4%	1%	0.20	4%	4%	0.00
Tuition preference: $4,000-$4,999	4%	2%	0.12	4%	4%	0.00
Tuition preference: $5,000-$7,499	8%	5%	0.12	8%	10%	−0.07
Tuition preference: $7,500-$9,999	6%	6%	0.00	6%	6%	0.00
Tuition preference: $10,000+	5%	7%	−0.08	5%	5%	0.00
Tuition preference: No preference	40%	53%	−0.25	40%	36%	0.08
Tuition preference: Missing	25%	24%	0.02	25%	28%	−0.07
Cost most important: Yes	21%	13%	0.22	20%	21%	−0.03
Cost most important: No	54%	63%	−0.17	54%	51%	0.06
Cost most important: Missing	25%	24%	0.02	26%	28%	−0.05
Living: Residence hall	35%	65%	−0.53	36%	32%	0.09
Living: Off-campus	17%	6%	0.36	17%	16%	0.03
Living: Parent’s house	23%	6%	0.51	22%	24%	−0.05
Living: Married housing	1%	<1%	0.12	1%	<1%	0.06
Living: Fraternity or sorority	3%	4%	−0.05	3%	2%	0.06
Living: Missing	22%	20%	0.05	23%	25%	−0.05

Note. Due to rounding, totals do not always add up to 100%.

Table 3.

College Preference.

	Unmatched sample			Matched sample with posttreatment adjustment
	Community college transfer	4-year rising junior		Community college transfer	4-year rising Junior
Baseline covariates	“Treated” (n = 2,154)	“Untreated” (n = 21,426)	SD	“Treated” (n = 2,117)	“Untreated” (n = 2,117)	SD
College preference
Location: In-state	64%	54%	0.22	62%	63%	−0.02
Location: Out-of-state	11%	22%	−0.29	14%	11%	0.09
Location: Missing	24%	24%	0.00	24%	27%	−0.07
Preferred distance: <10 miles	8%	3%	0.22	7%	8%	−0.04
Preferred distance: 10-25 miles	13%	5%	0.29	13%	15%	−0.06
Preferred distance: 26-100 miles	20%	15%	0.13	20%	19%	0.03
Preferred distance: 100+ miles	14%	32%	−0.41	14%	12%	0.06
Preferred distance: No college in mind	22%	25%	−0.07	23%	20%	0.07
Preferred distance: Missing	23%	22%	0.02	23%	26%	−0.07
Type: 4 year	61%	76%	−0.29	62%	59%	0.06
Type: 2 year	15%	1%	0.54	14%	15%	−0.03
Type: Missing	24%	23%	0.02	24%	26%	−0.05

Note. Due to rounding, totals do not always add up to 100%.

Regarding precollege demographic background characteristics, there were two notably large differences between the prematched group of community college transfer students and the prematched group of rising 4-year college juniors. Significantly higher proportions of the community college group were white and consequently the prematched group of rising 4-year college juniors was more racially/ethnically diverse. There were also large differences based on family income demonstrating that community college transfer students were less likely to emanate from high-income families.

As exhibited in Table 1, standardized differences specific to the academic performance baseline covariates were among the largest and suggested that the prematched sample of community college transfer students would have a lower likelihood of bachelor’s degree completion. We found substantially fewer of the prematched community college transfer students, in relative terms, fell into the highest high school GPA category. In terms of performance on the ACT, the prematched community college group as a whole had significantly lower scores on the individual subject tests that comprise the ACT.

Regarding academic resources, proportionally fewer of the community college transfer students were in a college prep program during high school, and consequently, substantially more were enrolled in a CTE or general curriculum program. However, the prematched community college transfer students had a higher mean number of semesters in which they were enrolled in community college–based dual-credit during high school relative to their peers directly enrolling at 4-year colleges. One explanation for this counterintuitive finding is that community college–based dual-credit increases one’s likelihood of enrollment at the community college providing dual-credit, as opposed to a 4-year college. Also, as previously mentioned, more of the treatment group members were enrolled in a high school CTE program and we could not make the distinction between traditional transfer dual-credit and CTE dual-credit. It very well could be that their higher participation rates in CTE resulted in higher participation rates in community college–based dual-credit.

As featured in Table 2, there were also major differences between the prematched groups in terms of high school context. The community college group, as a whole, tended to emanate from high schools with significantly higher proportions of White students and their high schools, on average, had significantly lower aggregate ACT scores. In terms of student-level contextual factors, the prematched sample of community college transfer students had significantly higher proportions expecting to work during college, as well as significantly higher proportions who were arguably more sensitive to cost, based on tuition preference and the importance of cost in selecting a college. Significantly higher proportions of community college transfer students expected to live with their parents or off-campus their first year in college, as opposed to living on campus in a dormitory, relative to prematched rising 4-year college juniors.

