EASEing Students Into College: The Impact of Multidimensional Support for Underprepared Students

Abstract

Extensive theoretical literature and qualitative evidence nominate learning communities as a promising strategy to improve persistence and success among at-risk populations, such as students who are academically underprepared for college-level coursework. Yet rigorous quantitative evidence on the impacts of these programs is limited. This paper estimates the causal effects of a first-year STEM learning communities program on both cognitive and noncognitive outcomes at a large public 4-year institution. We use a regression discontinuity design based on the fact that students are assigned to the program if their math SAT score is below a threshold. Our results indicate that program participation increased the academic performance and sense of belonging for students around the cutoff. These results provide compelling evidence that learning communities can support at-risk populations when implemented with a high level of fidelity.

Keywords

achievement gap econometric analysis educational policy engagement learning communities program motivation policy analysis postsecondary education program evaluation quasi-experimental analysis regression discontinuity retention STEM persistence student outcomes

“I have made amazing friends in my cohort. EASE has helped me find my ‘people’ at UCI that I can have fun with as well as study and learn from. EASE gave me a ton of resources.”

Researchers and policymakers have been concerned about the lower representation of underrepresented racial and ethnic minority (URM) students in science, technology, engineering, and mathematics (STEM) for decades. Indeed, while URM students collectively make up almost one third of the higher education population, those groups together make up only 12% of STEM workers (National Center for Education Statistics [NCES], 2013). The implication of such a discrepancy for the national equity agenda is substantial, considering that occupations in STEM fields on average confer considerably higher earnings than non-STEM fields and therefore promise disproportionate gains in long-term economic mobility.

Why aren’t there more URMs in STEM? One source of this discrepancy is the low graduation rates among racial minorities in STEM disciplines. National reports consistently found that while URM students have reached parity with their White and Asian American counterparts in their proportional interest in majoring in STEM disciplines at the beginning of their undergraduate career, their completion rates are only half to two thirds of the completion rates among White and Asian Americans (Center for Institutional Data Exchange and Analysis [C-IDEA], 2001; Higher Education Research Institute [HERI], 2010; Huang, Taddese, & Walter, 2000). Therefore, while the attrition rates for STEM fields are generally higher than non-STEM fields across all subgroups of students, attaining a bachelor’s degree in STEM is especially unlikely for URMs.

The low graduation rates among URMs in STEM have spurred decades of academic studies and policy analyses (Chen, 2013; Seymour & Hewitt, 1997), but researchers have yet to achieve consensus on either the biggest challenges faced by these students or most effective ways to better support them. Historically, colleges have mainly attributed the low graduation rates to students’ inadequate academic preparedness upon college enrollment. This argument has been increasingly challenged in the literature on college persistence and success; indeed, in the past two decades, a greater number of researchers have highlighted the importance of engaging students both academically and socially (e.g., Bergen & Milem, 1999; Kuh, Kinzie, Schuh, & Whitt, 2005; Robinson, 1996; Tinto, 2017).

A number of institutionalized efforts have been made to provide students with academic and social support, such as peer advising, intrusive academic advising, student coaching, and cohorting programs (Bettinger & Baker, 2014; DiMaria, 2006; Smith, MacGregor, Matthews, & Gabelnick, 2004; MDRC, 2012). One promising institutionalized effort to support students in both domains simultaneously is through learning communities (Smith et al., 2004). Broadly defined as a group of students sharing learning activities together, learning communities intend to support students academically as well as connecting students emotionally. However, while an extensive theoretical literature and qualitative evidence nominates learning communities as a promising strategy to improve persistence and success among at-risk populations in college, such as students who are academically underprepared for college-level coursework (see Dagley, Georgiopoulos, Reece, & Young, 2015; Engstrom & Tinto, 2007; Scrivener et al., 2008; Shapiro & Levine, 1999; Smith et al., 2004; Taylor, Moore, MacGregor, & Lindblad, 2003; Tinto & Goodsell, 1993; Tinto & Russo, 1994; Zhao & Kuh, 2004), experimental and quasi-experimental evidence on the impacts of existing programs on students’ academic performance and social integration is still limited. This is particularly the case in regard to learning communities in specific fields of study, such as STEM-focused learning communities.

This paper estimates the causal effects of a first-year learning communities program—Enhanced Academic Success Experience initiative (EASE)—provided to freshmen majoring in biological sciences (bio sci) at a large public Hispanic-serving institution (HSI), using a regression discontinuity (RD) design based on the fact that students are assigned to the program if their math SAT score falls below a threshold. EASE represents a strong institutionalized effort, and two key features distinguish it from most of the existing learning communities. First, instead of allowing students to volunteer into the program, EASE students were grouped into cohorts of 30. Each student cohort was enrolled in the same lecture and discussion sections of all major-related courses throughout the freshman year. The main goal was for these small groups to engage in shared learning activities and develop stronger student relationships, thereby increasing students’ sense of belonging to the major and at the college. Second, additional academic and social support was provided to students enrolled in EASE, where each cohort of 30 was connected with a senior bio sci major who mentored them by providing advice and conducting 1-hour weekly meetings with the cohort. The personalized advising and guidance intends to not only bridge students’ informational gaps but also support the students in developing academic and social-emotional skills throughout their first year in college.

We examine the impact of the program on a variety of outcome measures, including course performance in gateway biology courses, cumulative first-year GPA, and retention within the bio sci major, as well as noncognitive measures of social integration, such as a sense of belonging as a bio sci major. Our results indicate that EASE participation significantly increased students’ grades in core college-level courses and first-year cumulative GPA, boosted students’ sense of belonging to the bio sci major, and most importantly, improved first-year major persistence for students on the margin of being eligible to participate in the program. We conduct several robustness checks to rule out alternative explanations for these findings. These promising results provide compelling evidence that learning communities programs can provide effective support to at-risk populations when implemented with a high level of fidelity.

Theoretical Framework and Extant Research

Learning communities represent an educational strategy intended to improve college student engagement and success. It is informed by Tinto’s (1975) integration theory, which posits that academic and social integration are key components of academic performance and both college student retention and persistence (Astin, 1984; Braxton & McClendon, 2001; Carini, Kuh, & Klein, 2006; Tinto, 1975; Tinto & Goodsell, 1993).

Tinto’s theory offers colleges and universities a model of student persistence in which academic integration (AI), referring to a student’s academic performance and course-related conduct, is marked by behaviors such as attending class, completing courses, interacting with faculty, and attending study groups. Social integration, the other main component of Tinto’s theory, consists of two components: a student’s level of participation in the social culture of a college, such as through campus-related activities, and perceived social integration.

