Community college transfer and degree attainment

... Chapter Community College Transfer and Baccalaureate Attainment Chapter 20 Institutional Variation in Transfer Rates Among California Community Colleges Chapter 78 Reverse Transfer and the Community. .. attending a community college Reverse transfers improve their grade point averages upon transitioning to a community college, 20% transfer back to a four-year college and earn a bachelors degree, and. .. clear that a large proportion of community college entrants plan to use the community college transfer pathway to attain a baccalaureate degree The community college transfer function has taken on

Community College Transfer and Degree Attainment By DEMETRA MARIANNE KALOGRIDES B.S. (Santa Clara University) 2003 M. A. (University of California, Davis) 2005 DISSERTATION Submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in Sociology in the Office of Graduate Studies of the UNIVERSITY OF CALIFORNIA DAVIS Approved: ^ j^Uv4si^JiM~ ^f^ou^nsJi^_ Committee in Charge 2008 i UMI Number: 3329625 Copyright 2008 by Kalogrides, Demetra Marianne All rights reserved. INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. ® UMI UMI Microform 3329625 Copyright 2008 by ProQuest LLC. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 E. Eisenhower Parkway PO Box 1346 Ann Arbor, Ml 48106-1346 Demetra Marianne Kalogrides June 2008 Sociology Community College Transfer and Degree Attainment Abstract Supporters of community colleges have argued that these institutions offer upward mobility by opening the door to higher education for disadvantaged students and providing them with additional job training, remedial education, and preparation for transfer to four-year schools. Critics, on the other hand, maintain that rather than democratizing access to higher education, community colleges actually divert disadvantaged students away from four-year colleges and from attaining bachelor's degrees. Given their potential to undermine stratification, understanding whether community colleges have lived up to their democratizing ideal is imperative. In this dissertation I study pathways to and from community colleges both nationally and in California. I find that the community college is an effective route to a baccalaureate degree for students who transfer. Transfer rates remain low, however, and although they vary considerably among different community colleges, it remains unclear which policies or practices distinguish colleges that are relatively more successful in promoting transfer among their students. Disadvantaged students and those with weak academic backgrounds often make their way to community colleges after struggling in four-year institutions. Although these reverse transfers do not fare as well as students with exclusive four-year college enrollment, they do appear to have more favorable outcomes than otherwise similar students who drop out of postsecondary school altogether after initially enrolling in a four-year school. ii Table of Contents Abstract ii Table of Contents iii List of Tables iv List of Figures vi Chapter 1 Introduction 1 Chapter 2 Community College Transfer and Baccalaureate Attainment Chapter 3 20 Institutional Variation in Transfer Rates Among California Community Colleges Chapter 4 78 Reverse Transfer and the Community College as a Second Chance Institution 139 iii List of Tables Table 2.1 Descriptives by First College Type 65 Table 2.2 First Stage Instrumental Variables Models 67 Table 2.3 First Stage Propensity Score Models 68 Table 2.4 Discrete-Time Hazard Model of Transfer to a Four-Year College Among Community College Entrants. 69 Table 2.5 Mean Number of Terms Enrolled to Earn Credits 71 Table 2.6 Discrete-Time Hazard Model of Attrition from a Four-Year Institution Table 2.7 72 Comparison of Community College Transfer Coefficients Across Alterative Approaches to Controlling for Selection Table 3.1 74 Means of Student-Level Measures by Whether Students Transferred 123 Table 3.2 Student-Level Logit Model Predicting Transfer 125 Table 3.3 Descriptives by Clusters of Community Colleges 128 Table 3.4 Comparison of Means for Community College Characteristics for High and Low Ranking Schools 130 Table 3.5 OLS Regression Coefficients for Community College Fixed Effects 132 Table A3.1 Means of Independent Variables by Transfer Status & Quintiles of Predicted Probability of Transfer 133 Table A3.2 Transfer Rates Using Alternative Definitions of Those At Risk 134 Table 4.1 Descriptive Statistics by Postsecondary Enrollment Path 183 Table 4.2 First Stage Propensity Score Models: Logistic Regression of First College iv Type Table 4.3a 185 Multinomial Logistic Regression of Transitions from First Four-Year College Table 4.3b 187 Multinomial Logistic Regression of Transitions from First Four-Year College: Reverse Transfer vs. Dropout Table 4.4 Regression of Income and Credits Earned on Types of Transitions from Initial Four-Year College Table 4.5 193 Status of Respondents Excluded from Income Analysis by Postsecondary Enrollment Pathway