In terms of the factors related to college preference (see Table 3), proportionally fewer community college transfer students had originally stated that a 4-year college was their preferred choice, and mentioned a community college as their preferred destination. As a whole, the prematched sample of community college transfer students was significantly more homebound based on their plans in high school. Significantly higher proportions specified in-state colleges as their preferred destination, as well as colleges that have a shorter geographic distance from home.

There were significantly large standardized differences between the prematched samples regarding college selectivity, as highlighted in Table 4. This was a solid justification for our posttreatment adjustment. Community college transfers were significantly more likely to enroll at competitive 4-year colleges; whereas their rising junior counterparts were significantly more likely to have enrolled at more competitive institutions. In fact, two thirds of the prematched sample of community college transfers enrolled at a competitive 4-year college, relative to 39% of the prematched 4-year college juniors.

Table 4.

Barron’s Selectivity and Propensity Scores.

	Unmatched sample			Matched sample with posttreatment adjustment
	Community college transfer	4-year rising junior		Community college transfer	4-year rising junior
Baseline covariates	“Treated” (n = 2,154)	“Untreated” (n = 21,426)	SD	“Treated” (n = 2,117)	“Untreated” (n = 2,117)	SD
Most/highly competitive	8%	28%	−0.50	8%	8%	0.00
Very competitive	11%	25%	−0.35	11%	11%	0.00
Competitive	66%	39%	0.64	66%	66%	0.00
Less/not competitive	12%	6%	0.21	12%	12%	0.00
Other/not defined	3%	1%	0.14	3%	3%	0.00
Propensity scores	0.36	0.06	1.52	0.35	0.35	0.00

Note. Due to rounding, totals do not always add up to 100%.

Balancing Diagnostics and Differences After the Match

Using our matching criteria, we were able to match over 98% of the prematch sample of community college transfer students. As mentioned previously, the final matched sample included 2,117 community college transfer students and an equivalent number of rising juniors. To determine if adequate balance between the groups was attained, we compared the mean propensity scores both before and after the match. As shown in Table 4, prior to the match, the mean propensity score for the community college transfer students was .36, whereas the mean propensity score for the group of rising 4-year college juniors was .06. This equated to a standardized difference of 1.52, which is more than seven times greater than the threshold for being considered significantly large (.20). After the match, the mean propensity score for 4-year rising juniors was .35, which was approximately the same as that of their community college transfer counterparts. The matched sample of rising 4-year college juniors demonstrated considerable movement toward the community college transfer group in terms of their mean propensity scores. Relatedly, similar movement toward the community college transfer group was evident in the baseline covariates as well. After the matching process, all of the standardized differences between the community college transfers and the rising 4-year college juniors were under the .10 threshold, with the exception of two of the baseline covariates. The matching process appeared to provide an overcorrection in terms of two of the ACT subject matter tests. After the match, the community college transfer group had slightly higher mean ACT English and science scores. Nonetheless, the standardized differences were fairly close to the .10 threshold, suggesting a slight imbalance, rather than a significant bias (as was the case prior to match). There was a .11 difference in terms of score on ACT English and a .13 difference in terms of ACT science. We argue that our matching procedures, overall, provided adequate balance on the observed baseline covariates that were used in the PSM model. In fact, as illustrated in Tables 1 through 3, most of the standardized differences were fairly close to zero. Moreover, as a result of the posttreatment adjustment, there was perfect balance in terms of institutional selectivity.

Was There a Time-Based Penalty?

To determine when a community college penalty began and for how long it continued, we used descriptive statistics (cumulative rates of bachelor’s degree completion) and inferential statistics (Pearson chi-squares). This allowed for a semester-by-semester comparison regarding bachelor’s degree completion between the observationally equivalent groups (see Table 5).

Table 5.

Cumulative Time to Bachelor’s Degree Completion.

	Fall 2003	Spring 2004	Fall 2004	Spring 2005	Fall 2005	Spring 2006	Fall 2006	Spring 2007	Fall 2007	Spring 2008	Fall 2008	Spring 2009	Fall 2009	Spring 2010
Community college transfers (n = 2,117)	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%	0.5%	29.3%	51.1%	69.2%	76.1%	80.5%	82.6%	83.7%
Rising 4-year juniors (n = 2,117)	0.0%	0.0%	0.0%	0.0%	0.4%	1.7%	5.3%	47.6%	63.8%	74.3%	78.7%	80.4%	81.8%	82.2%
Difference from community college transfers	0	0	0	0	−0.4	−1.7*	−4.8*	−18.3*	−12.7*	−5.1*	−2.6	0.1	0.8	1.5

Statistically significant at the p ≤ .001 level based on Pearson chi-square. Degrees awarded during the summer semester are included in the following fall semester’s total.