Modern theorists have placed greater emphasis on the second component, positing that a subjective sense of belonging (defined as feeling acceptance, respect, and inclusion as well as feeling valued within a group) is a critical component of social integration and is particularly relevant to student learning outcomes (Hurtado & Carter, 1997; Strayhorn, 2012). Indeed, research has shown that the psychological processes associated with sense of belonging have the potential to interfere with academic function, as common challenges become much more severe when students feel they are the only ones dealing with them or when they feel that those of their demographic do not often succeed (Strayhorn, 2008, 2012). Low-income, first-generation, and URM college students—as well as women in male-dominated majors—often fit these criteria (Darling, Molina, Sanders, Lee, & Zhao, 2008; Freeman, Anderman, & Jensen, 2007; Hoffman, Richmond, Morrow, & Salomone, 2003; Hurtado & Carter, 1997; Master, Cheryan, & Metzloff, 2016; Ostrove & Long, 2007; Strayhorn, 2008, 2012). Further, research has shown that sense of belonging is particularly relevant in STEM courses, as it is common for STEM students to find that their racial or gender groups are underrepresented (Murphy & Zirkel, 2015). In short, all existing studies point to the idea that developing a strong sense of belonging early in one’s college career is key.

Learning communities emerge in the midst of this discussion as a way for colleges and universities to promote academic and social integration simultaneously. They are intentionally designed to increase opportunities for students to interact with peers, faculty, and the curriculum, therefore allowing for the construction of a strong support system and the development of a strong sense of belonging (Smith et al., 2004).

The majority of learning communities incorporate active and collaborative learning activities (e.g., students co-enrolling in courses) and promote involvement in complementary academic and social activities that extend beyond the classroom (e.g., students meet weekly in a study skills course and/or with a group mentor) (Zhao & Kuh, 2004). Indeed, prior research has documented that first-year students who participate in a learning community have higher grades, retention rates, and self-reported levels of engagement when compared to peers who have not had a learning community experience. Further, learning community students have reported studying more with peers outside of class and becoming more involved in university activities (Engstrom & Tinto, 2007; Shapiro & Levine, 1999; Taylor et al., 2003; Tinto & Goodsell, 1993; Tinto & Russo, 1994; Zhao & Kuh, 2004).

Despite these promising factors, most of the existing research that examines the correlation between participation in learning communities and student college experiences is based on voluntary enrollment in the communities. While this research has generally identified a positive correlation between participation in learning communities and improved academic and social outcomes, it is possible that these results are subject to biases regarding students’ self-selection into these programs.

The most rigorous evaluation of learning communities to date is the Learning Communities Demonstration by MDRC (2012), an evaluation of a one-semester learning communities program for students assigned to developmental English classes at six community colleges. Within each college, students from a variety of fields were randomly assigned to participate in the learning communities program during their initial semester in college. In general, the study found a small positive impact on academic-related measures, such as credits earned, but failed to find any consistent evidence demonstrating that learning communities positively influenced students’ persistence. Importantly, however, the study also found that students in the Kingsborough Opening Doors Learning Communities program, which provided enhanced support services such as academic counseling and tutoring to students—a feature distinguishing it from the other five programs—reported a noticeably larger impact on academic credits earned than that reported by the other campuses. This suggests that more comprehensive learning community programs providing intensive support services may yield greater benefits; in fact, a separate report from MDRC (Scrivener et al., 2008) discussing the 6-year follow-up results of the Kingsborough program identified the positive impact of program participation on graduation rates.

Overall, researchers have theorized that learning communities—particularly those with enhanced support services—are a promising way to improve the college experience and student learning outcomes. To date, the rigorous evaluations that do exist demonstrate what a typical learning communities program looks like and delineate the average impact in the community college context. In this study, we contribute to the discussion about learning communities by documenting and rigorously evaluating a first-year STEM learning communities program in a 4-year public university that not only groups students into cohorts but also provides enhanced support systems, such as peer mentoring resources. In addition to measuring student academic performance measures that are used in most of the existing studies to evaluate the impact of learning communities, we also measure the impact of learning communities on manifestations of social integration, such as a sense of belonging obtained through student surveys.

Data and Research Background

Research Setting and the EASE Learning Communities Program

This study was conducted at the University of California, Irvine (UCI). As an HSI, roughly a quarter of the student body belongs to an underrepresented minority group. When addressing inequalities in STEM education, a heavily impacted target population includes majors in the bio sci. A typical incoming class in the Ayala School of Biological Sciences at UCI consists of over 50% first-generation, over 30% URM, and over 40% low-income students.¹ Roughly 35% of incoming freshmen who enrolled as bio sci majors over the past decade did not graduate with this major in 4 years, with at-risk students being disproportionately represented in this number.

Considering the large proportion of historically at-risk students admitted into UCI, the School of Biological Sciences created the EASE initiative. The EASE program evolved through two stages—a pilot stage and a full-implementation stage. During the 2015–16 academic year, a pilot stage of the EASE initiative was implemented and 120 students volunteered into the program. These students were grouped into cohorts of 30 and were enrolled in the same lecture and discussion sections of biology and chemistry courses during their freshman year. In Fall 2016, the EASE program was further expanded to include all freshmen with a math SAT score of 600 or less, as math SAT has been found to be the most significant predictor of major success.²

During the full-implementation stage, the core components of the EASE program were further expanded to provide multidimensional support to students. Specifically, the program included two key components: (i) an academic component, where all EASE students were required to take an additional developmental chemistry course (Chem 1P) during the fall quarter, intended to prepare bio sci majors for subsequent college-level courses in chemistry and biology, and (ii) a social-psychological component, where EASE students were grouped into cohorts of 30 students. These student cohorts were enrolled in the same required biology and chemistry classes and discussion sections during their freshman year. In addition, EASE students also received increased academic counseling. Each cohort of 30 was connected with a senior bio sci major who mentored them by providing advice and conducting 1-hour weekly meetings with the cohort. The weekly meetings included a variety of academic-related topics, for example study skills, metacognition, and issues encountered at UCI.

The EASE program was well received by participants, and students explicitly indicated that the learning communities program effectively facilitated social integration. For example, one EASE participant mentioned³:

I liked how I always knew someone in my class before I had it and how everyone was friendly with each other. UCI is a big school and it’s nice to feel like you know a couple of familiar faces.

Another student also made a similar comment:

I have made amazing friends in my cohort. EASE has helped me find my “people” at UCI that I can have fun with as well as study and learn from. EASE gave me a ton of resources.