Table A4.1 195 Marginal Effects from Multinomial Logistic Regression Model: Reverse Transfer vs. BA from First Four-Year School Table A4.2 197 Marginal Effects from Multinomial Logistic Regression Model: Lateral Transfer vs. BA from First Four-Year School Table A4.4 195 Marginal Effects from Multinomial Logistic Regression Model: Dropout vs. BA from First Four-Year School Table A4.3 190 198 Marginal Effects from Multinomial Logistic Regression Model: Reverse Transfer vs. Dropout from First Four-Year School v 199 List of Figures Figure 2.1 Relationship between College Distance and the Predicted Probability of Community College Entrance Figure 2.2 75 Relationship between Differences in Tuition between Public Four-Year Institutions and Community Colleges and the Predicted Probability of Community College Entrance Figure 2.3 Predicted Probability of Transfer by Number of Credits Earned in Community College Figure 2.4 76 Predicted Probability of Attrition from a Four-Year Institution, by Number of Credits Earned and Transfer Status Figure 2.6 77 Predicted Probability of Attrition From a Four-Year Institution, by Years of Enrollment and Transfer Status Figure 3.1 76 Differences in Academic Credits Earned, by Year of Enrollment Figure 2.5 75 77 Proportion of Variance Explained by Various Cluster Solutions 135 Figure 3.2 Kernel Density Plot of Observed Institutional Transfer Rates 135 Figure 3.3 Distributions of Proportion of Students each College Enrolled in each Quintile of the Predicted Probability of Transfer Figure 3.4 Variation in Observed Transfer Rates within Quintiles of The Predicted Probability of Transfer Figure 3.5 136 136 Kernel Density Plots of Observed Transfer Rates by Community College Cluster 137 vi Figure 3.6 Coefficients and Confidence Intervals for the 20 Schools that Have the Most Positive Impact on Transfer, Net of Student Attributes Figure 3.7 137 Coefficients and Confidence Intervals for the 20 Schools that Have the Most Negative Impact on Transfer, Net of Student Attributes 138 vn 1 CHAPTER 1: Introduction 2 Rates of college enrollment have risen substantially over the past 25 years. In 1980 about 47% of students entered postsecondary school immediately following high school graduation, whereas 67% of students did so in 2005 (National Center for Education Statistics 2007). Community colleges have accommodated a large share of the increase in postsecondary enrollment (National Center for Education Statistics 2007). More than half of the students who enter college shortly after high school graduation will begin their postsecondary education at a community college (Adelman 2005) and in some states, such as California, this figure is close to 70 percent (National Center for Education Statistics 2005). Community colleges' low cost and close geographic proximity to most potential students open the door to higher education for populations that have traditionally had limited access to college such as racial/ethnic minorities and students from disadvantaged socioeconomic backgrounds (Kane and Rouse 1999). In their idealized form, therefore, community colleges have the potential to function as a democratizing force in American society by extending opportunities for access to education and job training that would not otherwise exist. Whether community colleges have lived up to this democratizing ideal, however, has been a source of contention among researchers. Critics maintain that rather than democratizing access to higher education, community colleges actually divert disadvantaged students away from four-year colleges and from attaining bachelor's degrees. Noting high rates of attrition from community colleges, low rates of transfer to four-year colleges, and high rates of attrition among students who do transfer, critics contend that students who attend community colleges attain less education than they 3 would have had they attended a four-year college. Understanding the role of community colleges in social stratification is an important policy concern yet questions still remain about the efficacy of community colleges in improving the educational and economic outcomes of their students. In this project I study pathways to and from community colleges both nationally and in California. I study student and institution-level predictors of transfer to four-year colleges, success of transfer students in obtaining baccalaureate degrees, and the predictors and outcomes of "reverse transfers"- students who attend community colleges after attending a four-year school without obtaining a degree. I use alternative and more robust methods for addressing self-selection and for identifying appropriate counterfactuals than have been used in prior work. Taken as a whole, the findings from this research both confirm prior results and shed new light on the impact community colleges have on the educational outcomes of their students. Literature Review Community colleges offer a wide variety of degree programs and have diverse missions that range from the provision of transfer education