While the community college transfer students were not penalized at 7 years post high school graduation (175% of normal graduating time) or the end of the tracking period, there were differences indicative of a penalty from 75% of normal time until 125% of normal time or 5 years. As shown in Table 5, statistically significant differences (based on chi-squares) indicative of a community college penalty were evident starting in the spring semester of 2006 (75% of normal graduating time), as a limited number of the rising 4-year college juniors completed their bachelor’s degree early, whereas none of the community college transfer students met that distinction.

The largest difference between the community college transfer students and the comparison group was at the end of the spring semester of 2007 or at 100% of normal time. Slightly less than 30% of community college transfer students had finished their bachelor’s program within that 4-year timeframe compared with nearly one half of the comparison group (48%), which equated to a significantly large difference of roughly 18 percentage points.

However, the community college transfer group began to eliminate the gap with their comparison group counterparts at a fairly rapid pace. At the end of spring semester of 2008 (125% of normal time), the previously mentioned 18 percentage point difference was reduced to five and a half percentage points. At the end of the spring semester of 2009 (150% of normal time), the difference favoring the rising 4-year college juniors, or the community college penalty, was fully eliminated. In the remainder of the study, the difference between the community college transfer students and the comparison group of direct 4-year college entrants lacked statistical and practical significance. Figure 2 displays the penalty in terms of semester difference as a function of the cumulative graduation rate. Note that the extra time required for community college transfer students to achieve a comparable graduation rate to that of students who initially enroll in a 4-year institution never exceeded two semesters for cumulative graduation rates between 0.1 and 0.8.

Figure 2.

Semester penalty as a function of cumulative graduation rate.

Limitations

This study focused on a cohort of Illinois high school graduates, which necessarily limits its generalizability to Illinois. We believe that findings from this study could still be applied to other groups of students in the United States, given that higher education trends in Illinois resemble national and regional trends. Long and Kurlaender (2009) used similar logic regarding the broad applicability of Ohio data to other groups of students. An example of the applicability of Illinois data was found in Mortenson (2013) who noted that the Illinois college attendance rates and changes in public high school graduation rates were similar to that of the United States as a whole. Moreover, the Illinois college going rate is fairly similar to those of nearby Midwestern states such as Michigan, Indiana, Missouri, and Kentucky.

It should also be noted that the treatment group was a fairly select group of community college students, namely those who enrolled full-time and maintained that enrollment for 2 years prior to transferring to a 4-year college. Therefore, these results should not be applied broadly to all community college students, many of whom enroll on a part-time basis, do not maintain continuous enrollment, do not transfer to a 4-year institution, or take longer than 2 years to transfer if they do. Nonetheless, the comparison group was an equally, if not more, select group of rising 4-year college juniors. The high rate of bachelor’s degree completion overall in both groups signals the importance of initially enrolling full-time and maintaining full-time enrollment.

Finally, while the comprehensive model we used for the PSM took into consideration one psychological attribute—degree aspirations—there are other non-cognitive psychological attributes that could be included in future studies since previous studies have established their importance (Wang, 2012, 2013). Additional non-cognitive variables could not be included in the present study as these variables are not yet included in the Illinois longitudinal data systems.

Discussion

Our study contributes to the literature in several ways. First of all, we demonstrate the importance of studying time explicitly, which adds to the literature that examines the existence of a community college penalty. Furthermore, we built on the work of Bound et al. (2012) and considered all students who transfer from community college to a 4-year institution rather than focus only on those who eventually complete a bachelor’s degree. Considering all community college transfers allowed for a more complete examination of differences in the likelihood of degree completion between community college transfers and direct, 4-year college entrants.

Methodologically, this study contributes to the literature through the use of PSM. The model used to create the propensity scores was a hierarchical generalized linear model, which helped to account for the effect of differences across high schools on initial college enrollment. This study also used a posttreatment adjustment, which helped account for differences in college selectivity. Being able to create observationally equivalent groups of community college students and students who directly enroll in 4-year institutions enabled a descriptive survival analysis showing the time points at which there existed a penalty for initial community college enrollment.

Policy Implications and Directions for Future Research

In revisiting the community college penalty issue, we found that while there is an overall lack of penalty with respect to bachelor’s degree completion, there is an initial penalty from 75% of the normal time through 125% of normal graduating time. In this section, we hypothesize several reasons for the existence of this penalty and offer a number of corrective policy recommendations to help facilitate the timely success of community college transfer students. We also offer suggestions for future research.