Data and Sample Description

Data for this study came from multiple sources. Specifically, the UCI Registrar’s Office provided information on student demographic characteristics and SAT scores as well as their UCI transcript and declared major following the academic year. We also collected program-level data that recorded students’ actual enrollment in the EASE program. Finally, students’ academic and social-psychological engagement was collected through two waves of surveys (see Appendix A for survey items and details): (i) a presurvey conducted at the beginning of the fall quarter, and (ii) a follow-up survey conducted at the end of the fall quarter. These surveys took roughly 10 minutes to complete. They consisted of roughly 100 items exclusively from existing, validated measures whose reliability coefficients were consistent with those of previous studies. We adapted previously established instruments targeting three noncognitive measures, including sense of belonging (Hoffman et al., 2002), motivation (Eccles et al., 1983), and attitudes about science (Adams et al., 2006). Items on motivation and belongingness were adapted to be domain-specific, asking explicitly about these constructs with respect to the subject of biology. In the follow-up survey, students were asked about the frequency of their course-related and social behaviors since the beginning of the academic year. Student completion of the surveys was tied to course credit and therefore yielded an extremely high response rate of roughly 95%.

The sample of this study includes all bio sci freshmen majors who entered during Fall 2016 with an available math SAT score (N = 907). Nearly half of the entering cohort were below 600 in SAT math and were hence required to enroll in the EASE program. Panel A in Table 1 provides descriptive statistics on demographic characteristics for the full sample of bio sci students and breaks down these numbers by EASE program assignment. Students assigned to the EASE program were significantly more likely to be from underrepresented (55%), first-generation (63%), and low-income (55%) backgrounds. Additionally, EASE students were more likely to be female (78%).

Table 1

Summary Statistics

			EASE Status			EASE Eligibility Status
	Full Sample		Control	Participant		Ineligible	Eligible
	M	SD	M	M	t Test of Difference	M	M	t Test of Difference
Panel A. Descriptive statistics of the full sample of students, EASE participants, and EASE-eligible students
Female	68.4	(46.5)	61.3	78.0	−5.452	58.8	77.5	−6.187
White	12.5	(33.1)	13.1	11.6	0.647	11.3	13.6	−1.02
URM	33.7	(47.3)	17.7	55.0	−12.613	13.6	52.3	−13.282
Asian	53.9	(49.9)	69.2	33.3	11.319	75.1	34.1	13.342
First-generation status	48.4	(50.0)	37.2	63.4	−8.085	31.9	63.9	−10.16
Low-income status	40.7	(49.2)	30.3	54.6	−7.623	26.9	53.6	−8.509
SAT read score	558.9	(76.45)	581.8	528.2	11.137	589.2	530.4	12.55
SAT math score	598.0	(88.48)	653.4	523.9	31.59	672.0	528.5	41.696
N	907		519	388		439	468
Panel B. Summary statistics for outcome measures by full sample and EASE participation status
Academic-related measures
Bio93 course grade	3.051	(0.952)	3.314	2.702	10.088
Bio94 course grade	2.818	(0.900)	2.895	2.716	2.897
Year 1 GPA	3.056	(0.607)	3.128	2.952	4.153
Retained	0.933	(0.250)	0.959	0.898	3.654
Non-academic-related measures
Sense of belonging	4.544	(1.213)	4.503	4.599	1.164
Academic integration	0.854	(0.573)	0.775	0.960	4.754
Academic and social concerns	3.902	(1.554)	3.935	3.859	0.713
Interest	5.837	(1.182)	5.790	5.868	0.917

Note. Non-academic-related measures were surveyed at Wave 2. For academic and social concerns, higher values indicate more concern.

Outcome Measures

We focus on two sets of student outcome measures: academic outcomes and social-psychological outcomes. We describe below how each of the key outcome measures is constructed. More detailed information regarding the specific items included in each of the social-psychological constructs is in Appendix A.

Academic outcomes consisted of performance and persistence data, including the following:

Bio 93 grade was the student’s final score for the first introductory biology course, “From DNA to Organisms.” This course took place during the fall quarter and was measured on a standard 4-point scale.

Bio 94 grade was the student’s final score for the second introductory biology course, “From Organisms to Ecosystems.” This course took place during the winter quarter and was also measured on a standard 4-point scale.

Overall GPA was the student’s cumulative GPA at the end of the first academic year. Bio sci majors typically take chemistry, biology, and mathematics courses during their freshman year.

Retention was a dichotomous variable, measuring whether students were still bio sci majors during the summer after their first year (coded as 1) or had switched majors (coded as 0).

Social-psychological outcomes consisted of a variety of students’ attitudes and behaviors regarding the field of biology. These were scaled measures, for which individual items and Cronbach’s alphas⁴ are listed in Appendix A.

Belonging in biology assessed the extent to which students felt they belonged in the discipline of biology at UCI. These items, from Hoffman and colleagues’ (2002) Sense of Belonging Scale, have been validated in multiple college-aged samples and focus on belonging derived from students’ perceptions of peer support, classroom comfort, and faculty support (Tovar & Simon, 2010). Items were adapted to be specific to the biology discipline, rather than to the university in general.

Academic and social concerns (ASC) measured the extent to which participants worried that on campus, in general, others would dislike them or form negative evaluations of their academic abilities. This measure is a shortened version of Heatherton and Polivy’s (1991) State Self-Esteem Scale, which was originally validated and continues to be used among undergraduate populations (e.g., Cohen & Garcia, 2005; Sherman, Bunyan, Creswell, & Jaremka, 2009).

AI measured the frequency with which participants engaged in various school-related activities during their first term on campus. The academic behaviors are considered to be representative of Tinto’s (1975) theory of AI, such as the amount of time talking with faculty, planning with academic advisors, and attending study groups. These items have been employed in nationally representative samples of postsecondary students, often aggregated to create an “academic integration index” (Flynn, 2014).

Interest in biology assessed students’ fascination and excitement regarding their impending studies in the field of biology. Rooted in the Eccles et al. (1983) Expectancy-Value Model of motivation, interest is a critical indicator of how much students value a field of study. Items were drawn from a recent study specifically focusing on the interest of college students in STEM disciplines (Harackiewicz et al., 2015).

Panel B of Table 1 provides descriptive statistics on key outcome measures for the full sample of bio sci students, also breaking down the numbers by EASE program assignment. EASE participants on average had lower grades in their first two biology courses, lower first-year GPA, and a weaker likelihood of continuing in the bio sci major compared to students who did not participate in the EASE program. In terms of non-academic-related items, EASE participants reported a statistically different value for the measure of AI, as compared to their non-EASE counterparts.