to the provision of adult, remedial, and vocational education (Dougherty 2002). The ambitions of community college attendees are diverse and some students enroll without the intention of transferring to a four-year college and attaining a bachelor's degree (Alfonso, Bailey, and Scott 2005). However, despite the diversity of the community college mission, transfer education remains one of the most important services provided by these institutions, especially for traditional aged students. Among members of the class of 1992 who entered college shortly after high school graduation, more than three quarters of those whose initial postsecondary school was a community college indicated that they aspired to transfer and earn a bachelor's degree (author's calculations based on National Education Longitudinal Study). Although the validity of students' stated aspirations may be questionable (Reynolds, Stewart, MacDonald, and Sischo 2006), it is clear that a large proportion of community college entrants plan to use the community college transfer pathway to attain a baccalaureate degree. The community college transfer function has taken on increased importance in recent years due to rising four-year college tuition, stagnating need-based student aid, declining incomes for disadvantaged families, and the elimination of remedial education courses in many public four-year colleges (Dougherty 2002; Dougherty and Kienzl 2006; Richardson 2005; Sturrock 2003; Wellman 2002). The transfer function is also imperative given the rising importance of a baccalaureate degree in today's economy. Changes in the domestic and international economy have rendered a college education more important than ever before. The college wage premium rose dramatically in the 1980s (Card and Lemieux 2001; Katz and Autor 1999) and is currently as high as at any other point in U.S. history (Goldin and Katz 2007). Among men aged 31-35, for example, the wage differential between high school and college graduates grew from 18 percent in 1979-81 to 41 percent in 1989-91 (Card and Lemieux 2001). As of 1999, full-time yearround workers with a bachelor's degree earn nearly twice as much as those with only a high school diploma and those with an advanced degree earned nearly three times as much (Cheeseman Day and Newburger 2002). These increases are driven, in part, by the current demand for highly skilled and high ability workers in today's economy (Goldin and Katz 2007; Johnson and Reed 2007; Taber 2001). By facilitating transfer to four-year 5 institutions, community colleges have the potential to provide an important alternative pathway to the baccalaureate degree, especially for disadvantaged youth. Given the rise of community college enrollment, the growing importance of a baccalaureate degree in today's economy and the role of community colleges in providing access to the baccalaureate by facilitating transfer to four-year institutions, the question remains whether community college entrants fare as well as otherwise similar students who initially enroll in a four-year institution. Community colleges are often seen as contradictory institutions and their effect on educational attainment is unclear. Supporters argue that community colleges serve a democratizing role by increasing access to higher education for disadvantaged students who would have otherwise not attended college (Cohen and Brawer 2003; Rouse 1995), while critics argue that these institutions lower rates of baccalaureate attainment among students who would have otherwise attended four-year colleges (Alba and Lavin 1981; Alfonso 2006; Brint and Karabel 1989; Karabel 1972b). Critics' claims are bolstered by descriptive evidence. Although many community college students aspire to a bachelor's degree, most do not receive any type of degree, few transfer to a four-year college and their rates of bachelor's degree attainment trail far behind those of students who initially enter four-year colleges (Brint and Karabel 1989; Dougherty 1987; Dougherty and Kienzl 2006; Karabel 1972a; Lee and Frank 1990). However, the types of students who enter community colleges are "non-traditional" in many respects, face many barriers to degree completion and transfer, and are very different than the types of students who initially enter four-year schools. Compared to four-year college students, community college students come from less advantaged social 6 class backgrounds, are more likely to be racial/ethnic minorities, have lower levels of secondary school academic achievement and educational and occupational aspirations, are less likely to attend college full-time, and are more likely to delay post-secondary enrollment after high school graduation (Adelman 2005; Chen and Carroll 2007; Dougherty 1992; Horn, Cataldi, and Sikora 2005). These factors are all negatively associated with bachelor's degree attainment (Adelman 1999; Choy 2002). The substantial differences in socioeconomic status, enrollment pathways, and high school achievement between community college entrants and four-year