This study shows that policymakers should not just consider whether community college transfer students are able to complete bachelor’s degrees but also when they do so. The initial penalty that exists for community college transfer students suggests the need for policies to help community college students transition more efficiently to 4-year institutions, and move more quickly toward degree completion. Several recent studies have noted challenges for community college students during transfer to a 4-year institution and have suggested strategies for connecting and integrating community college students into these institutions (Chrystal, Gansemer-Topf, & Laanan, 2013; Flaga, 2006; Packard, Gagnon, & Senas, 2012; Townsend & Wilson, 2006). In particular, scholars have noted the stress and difficulty of the social adjustment of community college transfer students, which should be addressed at the 4-year receiving institution (Bahr, Toth, Thirolf, & Massé, 2013; Chrystal et al., 2013; Laanan, 1996).

Policies that foster the development of bridge programs between 4-year institutions and feeder community colleges should be encouraged. Ackermann (1991) argued that bridge programs allow for transfer students to gain more control of their academic futures through providing for a better understanding of the resources available and how to access those resources upon transfer. Some bridge programs have been extended back into the sending community college. An emerging trend is for 4-year institutions to include potential transfer students in bridge programs and have academic advisors embedded at feeder community colleges, who can assist community college students in navigating the transfer process. Karp (2013) advocates for establishing academic bridges with more targeted academic advising for transfer students and suggests advisors should, when possible, have expertise in the specific disciplines and have knowledge about the labor market for the profession associated with the given discipline. This kind of discipline-specific bridge program can be found in Jefferson, Steadman, and Laier (2014) who described a program focused on transfer students in engineering.

Bridge programs can also provide for the social and cultural integration of community college students into the prevailing culture of the receiving 4-year institution and perhaps allow transfers to overcome that initial awkward fit as described by Townsend and Wilson (2006). This is important, as some have suggested that transfer students have a “late comer” status (Bahr et al., 2013, p. 479) and may have difficulty integrating into social networks that native students have developed during their first 2 years of college (Ackermann, 1991; Bahr et al., 2013; Townsend & Wilson, 2006). A contributing factor is that community college transfer students traditionally fall outside the scope of most retention and integration programs at 4-year institutions, which are typically designed with the first-time direct entrant in mind. Pull factors such as family responsibilities, living arrangements, and employment have also been used to describe some of the barriers to social integration among transfer students (Bahr et al., 2013). Institutions should consider both types of barriers—pull factors and late comer status—when developing bridge programs, such as having customized orientation for transfer students (Eggleston & Laanan, 2001) and building integration into academic programs to contend with the pull factors so that integration activities are also curricular as opposed to being solely extra-curricular (Jefferson et al., 2014).

While addressing experiential aspects of the student transfer process is important, it is also essential to address institutional policies that may hinder timely graduation. In the case of transfer students, policies should be put in place to help ensure that community college students receive credit for courses completed and avoid duplication of coursework. Monaghan and Attewell (2015) noted that community college transfer students tend to lose credits when they transfer to a 4-year institution. To address potential loss of credits at transfer, two types of statewide transfer policies should be considered. Johnson (2011) advocated statewide policies that establish a general curriculum that is transferrable from institution to institution. However, in many states, such general curriculum packages only account for 12 to 14 courses or 37 to 41 credits. A further consideration would be to develop expert-driven major-specific recommendations for the first 2 years of college for the most popular majors. These recommendations are typically driven by faculty, which was one of the recommendations set forth by Ignash and Townsend (2000). Taken together, both types of articulation packages would ensure more of the credits transfer. State policy can also be used to facilitate the transfer process by centralizing and standardizing transfer information to ensure it is more readily accessible and easy to understand. In addition to implementing these types of articulation agreements, more research is needed to establish the effects of these agreements on community college transfer students and timely degree completion (Bahr et al., 2013; Roksa & Keith, 2008).

Future studies should also consider the new reality of college enrollment or the fact that many college students take non-linear pathways while enrolling at multiple institutions, sometimes simultaneously (Goldrick-Rab & Pfeffer, 2009; Hillman, Lum, & Hossler, 2008; Lichtenberger, 2011). One limitation of the current study was that we did not take into consideration summer school enrollment as a predictor of bachelor’s degree completion, nor any of the other patterns in which students transfer to and from community colleges and 4-year institutions, such as reverse transferring. Future research should focus on the myriad college enrollment patterns and how each may be related to bachelor’s degree completion as a function of time. Future research should also account for credit hour accumulation prior to transfer and the number of transfer hours accepted by the receiving institution in examining the posttransfer outcomes, which could not be addressed due to some of the limitations of the current study.

Finally, while a quasi-experimental approach can reveal temporal patterns of bachelor’s community college transfer students, it does not fully explore how and why community college transfer students face an initial time-based disadvantage with respect to bachelor’s degree completion. More qualitative research should be conducted across a variety of 4-year institutions to explore how community college students experience the transfer process (see Chrystal et al., 2013).