Methodology

Because the EASE program follows an SAT cutoff score for assignment, we are able to utilize an RD design to estimate the causal impacts of the program on student outcomes. In the current research context, if we assume that nothing other than the EASE assignment varies discontinuously at the cutoff, we may attribute any observed discontinuity in outcomes at the cutoff to the EASE program. For example, while we might expect first-year major persistence would be positively related to students’ math SAT scores, there is no reason other than the EASE program to expect a discontinuous jump in this relationship for students that fall just around the score cutoff (math SAT score of 600). The basic implementation of the RD design identifies the impact of the EASE program by comparing outcomes of students who score barely above the 600 SAT cutoff score with those who score barely below; these students sharply differ in EASE assignment yet are otherwise very similar. As a result, the regression coefficient can be then interpreted as the causal impact of the intervention for students on the margin of passing the cutoff (Levin & Calcagno, 2008).

Following Imbens and Lemieux (2008), we focus primarily on a local linear estimation that is limited to a narrow bandwidth around the cutoff to estimate the effect of EASE assignment on our outcome measures and therefore represent our intent-to-treat (ITT) estimates. The equation writes as follows:

\begin{array}{l} \begin{array}{l} Y_{i} = α + β_{1} ({Below}_{i}) + β_{2} ({ScoreDistance}_{i}) + β_{3} \\ ({ScoreDistance}_{i} * {Below}_{i}) \end{array} \\ {+ X}_{i} + μ \end{array},

(1)

where Y is the outcome measure (i.e., course grade, first-year GPA, or first-year persistence); Below is a binary indicator of whether or not the student was assigned to the EASE program; ScoreDistance is the difference between the student’s math SAT score and the EASE cutoff score (i.e., 600); the interaction term between EASE assignment (Below) and the running variable (ScoreDistance) allows different slopes above and below the cutoff score. X_i is a vector of individual-level covariates listed in Table 1, Panel A.

Our main model is a local linear regression within a bandwidth of ± 60.⁵. We have also conducted separate robustness checks where we added quadratic terms for the test score distance variable to allow the regression function to be more flexible in capturing possible nonlinear relationships between the running variable and the outcome measures. Results from these robustness checks are presented in Tables C1 and C2 in Appendix C.

The traditional RD method, known as a “sharp RD,” assumes full compliance with recommendations based on the test cutoff. In the context of the current study, however, not all students below the 600 cutoff score followed the EASE assignment.⁶

Accordingly, the average probability of enrollment in EASE is less than one below the cutoff. To deal with potential bias associated with noncompliance, we followed existing literature for a “fuzzy RD” design (see Imbens & Lemieux, 2008, for a detailed discussion), where we used EASE assignment as an instrumental variable for actual participation in EASE and employed a two-stage least squares strategy to provide a consistent estimate of EASE on student outcomes:

\begin{array}{l} {Enroll}_{i} = γ_{0} + γ_{1} ({Below}_{i}) + γ_{2} ({ScoreDistance}_{i}) + \\ γ_{3} ({Below}_{i} * {ScoreDistance}_{i}) \\ + X_{i} + μ_{i} \end{array}

(2)

\begin{array}{l} Y_{i} = δ_{0} + δ_{1} ({\hat{Enroll}}_{i}) + δ_{2} ({ScoreDistance}_{i}) + \\ δ_{3} ({Below}_{i} * {ScoreDistance}_{i}) + X_{i} + ε_{i} \end{array}

(3)

Enroll indicates enrollment in EASE. Equation (2) represents the first stage, where we use a linear probability model to predict EASE enrollment as a function of EASE assignment. Equation (3) then estimates the local average treatment effect of the predicted probability of enrollment on student outcome measures in the second stage. δ₁ captures the impacts of participating in EASE on student outcomes.

We conduct several tests to ascertain the validity of the RD design. Detailed explanations of these validity checks and their results are presented in Appendix B. Finally, it is important to note that RD design provides estimates of the “local average treatment effects” (Imbens & Angrist, 1994). This means that the conclusions drawn from our analysis only speak to the subpopulation of students around the cutoff points. We discuss the implications of this limitation in the Discussion section.

Results

Impacts of EASE on Academic Outcomes

Academic outcomes include three primary measures: (i) course grades in required first-year biology courses (Biology 93 and Biology 94), (ii) cumulative GPA at the end of the freshmen year, and (iii) whether the student remained as a bio sci major at the end of the first year. We first begin by examining graphical plots of these outcomes by SAT score for visual evidence of discontinuities at the cutoff, shown in Figure 1. Overall, the clearest pattern coming out of these graphs is that students do seem to benefit from EASE: There is some hint of possible discontinuity on the course grade from Biology 93 and first-year retention as a bio sci major in favor of students assigned to the EASE program. The discontinuity becomes more pronounced when it comes to course grade from Biology 94 and cumulative first-year GPA, as students right below the 600 cutoff clearly outperformed their counterparts who scored right above 600 on SAT math and were hence ineligible to participate in EASE.

Figure 1.

Academic outcomes by SAT math score.

The positive impacts of EASE on students’ academic outcomes shown in the plots are further supported by statistical estimates. Panel A in Table 2 reports both the EASE assignment and EASE participation on the previously mentioned academic measures within a ±60-point bandwidth. For each measure, the first row reports the ITT estimate, which measures the average differences between individuals above and those below the cutoff score, controlling for baseline characteristics; the second row then reports the two-stage least squares estimate using the fuzzy RD design. To test the sensitivity of our results to different bandwidths, we further present results based on a narrower bandwidth (±50 points) in Appendix C, Table C1 and a wider bandwidth (±70 points) in Appendix C, Table C2.

Table 2

Impacts of EASE Program on Academic and Nonacademic Outcomes (Bandwidth ± 60 Points)

	Biology 93 Course Grade	Biology 94 Course Grade	Year 1 GPA	Retained
Panel A. Academic outcome measures
Intent-to-treat estimates	0.34**	0.44**	0.33***	0.064**
Intent-to-treat estimates	(0.17)	(0.19)	(0.12)	(0.028)
Instrumental variable estimates	0.64**	0.77**	0.63***	0.13**
Instrumental variable estimates	(0.33)	(0.32)	(0.23)	(0.057)
N	393	375	376	393
	Sense of Belonging	AI	ASC	Interest
Panel B. Nonacademic outcome measures
Intent-to-treat estimates	0.55***	0.34	−0.040	0.17
Intent-to-treat estimates	(0.19)	(0.22)	(0.18)	(0.19)
Instrumental variable estimates	1.08**	0.65	−0.077	0.33
Instrumental variable estimates	(0.42)	(0.44)	(0.33)	(0.38)
N	359	376	359	356

Note. Each cell represents a separate regression within a ±60-point band width. Each analysis includes the following covariates: SAT math score, an interaction term between SAT math score and whether the student was assigned to EASE, a gender dummy variable, race dummy variables, a dummy variable for low-income status, a dummy variable for first-generation status, SAT read score, and Wave 1 survey measure (for noncognitive measures). Nonacademic outcome measures are standardized to have a mean of 0 and standard deviation of 1. AI = academic integration; ASC = academic and social concerns. For ASC, higher values indicate more concern. Robust standard errors are reported in parentheses.

The results from the ITT analyses indicate assignment to EASE has decent, statistically significant positive effects on all academic outcome measures. These estimates remain significant after using fuzzy RD to address noncompliance (Row 2). Focusing on the two-stage least squares estimates in the second row, EASE participation on average increases course grade from Bio 93 by 0.64 grade points on a 0–4 grading scale, which represents more than two letter grades higher, such as from a B– to a B+. Considering that the average grade from Bio 93 has a mean of 3.05 and a standard deviation of 0.95, an effect of 0.64 grade points represents an increase of 0.67 standard deviations. The effect size is even larger for Bio 94, where EASE participation increases course grades by 0.77 grade points, or 0.86 standard deviations. The effect is also reflected in cumulative first-year GPA, with a fairly similar effect size of 0.63 grade points, or 1.05 standard deviations. Finally, EASE participation increases the first-year retention rate as a bio sci major by 13 percentage points.

Impacts of EASE on Social-Psychological Outcomes

Based on students’ responses to the pre- and postsurveys during the fall quarter, we further examine the impacts of EASE on students’ social-psychological outcomes on four measures: (1) sense of belonging to bio sci, (2) AI, (3) ASC, and (4) interest in biology. Figure 2 shows graphical plots of these outcomes by SAT score for visual evidence of discontinuities at the cutoff. Compared to academic outcome measures, all the social-psychological measures are noticeably noisier, with substantially larger variations among individuals. Despite the larger variations, however, there is still a fairly clear discontinuity at the cutoff on sense of belonging and some hint of a possible discontinuity on AI and interest in favor of students assigned to the EASE program.

Figure 2.

Nonacademic outcomes by SAT math score.

Panel B in Table 2 further reports statistical estimates from a local linear regression with a bandwidth of ±60 points. The pattern of results is fairly consistent across the four social-psychological measures, where assignment to EASE is associated with higher levels of sense of belonging, academic integration, and interest in biology, and with a lower level of academic concerns, although only the effect on sense of belonging reaches statistical significance at the 5% level.⁷

Focusing on the two-stage least squares estimates, EASE participation on average increases students’ sense of belonging by more than 1 standard deviation.

Mechanisms: Remediation vs. Social-Psychological Component

Having established the causal impacts of EASE on students’ academic outcomes and social integration, we further examine which component of the EASE program benefits students the most. It is worth noting that the impact of the academic component and the social-psychological component of the EASE program on student outcomes are likely to be amplified and sustained when implemented together. Hence, our exploration to tease out the impact of one component from that of the other only intends to provide suggestive insights rather than to make a defining conclusion.

Specifically, we exploit the fact that prior to the 2015–16 academic year when the EASE program was first piloted, the School of Biological Sciences used the same math SAT threshold to assign students to developmental coursework but did not provide any activity or program to engage students social-psychologically. This setting allows us to use a fuzzy RD design to examine the impact of being assigned to developmental education alone on student outcomes based on data from the 2014–15 cohort.

Descriptive information about the 2014–15 cohort, presented in Appendix E, Table E1, demonstrates similarities between this earlier cohort and the 2016–17 cohort, except the proportion of URM students was noticeable lower for the 2014–15 cohort (20% vs. 34%). When examining significant differences between students above and below the math SAT cutoff score of 600, the 2014–15 cohort exhibited similar demographic differences, as compared to the 2016 cohort.

Figure 3 shows graphical plots of Bio 93 and 94 grades by SAT score for the 2014 cohort, and Table 3 further presents statistical estimates based on a local linear regression with a bandwidth of ±60 points. Not only do we find little evidence of assignment to developmental education on performance in either of the two courses, but all the coefficients are also negative. These analyses on the 2014 cohort therefore provide suggestive evidence that academic remediation, at least alone, is not sufficient to support students; instead, a well-designed and implemented social-psychological component is critical to the effectiveness of a STEM learning communities program.

Figure 3.

Academic outcomes by SAT math score (2014 cohort).

Table 3

2014 Cohort: Impacts of Assignment to Developmental Education on Academic Outcomes (Bandwidth ± 60 Points)

	Biology 93 Course Grade		Biology 94 Course Grade
	(1)	(2)	(1)	(2)
Intent-to-treat estimates	−0.062	−0.33	−0.23	−0.38
Intent-to-treat estimates	(0.19)	(0.33)	(0.16)	(0.28)
N	408	408	357	357

Note. Each cell represents a separate regression within a ±60-point bandwidth. Each analysis includes the following covariates: SAT math score, an interaction term between SAT-math score and whether the student was above or below the SAT math cutoff score, a gender dummy variable, race dummy variables, and SAT read score. Model 2 includes the following additional controls: squared term of the SAT math score and an interaction term between the squared term of the SAT math score and whether the student was above or below the SAT math cutoff score. Robust standard errors are reported in parentheses.

Discussion and Conclusion

Learning communities have been nominated as a promising way to engage students both academically and social-psychologically. Yet there is limited experimental and quasi-experimental evidence on the causal impacts of learning communities on student academic outcomes and social-psychological engagement in the 4-year setting and within specific fields of study. In this paper, we estimated the effects of a first-year STEM learning communities program on a variety of student outcomes based on a quasi-experimental design. We view our main contribution as threefold.

First, we provide detailed description about how a successful field-specific learning communities program was administered and implemented. The EASE program combined several of the active ingredients of interventions that are consistent with the theoretical frameworks about college retention and persistence. Other departments or colleges that intend to better support and engage students may wish to draw on the key components of EASE in building their own learning communities program. Second, we use a rigorous RD design to estimate the effects of EASE on a variety of student performance measures as well as noncognitive measures of social belonging and academic integration. We conduct several validity and robustness checks that affirm the strength of our method. Our results indicate that participation in the EASE learning communities program significantly increased the grades in core college-level courses for students around the cutoff, improved their first-year major retention rate, and boosted their sense of belonging. To our knowledge, this is the first study that is able to provide a “causal” estimate of the benefits of a learning communities program offered to students in a particular field of study.