Footnotes

Acknowledgements

The authors wish to thank the anonymous reviewers for their thoughtful comments on an earlier version of this manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Eric Lichtenberger, PhD, is the deputy director of Information Management and Research at the Illinois Board of Higher Education. He previously served as the associate director for Research at the Illinois Education Research Council at Southern Illinois University Edwardsville. His background and interests include high school to college and college to workforce transitions.

Cecile Dietrich, PhD, RN, is an applied statistician at Radford University. Her background and interests include higher education research and community mental health research. She has conducted applied research and evaluation for a variety of organizations, including the Virginia Department of Education, the Illinois Educational Research Council, the World Bank, faith-based and other non-profit organizations.

References

Ackermann

S. P.

(1991). The benefits of summer bridge programs for underrepresented and low-income transfer students. Community/Junior College Quarterly of Research and Practice, 15, 211-224. doi:10.1080/0361697910150209

Allison

P. D.

(2009). Missing data. In Millsap

Maydeu-Olivares

(Eds.), The SAGE handbook of quantitative methods in psychology (pp. 72-89). Thousand Oaks, CA: SAGE.

Alwin

D. F.

Otto

L. B.

(1977). High school context effects on aspirations. Sociology of Education, 50, 259-273. doi:10.2307/2112499

Arpino

Mealli

(2011). The specification of the propensity score in multilevel observational studies. Computational Statistics & Data Analysis, 55, 1770-1780. doi:10.1016/j.csda.2010.11.008

Avery

Hoxby

C. M.

(2004). Do and should financial aid packages affect students’ college choices? In Hoxby

C. M.

(Ed.), College choices: The economics of where to go, when to go, and how to pay for it (pp. 239-302). Chicago, IL: University of Chicago Press.

Bahr

P. R.

Toth

Thirolf

Massé

J. C.

(2013). A review and critique of the literature on community college students’ transition processes and outcomes in four-year institutions. In Paulsen

(Ed.), Higher education: Handbook of theory and research (pp. 459-511). Dordrecht, The Netherlands: Springer Netherlands. doi:10.1007/978-94-007-5836-0_10

Bailey

Jenkins

Leinbach

(2005). What we know about community college low-income and minority student outcomes: Descriptive statistics from national surveys. New York, NY: Teachers College, Community College Research Center, Columbia University. Retrieved from http://ccrc.tc.columbia.edu/Publication.asp?UID=207

Barreno

Traut

C. A.

(2012). Student decisions to attend public two-year community colleges. Community College Journal of Research and Practice, 36, 863-871. doi:10.1080/10668920903505007

Barron’s. (2003). Profile of American colleges. Woodbury, NY: Barron’s Education Series.

10.

Baryla

Dotterweich

(2001). Student migration: Do significant factors vary by region? Education Economics, 9, 269-280. doi:10.1080/09645290110086135

11.

Beal

S. J.

Crockett

L. J.

(2010). Adolescents’ occupational and educational aspirations and expectations: Links to high school activities and adult educational attainment. Developmental Psychology, 46, 258-265. doi:10.1037/a0017416

12.

Bean

J. P.

(1990). Why students leave: Insights from research. In Hossler

J. P.

Bean Associates (Eds.), The strategic management of college enrollments (pp. 170-185). San Francisco, CA: Jossey-Bass.

13.

Bean

J. P.

Metzner

B. S.

(1985). A conceptual model of nontraditional undergraduate student attrition. Review of Educational Research, 55, 485-540. doi:10.3102/00346543055004485

14.

Bettinger

(2004). How financial aid affects persistence. In Hoxby

C. M.

(Ed.), College choices: The economics of where to go, when to go, and how to pay for it (pp. 207-238). Chicago, IL: University of Chicago Press. doi:10.3386/w10242

15.

Bound

Lovenheim

M. F.

Turner

(2012). Increasing time to baccalaureate degree in the United States. Education Finance and Policy, 7, 375-424. doi:10.1162/EDFP_a_00074

16.

Boxer

Goldstein

S. E.

DeLorenzo

Savoy

Mercado

(2011). Educational aspiration–expectation discrepancies: Relation to socioeconomic and academic risk-related factors. Journal of Adolescence, 34, 609-617. doi:10.1016/j.adolescence.2010.10.002

17.

Brand

J. E.

Pfeffer

F. T.

Goldrick-Rab

(2014). The community college effect revisited: The importance of attending to heterogeneity and complex counterfactuals. Sociological Science, 1, 448-465. doi:10.15195/v1.a25

18.

Chrystal

L. L.

Gansemer-Topf

Laanan

F. S.

(2013). Assessing students’ transition from community college to a four-year institution. Journal of Assessment and Institutional Effectiveness, 3, 1-18.

19.

Cohen

A. M.

Brawer

F. B.

(2008). The American community college (5th ed.). San Francisco, CA: Jossey-Bass.

20.