Finally, we conduct exploratory analyses to provide insight into the specific component of the EASE program that benefits students the most. Although our study cannot definitively “prove” which ingredient is driving the impact of EASE, we find little positive influence of remedial education alone on students’ learning outcomes, which echoes the existing studies evaluating developmental coursework that fails to find consistent evidence for the benefits of receiving college remediation (Calcagno & Long, 2008; Martorell & McFarlin, 2011; Scott-Clayton & Rodriguez, 2015; Xu, 2016; Xu & Dadgar, 2018). While we cannot directly attribute the success of the EASE program completely to the social-psychological component, our explorations indicate that social factors such as a sense of belonging may have been especially important for students in this study’s learning community program that enrolls a substantial proportion of URM students. Indeed, previous work has found that opportunities for academic engagement with others is especially useful for boosting URM students’ sense of belonging (Strayhorn, 2012), and the association between sense of belonging and academic outcomes has been found to be significantly stronger among students of color compared to White students (Murphy & Zirkel, 2015). Therefore, cohorting students across all the first-year courses, a key component of learning communities, may have been particularly helpful for integrating URM students socially, which, in turn, would further boost their academic integration.

While the EASE program provides a successful model to support academically underprepared populations in a specific field of study, the findings from this study should be interpreted in light of several important caveats when it comes to policy and practice. First, as discussed earlier, with the RD analysis, our results only speak to students around the threshold and therefore may not generalize to individuals far away from the cutoff. Future studies may wish to use random controlled trials to identify possible heterogeneous effects of EASE by student academic preparation.

In addition, we have a relatively short follow-up window in assessing the impacts of EASE. Some of the short-term effects observed in this study may either become stronger or fade out as one’s academic career evolves. In particular, it would be important to examine whether the positive impacts of EASE on major retention in the first year still hold by the end of the second year, when students are required to decide their major.

Finally, if the benefits of EASE are primarily due to the social-psychological component rather than developmental education, it implies that the remedial courses may not be providing the optimal content, at least not as currently practiced. During the past few years, there has been an ongoing national effort in reforming traditional developmental education. Some of the most popular approaches include using multiple measures for more accurate placement into developmental education, curriculum pathways to better align developmental coursework with college-level course content, acceleration to substantially shorten the duration of the developmental education or to provide it during the summer before college enrollment, and modularization, where students only took the modules in which the diagnostic placement test had indicated a need for improvement. A question for future research is thus how to provide developmental education more effectively in a learning communities programs such as EASE and what type of content would be optimal.

Footnotes

Appendix

Appendix B : Details About the RD Design

Appendix C : Robustness Checks

Table C2

Impacts of EASE Program on Academic and Nonacademic Outcomes (Bandwidth ±70 Points)

Academic Outcome Measures
	Biology 93Course Grade		Biology 94Course Grade		Year 1 GPA		Retained
	(1)	(2)	(1)	(2)	(1)	(2)	(1)	(2)
Intent-to-treat estimates	0.29*	0.31	0.45***	0.31	0.32***	0.27	0.04	0.12***
Intent-to-treat estimates	(0.15)	(0.24)	(0.17)	(0.27)	(0.11)	(0.17)	(0.03)	(0.05)
Instrumental variable estimates	0.54*	0.67	0.77***	0.58	0.61***	0.58	0.075	0.26**
Instrumental variable estimates	(0.29)	(0.54)	(0.29)	(0.50)	(0.20)	(0.37)	(0.050)	(0.12)
N	467	467	448	448	447	447	467	467
	Nonacademic Outcome Measures
	Sense of Belonging		AI		ASC		Interest
	(1)	(2)	(1)	(2)	(1)	(2)	(1)	(2)
Intent-to-treat estimates	0.52***	0.61**	0.39**	0.11	−0.022	−0.087	0.13	0.15
Intent-to-treat estimates	(0.17)	(0.28)	(0.20)	(0.33)	(0.16)	(0.27)	(0.17)	(0.28)
Instrumental variable estimates	1.02***	1.30*	0.74*	0.23	−0.042	−0.18	0.25	0.31
Instrumental variable estimates	(0.37)	(0.72)	(0.39)	(0.72)	(0.30)	(0.54)	(0.33)	(0.58)
N	427	427	448	448	428	428	425	425

Note. Each cell represents a separate regression within a ±70-point bandwidth. Each analysis includes the following covariates: SAT math score, an interaction term between SAT math score and whether the student was assigned to EASE, a gender dummy variable, race dummy variables, a dummy variable for low-income status, a dummy variable for first-generation status, SAT read score, and Wave 1 survey measure (for noncognitive measures). Model 2 includes the following additional controls: squared term of the SAT math score and an interaction term between the squared term of the SAT math score and EASE assignment. Nonacademic outcome measures are standardized to have a mean of 0 and standard deviation of 1. AI = academic integration; ASC = academic and social concerns. For ASC, higher values indicate more concern. Robust standard errors are reported in parentheses.

Appendix D : Exploration of Alternative Explanations

We have conducted a series of explorations to rule out the possibility that the estimated impacts were driven by alternative explanations other than EASE participation. First, although EASE has shown promising impacts on students’ course grades, one concern regarding this outcome measure is that EASE students might be assigned to different sections of the course taught by different instructors. Indeed, for Biology 93, three sections were offered and EASE students were enrolled in one of three sections. For Biology 94, six sections were offered, taught by four instructors. EASE students were assigned to two of those six sections. As a result, the positive impacts identified might be completely driven by the benefits of having more supportive and engaging instructors rather than participating in the EASE program. This is unlikely to be the case, considering that all the analyses on non-course-specific outcomes, such as first-year major persistence and sense of belongingness, all yielded consistent results with course-grade outcomes. People might argue that the “sense of belonging” measure may also partly be influenced by course instructors. As shown in Appendix A, several items of our “sense of belonging” construct asked students about their interactions with the faculty. We therefore conducted a robustness check where we took out all the faculty-related items from the sense of belongingness instrument and the results remain the same.

Second, we are concerned that the EASE assignment might influence students through channels other than the EASE participation, which would violate the exclusion assumption of the two-stage least squares regression. For example, if the department also provides students below the SAT threshold support in addition to the EASE program, we would not be able to tease out the impacts of participating in EASE from the additional support made available to the lower performing students. Although the school has confirmed that no additional support was provided to students, we further conducted a falsification test on outcome measures that are not expected to be influenced by the EASE program on a theoretical basis.