Cohen

(1977). Statistical power analysis for the behavioral sciences. Toronto, Canada: Ontario Academic Press.

21.

Cohen

(1975). Applied multiple regression correlation analysis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum.

22.

Cohen

(1985). Applied multiple regression correlation analysis for the behavioral sciences (2nd ed.). Mahwah, NJ: Lawrence Erlbaum.

23.

College Board. (n.d.). Community college: Answers to frequently asked questions about community colleges. Retrieved from Https://bigfuture.collegeboard.org/find-colleges/how-to-find-your-college-fit/community-college-faqs

24.

Cross

(1971). Access and accommodation in higher education. Research Reporter, 6(2), 6-8.

25.

Dadgar

(2012). Essays on the Economics of Community College Students’ Academic and Labor Market Success (Unpublished doctoral dissertation). Columbia University, New York, NY.

26.

de Araujo

Murray

. (2010). Channels for improved performance from living on campus. American Journal of Business Education, 3(12), 57-64. doi:10.19030/ajbe.v3i12.965

27.

Dehejia

R. H.

Wahba

(2002). Propensity score-matching methods for nonexperimental causal studies. The Review of Economics and Statistics, 84, 151-161. doi:10.1162/003465302317331982

28.

DesJardins

S. L.

Ahlburg

D. A.

McCall

B. P.

(2002). A temporal investigation of factors related to timely degree completion. The Journal of Higher Education, 73, 555-581. doi:10.1353/jhe.2002.0042

29.

DesJardins

S. L.

Ahlburg

D. A.

McCall

B. P.

(2006). An integrated model of application, admission, enrollment, and financial aid. The Journal of Higher Education, 77, 381-429. doi: 10.1353/jhe.2006.0019

30.

Dietrich

Lichtenberger

(2015). Using propensity score matching to test the community college penalty assumption. The Review of Higher Education, 38, 193-219. doi:10.1353/rhe.2015.0013

31.

Dougherty

Kienzl

(2006). It’s not enough to get through the open door: Inequalities by social background in transfer from community colleges to four-year colleges. The Teachers College Record, 108, 452-487. doi:10.1111/j.1467-9620.2006.00658.x

32.

Eagan

M. K.

Hurtado

Chang

M. J.

Garcia

G. A.

Herrera

F. A.

Garibay

J. C.

(2013). Making a difference in science education: The impact of undergraduate research programs. American Educational Research Journal, 50, 683-713. doi:10.3102/0002831213482038

33.

Eggleston

L. E.

Laanan

F. S.

(2001). Making the transition to the senior institution. New directions for community colleges, 2001(114), 87-97. doi:10.1002/cc.23

34.

Flaga

C. T.

(2006). The process of transition for community college transfer students. Community College Journal of Research and Practice, 30, 3-19. doi:10.1080/10668920500248845

35.

Fletcher

J. M.

(2012). Similarity in peer college preferences: New evidence from Texas. Social Science Research, 41, 321-330. doi:10.1016/j.ssresearch.2011.11.001

36.

Flores

C. A.

Flores-Lagunes

(2009). Identification and estimation of causal mechanisms and net effects of a treatment. Gainesville, FL: IZA Discussion Paper. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1423353

37.

Frangakis

C. E.

Rubin

D. B.

(2002). Principal stratification in causal inference. Biometrics, 58, 21-29. doi:10.1111/j.0006-341X.2002.00021.x

38.

Gemici

(2011). Greatly underestimated: The role of vocational curriculum tracks in US high schools. International Journal of Vocational Education and Training, 19(2), 7-20.

39.

Glass

J. C.

Harrington

A. R.

(2002). Academic performance of community college transfer students and “native” students at a large state university. Community College Journal of Research and Practice, 26, 415-430. doi:10.1080/02776770290041774

40.

Goldrick-Rab

(2006). Following their every move: An investigation of social-class differences in college pathways. Sociology of Education, 79, 67-79. doi:10.1177/003804070607900104

41.

Goldrick-Rab

(2010). Challenges and opportunities for improving community college student success. Review of Educational Research, 80, 437-469. doi:10.3102/0034654310370163

42.

Goldrick-Rab

Pfeffer

F. T.

(2009). Beyond access: Explaining socioeconomic differences in college transfer. Sociology of Education, 82, 101-125. doi:10.1177/003804070908200201

43.

Greenland

Finkle

W. D.

(1995). A critical look at methods for handling missing covariates in epidemiologic regression analyses. American Journal of Epidemiology, 142, 1255-1264.

44.

Griswold

M. E.

Localio

A. R.

Mulrow

(2010). Propensity score adjustment with multilevel data: Setting your sites on decreasing selection bias. Annals of Internal Medicine, 152, 393-395. doi:10.7326/0003-4819-152-6-201003160-00010

45.