Specifically, we examined the impacts of EASE on students’ growth mindset. Growth mindset measured students’ theories of intelligence, assessing the extent to which they believed that intelligence was an immutable trait as opposed to a malleable feature of themselves (Dweck & Henderson, 1988). While mindset has been found to predict motivation, efforts, and achievement, the active ingredients of the EASE learning communities program that facilitates social-psychological integration into the specific biology communities at UCI are not designed to influence one’s general mindset. Indeed, as shown in the graphical demonstration (Appendix D, Figure D1), EASE does not have any impact on students’ growth mindset. Appendix D, Table D1 reports statistical estimates from a local linear regression with a bandwidth of ±60 points quantifying this result. All estimates are statistically insignificant, providing further evidence in support of the EASE program.

Appendix E : Descriptive Statistics for the 2014 Cohort

Table E1

2014 Cohort: Descriptive Statistics of the Full Sample of Students and by Development Education Eligibility

			Developmental Education Eligibility
	Full Sample		Ineligible: SAT Math≥600	Eligible: SAT Math<600
	M	SD	M	M	t Test of Difference
Female	66.0	(47.4)	62.0	72.9	−3.085
White	14.7	(35.5)	13.6	16.8	−1.199
URM	19.7	(39.8)	12.1	32.8	−7.094
Asian	65.5	(47.6)	74.4	50.4	6.847
SAT read score	565.7	(79.66)	585.5	531.4	9.533
SAT math score	625.4	(67.10)	665.0	556.8	34.003
N	761		483	278

Notes

ORCID iD

Peter McPartlan

Authors

DI XU, PhD, Economics and Education, is an assistant professor of educational policy and social context at the University of California, Irvine, 3200 Education Building, Irvine, CA 92697; "> dix3@uci.edu . Her research examines the impacts of educational programs and policies on student academic performance, persistence, and degree completion at the postsecondary education level, with a particular focus on students from disadvantaged backgrounds.

SABRINA SOLANKI is a doctoral candidate at the University of California, Irvine; "> ssolanki@uci.edu . Her research focuses on higher education policy, teacher effectiveness, STEM education, and the evaluation of education interventions.

PETER McPARTLAN is a doctoral candidate at University of California, Irvine, 3200 Education Building, Irvine, CA; "> pmcpartl@uci.edu . His research focuses on improving adolescent students’ motivation by leveraging the social processes that motivate them, such as sense of belonging.

BRIAN SATO, PhD, is an associate teaching professor in molecular biology and biochemistry at the University of California Irvine, 2238 McGaugh Hall MC3900, Irvine, CA 92697; "> bsato@uci.edu . His research focuses on improving STEM student outcomes at the classroom, program, and institutional levels.

References

Adams

W. K.

Perkins

K. K.

Podolefsky

N. S.

Dubson

Finkelstein

N. D.

Wieman

C. E.

(2006). New instrument for measuring student beliefs about physics and learning physics: The Colorado Learning Attitudes about Science Survey. Physical Review Special Topics–Physics Education Research, 2, 1–14. http://doi.org/10.1103/PhysRevSTPER.2.010101.

Astin

A. W.

(1984). Student involvement: A developmental theory for higher education. Journal of College Student Personnel, 25, 297–308.

Bergen

Milem

(1999). The role of student involvement and perceptions of integration in a causal model of student persistence. Research in Higher Education, 40(6), 641–664.

Bettinger

Baker

(2014). The effects of student coaching: An evaluation of a randomized experiment in student advising. Education Evaluation and Policy Analysis, 36(1), 3–19.

Braxton

McClendon

(2001). The fostering of social integration and retention through instructional practice. Journal of College Student Retention: Research, Theory, and Practice, 3, 57–71.

Calcagno

Long

(2008). The impact of postsecondary remediation using a regression discontinuity design: Addressing endogenous sorting and noncompliance (NBER Working Paper No. 14194). Cambridge, MA: National Bureau of Economic Research.

Carini

Kuh

Klein

(2006). Student engagement and student learning: Testing the linkages. Research in Higher Education, 47, 1–32.

Cattaneo

M. D.

Jansson

(2017). Simple local polynomial density estimators. Working paper, University of Michigan.

Center for Institutional Data Exchange and Analysis (C-IDEA). (2000). 1999-2000 SMET retention report. Norman, OK: University of Oklahoma.

10.

Chen

(2013). STEM attrition: College students’ paths into and out of STEM fields. Retrieved from http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2014001rev

11.

Cohen

G. L.

Garcia

(2005). “I am us”: Negative stereotypes as collective threats. Journal of Personality and Social Psychology, 89, 566–582.

12.

Dagley

Georgiopoulos

Reece

Young

(2015). Increasing retention and graduation rates through a STEM learning community. Journal of College Student Retention: Research, Theory & Practice, 18, 167–182.

13.

Darling

Molina

Sanders

Lee

Zhao

(2008). Belonging and achieving: The role of identity integration. Social Psychological Perspectives, Advances in Motivation and Achievement, 15, 241–273.

14.

DiMaria

(2006). Keeping our kids engaged, at-risk kids in college. Education Digest, 72(2), 52–57.

15.

Dweck

C. S.

Henderson

V. L.

(1988). Theories of intelligence: Background and measures. Unpublished manuscript.

16.

Eccles

J. S.

Adler

T. F.

Futterman

Goff

S. B.

Kaczala

C. M.

Meece

J. L.

Midgley

(1983). Expectancies, values, and academic behavior. In Spence

J. T.

(Ed.), Achievement and achievement motives: Psychological and sociological approaches (pp. 75–146). San Francisco, CA: W.H. Freeman.

17.

Engstrom

Tinto

(2007). Pathways to student success: The impact of learning communities on the success of academically under-prepared college students. Retrieved from the Lumina Foundation for Education website: https://www.luminafoundation.org

18.

Flynn

(2014). Baccalaureate attainment of college students at 4-year institutions as a function of student engagement behaviors: Social and academic student engagement behaviors matter. Research in Higher Education, 55(5), 467–493. http://doi.org/10.1007/s11162-013-9321-8

19.

Freeman

T. M.

Anderman

L. H.

Jensen

J. M.

(2007). Sense of belonging in college freshmen at the classroom and campus levels. The Journal of Experimental Education, 75, 203–220. http://doi.org/10.3200/JEXE.75.3.203-220

20.

Harackiewicz

J. M.

Canning

E. A.

Tibbetts

Priniski

S. J.

Hyde

J. S.

(2015, November 2). Closing achievement gaps with a utility-value intervention: Disentangling race and social class. Journal of Personality and Social Psychology. Advance online publication.