Hillman

Lum

Hossler

(2008). Understanding Indiana’s reverse transfer students: A case study in institutional research. Community College Journal of Research and Practice, 32, 113-134. doi:10.1080/10668920701707797

46.

Ignash

J. M.

Townsend

B. K.

(2000). Evaluating state-level articulation agreements according to good practice. Community College Review, 28(3), 1-21.

47.

Jefferson

G. D.

Steadman

Laier

J. E. (April, 2014)

. Addressing transfer student transition. Paper presented at The ASEE Southeastern Section Conference, Macon, Georgia. Retrieved from http://asee.cs.southern.edu/proceedings/ASEE2014/Papers2014/4/129.pdf

48.

Johnson

(2011). Three policies to reduce time to degree. Washington, DC: Complete College America. Available from www.completecollege.org

49.

Karp

(2013). Entering a program: Helping students make academic and career decisions (CCRC Working Paper No. 59). New York, NY: Teachers College, Community College Research Center, Columbia University. doi:10.7916/D8CJ8BGD

50.

Karp

Calcagno

Hughes

Jeong

D. W.

Bailey

(2007). The postsecondary achievement of participants in dual enrollment: An analysis of student outcomes in two states. St. Paul, MN: National Research Center for Career and Technical Education. Retrieved from http://ccrc.tc.columbia.edu/media/k2/attachments/dual-enrollment-student-outcomes.pdf

51.

Karp

Hughes

(2008). Dual enrollment can benefit a wide range of students. Techniques, 83(7), 14-17.

52.

Kolesnikova

N. A.

(2010). Community colleges and economic mobility. Federal Reserve Bank of St. Louis Review, 92, 27-53.

53.

Koranteng

(2015). Researchers find no evidence of “Community College penalty.” Retrieved from http://www.nacacnet.org/learning/communities/Admitted/Lists/Posts/Post.aspx?ID=694

54.

Laanan

F. S.

(1996). Making the transition: Understanding the adjustment process of community college transfer students. Community College Review, 23(4), 69-84.

55.

Lee

V. E.

Mackie-Lewis

Marks

H. M.

(1993). Persistence to the baccalaureate degree for students who transfer from community college. American Journal of Education, 102, 80-114.

56.

Lemons

C. J.

Fuchs

Gilbert

J. K.

Fuchs

L. S.

(2014). Evidence-based practices in a changing world: Reconsidering the counterfactual in education research. Educational Researcher, 43, 242-252. doi:10.3102/0013189X14539189

57.

Levesque

Laird

Hensley

Choy

S. P.

Cataldi

E. F.

Hudson

(2008). Career and Technical Education in the United States: 1990 to 2005 (Statistical Analysis Report). Washington, DC: National Center for Education Statistics. Retrieved from http://files.eric.ed.gov/fulltext/ED502109.pdf

58.

Lichtenberger

(2011). Reverse transfer students and postsecondary outcomes: A potential opportunity (IERC 2011-5). Edwardsville: Illinois Education Research Council.

59.

Lichtenberger

Dietrich

(2012). College readiness and the postsecondary outcomes of Illinois high school students (IERC 2012-1). Edwardsville: Illinois Education Research Council.

60.

Lichtenberger

Dietrich

(2013). The community college penalty and bachelor’s degree completion: Fact or fiction? (IERC 2013-1). Edwardsville: Illinois Education Research Council.

61.

Lichtenberger

Witt

Blankenberger

Franklin

(2014). Dual credit/dual enrollment and data driven policy implementation. Community College Journal of Research and Practice, 38, 959-979. doi:10.1080/10668926.2013.790305

62.

Long

B. T.

Kurlaender

(2009). Do community colleges provide a viable pathway to a baccalaureate degree? Educational Evaluation and Policy Analysis, 31, 30-53. doi:10.3102/0162373708327756

63.

Mattern

Wyatt

Shaw

(2013). College distance from home: Implications for student transfer behavior. Journal of The First-Year Experience & Students in Transition, 25, 77-92.

64.

Melguizo

Dowd

A. C.

(2009). Baccalaureate success of transfers and rising 4-year college juniors. Teachers College Record, 111, 55-89.

65.

Melguizo

Kienzel

G. S.

Alfonso

(2011). Comparing the educational attainment of community college transfer students and four-year college rising juniors using propensity score matching methods. Journal of Higher Education, 82, 265-291. doi:10.1353/jhe.2011.0013

66.

Monaghan

D. B.

Attewell

(2015). The community college route to the bachelor’s degree. Educational Evaluation and Policy Analysis, 37, 70-91. doi:10.3102/0162373714521865

67.

Mortenson

(2013). Chance for college by age 19 by state in 1986 to 2010. Postsecondary Education Opportunity, 247, 1-16.

68.