21.

Heatherton

T. F.

Polivy

(1991). Development and validation of a Scale for Measuring State Self-Esteem. Journal of Personality and Social Psychology, 60(6), 895–910.

22.

Higher Education Research Institute (HERI). (2010). Degrees of success: Bachelor’s degree completion rates among initial STEM majors. Los Angeles, CA: University of California, Los Angeles.

23.

Hoffman

Richmond

Morrow

Salomone

(2002). Investigating “sense of belonging” in first-year college students. Journal of College Student Retention, 4, 227–257.

24.

Huang

Taddese

Walter

(2000). Entry and persistence of women and minorities in college science and engineering education (No. NCES 2000601). Washington, DC: National Center for Education Statistics.

25.

Hurtado

Carter

D. F.

(1997). Effects of college transition and perceptions of the campus racial climate on Latino college students’ sense of belonging. Sociology of Education, 70, 324–345.

26.

Imbens

G. W.

Angrist

J. D.

(1994). Identification and estimation of local average treatment effects. Econometrics, 61, 467–476.

27.

Imbens

G. W.

Lemieux

(2008). Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142(2), 615–635.

28.

Kuh

G. D.

Kinzie

Schuh

J. H.

Whitt

E. J.

(2005). Student success in college: Creating conditions that matter. San Francisco, CA: Jossey-Bass.

29.

Lee

(2008). Randomized experiments from non-random selection in US house elections. Journal of Econometrics, 142, 675–697.

30.

Lee

Lemieux

(2010). Regression discontinuity designs in economics. Journal of Economic Literature, 48, 281–355

31.

Levin

Calcagno

J. C.

(2008). Developmental coursework in the community college: An evaluator’s perspective. Community College Review, 35, 181–207.

32.

Martorell

McFarlin

(2011). Help or hindrance? The effects of college remediation on academic and labor market outcomes. Review of Economics and Statistics, 93, 436–454.

33.

Master

Cheryan

Meltzoff

A. N.

(2016). Computing whether she belongs: Stereotypes undermine girls’ interest and sense of belonging in computer science. Journal of Educational Psychology, 108(3), 424–437. http://doi.org/10.1037/edu0000061

34.

MDRC. (2012). The effects of learning communities for students in developmental education report. Retrieved from http://www.mdrc.org/publication/effects-learning-communities-students-developmen tal-education

35.

Murphy

M. C.

Zirkel

(2015). Race and belonging in school: How anticipated and experienced belonging affect choice, persistence, and performance. Teachers College Record, 117, 1–40.

36.

National Center for Education Statistics. (2013). Digest of education statistics, 2013. Table 306.40. U.S. Department of Education. Washington, DC: Institute of Education Sciences.

37.

Ostrove

J. M.

Long

S. M.

(2007). Social class and belonging: Implications for college adjustment. The Review of Higher Education, 30(4), 363–389. http://doi.org/10.1353/rhe.2007.0028

38.

President’s Council of Advisors on Science and Technology (PCAST). (2012). Engage to excel: Producing one million additional college graduates with degrees in science, technology, engineering, and mathematics. Washington, DC: Author.

39.

Robinson

(1996). Underprepared students (ED433876). Washington, DC: Educational Resources Information Center.

40.

Scott-Clayton

Rodriguez

(2015). Development, discouragement, or diversion? New evidence on the effects of college remediation policy. Education Finance and Policy, 10, 4–45.

41.

Scrivener

Bloom

LeBlanc

Paxson

Rouse

Sommo

(2008). A good start: Two-year effects of a freshmen learning community program at Kingsborough Community College. Retrieved from http://www.mdrc.org/publication/effects-learning-communities-students-developmental-education

42.

Seymour

Hewitt

N. M.

(1997). Taking about leaving: Why undergraduates leave the sciences. Boulder, CO: Westview Press.

43.

Shapiro

N. S.

Levine

(Eds.). (1999). Creating learning communities: A practical guide to winning support, organizing for change, and implementing programs. San Francisco, CA: Jossey-Bass.

44.

Sherman

D. K.

Bunyan

D. P.

Creswell

J. D.

Jaremka

L. M.

(2009). Psychological vulnerability and stress: The effects of self-affirmation on sympathetic nervous system responses to naturalistic stressors. Health Psychology, 28(5), 554–562.

45.

Smith

B. L.

MacGregor

Matthews

R. S.

Gabelnick

(2004). Learning communities: Reforming undergraduate education. San Francisco, CA: Jossey-Bass.

46.

Strayhorn

T. L.

(2008). Sentido de pertenencia: A hierarchical analysis predicting sense of belonging among Latino college students. Journal of Hispanic Higher Education, 7(4), 301–320.

47.

Strayhorn

T. L.

(2012). College students’ sense of belonging: A key to educational success for all students. New York, NY: Routledge.

48.

Taylor

Moore

MacGregor

Lindblad

(2003). Learning community research and assessment: What we know now (National Learning Communities Monograph Series). Olympia, WA: Washington Center for Improving the Quality of Undergraduate Education.

49.

Tinto

(1975). Dropout from higher education: A theoretical synthesis of recent research. Review of Educational Research, 45, 89–125. http://doi.org/10.3102/00346543045001089

50.

Tinto

(2017). Through the eyes of students. Journal of College Student Retention: Research, Theory & Practice, 19(3), 254–269.

51.

Tinto

Goodsell

(1993). Freshman interest groups and the first year experience: Constructing student communities in a large university. Paper presented at the annual meeting of the College Reading and Learning Association, Kansas City, MO.

52.

Tinto

Russo

(1994). Coordinated studies programs: Their effect on student involvement at a community college. Community College Review, 22, 16–25.

53.

Tovar

Simon

M. A.

(2010). Factorial structure and invariance analysis of the sense of belonging scales. Measurement and Evaluation in Counseling and Development, 43(3), 199–217. http://doi.org/10.1177/0748175610384811

54.

Warner

(2013). Applied statistics: From bivariate through multivariate techniques. Thousand Oaks, CA: SAGE Publications.

55.

(2016). Assistance or obstacle? The impact of different levels of English developmental education on underprepared students in community colleges. Educational Researcher, 45, 496–507. http://doi.org/10.3102/0013189X16683401

56.

Dadgar

(2018). How effective are community college remedial math courses for students with the lowest math skills? Community College Review, 46, 62–81.

57.

Zhao

C. M.

Kuh

G. D.

(2004). Adding value: Learning communities and student engagement. Research in Higher Education, 45, 115–138.