National Center for Education Statistics. (2011). 2008-09 Baccalaureate and beyond longitudinal study. Retrieved from http://nces.ed.gov/pubs2011/2011236.pdf

69.

Nelson

J. I.

(1972). High school context and college plans: The impact of social structure on aspirations. American Sociological Review, 37, 143-148.

70.

Normand

S. L. T.

Landrum

M. B.

Guadagnoli

Ayanian

J. Z.

Ryan

T. J.

Cleary

P. D.

McNeil

B. J.

(2001). Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: A matched analysis using propensity scores. Journal of Clinical Epidemiology, 54, 387-398. doi:10.1016/S0895-4356(00)00321-8

71.

Oakes

J. M.

(2010). Applied propensity score analysis: Rationale and basic techniques. Academy Health. Retrieved from https://www.academyhealth.org/Training/ResourceDetail.cfm?itemnumber=3986

72.

S. R.

Reynolds

A. J.

(2008). Predictors of educational attainment in the Chicago longitudinal study. School Psychology Quarterly, 23, 199-229. doi:10.1037/1045-3830.23.2.199

73.

Packard

B. W. L.

Gagnon

J. L.

Senas

A. J.

(2012). Navigating community college transfer in science, technical, engineering, and mathematics fields. Community College Journal of Research and Practice, 36, 670-683. doi:10.1080/10668926.2010.495570

74.

Reese

(2008). Doubling the opportunity for success. Techniques, 83(7), 18-21.

75.

Roderick

Coca

Nagaoka

(2011). Potholes on the road to college high school effects in shaping urban students’ participation in college application, four-year college enrollment, and college match. Sociology of Education, 84, 178-211. doi:10.1177/0038040711411280

76.

Roksa

Keith

(2008). Credits, time, and attainment: Articulation policies and success after transfer. Educational Evaluation and Policy Analysis, 30, 236-254. doi:10.3102/0162373708321383

77.

Rosenbaum

P. R.

(1995). Observational studies. New York, NY: Springer-Verlag.

78.

Rosenbaum

P. R.

Rubin

D. B.

(1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41-55. doi:10.1093/biomet/70.1.41

79.

Rosenbaum

P. R.

Rubin

D. B.

(1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician, 39, 33-38. doi:10.1080/00031305.1985.10479383

80.

Rowan-Kenyon

H. T.

Perna

L. W.

Swan

A. K.

(2011). Structuring opportunity: The role of school context in shaping high school students’ occupational aspirations. Career Development Quarterly, 59, 330-344. doi:10.1002/j.2161-0045.2011.tb00073.x

81.

Sheehy

(2015, February 9). Dorms help give 2 year colleges a 4 year feel. U.S. News & World Report. Retrieved from http://www.usnews.com/education/community-colleges/articles/2015/02/09/dorms-help-give-2-year-colleges-a-4-year-feel

82.

Smalley

Lichtenberger

Brown

(2010). A longitudinal study of the Illinois high school class of 2002: A six-year analysis of postsecondary enrollment and completion (IERC 2010-3). Edwardsville: Illinois Education Research Council.

83.

Stuart

E. A.

(2010). Matching methods for causal inference: A review and a look forward. Statistical Science, 25, 1-21. doi:10.1214/09-STS313

84.

Surette

B. J.

(2001). Transfer from two-year to four-year college: An analysis of gender differences. Economics of Education Review, 20, 151-163. doi:10.1016/S0272-7757(00)00013-3

85.

Thoemmes

F. J.

West

S. G.

(2011). The use of propensity scores for nonrandomized designs with clustered data. Multivariate Behavioral Research, 46, 514-543. doi:10.1080/00273171.2011.569395

86.

Townsend

B. K.

Wilson

(2006). “ A hand hold for a little bit”: Factors facilitating the success of community college transfer students to a large research university. Journal of College Student Development, 47, 439-456. doi:10.1353/csd.2006.0052

87.

Tuckman

H. P.

(1970). Determinants of college student migration. Southern Economic Journal, 37, 184-189. doi:10.2307/1056128

88.

Wang

(2009). Baccalaureate attainment and college persistence of community college transfer students at four-year institutions. Research in Higher Education, 50, 570-588. doi:10.1007/s11162-009-9133-z

89.

Wang

(2012). Factors contributing to the upward transfer of baccalaureate aspirants beginning at community colleges. The Journal of Higher Education, 83, 851-875.

90.

Wang

(2013). Why students choose STEM Majors: Motivation, high school learning, and postsecondary context of support. American Educational Research Journal, 50, 1081-1121. doi:10.3102/0002831213488622

91.

Weldon

(2013, August). Reducing time to degree by cutting credit creep (Research Brief). Lexington, KY: The Council of State Governments. Retrieved from http://knowledgecenter.csg.org/kc/content/reducing-time-degree-cutting-credit-creep