entrants create challenges to identifying whether there is in fact an independent negative institutional effect of attending a community college on educational attainment. Community college students would have been less likely to have completed a bachelor's degree on average even if they had attended a four-year institution. The question remains, however, if their probability of attaining a bachelor's degree is lowered even further by virtue of attending a community college. Critiques of community colleges are based on the assumption that there are institutional features of these schools that impede educational attainment. Community college students tend to have limited social and academic integration and weak connections to the institution, factors that are associated with dropout (Chapman and Pascarella 1983; Dougherty 1992; Tinto 1993; Velez 1985). Community colleges are also unselective and these low levels of academic selectivity lead community college entrants to often be surrounded by peers who are not interested in or good at academic work (Dougherty 1992). Community colleges have also been criticized for failing to provide sufficient advice and counseling for students hoping to transfer and for tracking 7 some students with baccalaureate aspirations into vocational programs (Brint and Karabel 1989; Clark 1960). Students who transfer may also face significant challenges such as adjusting to increases in tuition and living expenses accompanying the move to a fouryear college, losing credits in their transition to a four-year college, and being poorly prepared for the kind of work they will encounter at a four-year college, which makes them prone to failure after they transfer (Cohen 1989; Dougherty 1992; Pascarella and Chapman 1983; Zumeta and Frankle 2007). A lack of strong articulation agreements between community colleges and four-year colleges in some states coupled with inadequate information regarding the prerequisites for academic success and timely degree completion may also create additional challenges for transfer students (Keith 1996). It is unclear from prior research whether any of these institutional features of community colleges do in fact inhibit the educational attainment of students. Studies that have evaluated the effects of community colleges on educational attainment while accounting for selection on observable and unobservable student attributes have come to contradictory conclusions. In general, findings suggest that community colleges may increase the educational attainment of students who would have otherwise not have attended college (Rouse 1995) but that they may decrease years of education attained for students who would have otherwise attended four-year institutions in their absence (Alba and Lavin 1981; Alfonso 2006). Recent work also suggests that community college students who transfer are equally as likely to earn a bachelors degree compared to similar students who begin in a four-year college (Lee, Mackie-Lewis, and Marks 1993; Melguizo 2008; Melguizo and 8 Dowd 2008). On the one hand, this is good news for those concerned with educational stratification and suggests that students who transfer do not face any additional barriers to baccalaureate completion after transitioning to a four-year college. On the other hand, transfer rates remain relatively low- below 40%, even among students who aspire to earn a baccalaureate. Transfer rates are particularly low among students with low levels of secondary school achievement and among socioeconomically disadvantaged students (Adelman 2005; Dougherty and Kienzl 2006; McCormick 1997). It is unclear whether transfer rates remain low because community colleges serve a population of relatively disadvantaged students with diverse educational aspirations or whether community colleges are doing something that creates barriers to transfer. From a policy perspective, it would be helpful to know whether there is anything that colleges can do to increase transfer among their students. From prior research we know a great deal about the characteristics of individual students that are associated with transfer. Students from higher socioeconomic backgrounds, those who enter when they are younger, enroll in a more rigorous high school curriculum and earn higher test scores, enroll continuously, full-time, and do not delay college entry after high school graduation are significantly more likely to transfer than other students (Dougherty and Kienzl 2006; Grubb 1991; Lee and Frank 1990; McCormick 1997). Community college students who transfer are much more similar to students who begin postsecondary school in a four-year institution than students who do not transfer (Dougherty and Kienzl 2006). However, as open-door institutions, community colleges can do little to control the attributes of the students they enroll. If we want to increase transfer rates, therefore, it is important to know what characteristics 9 and programs of individual community colleges promote successful transfer among their students. A better understanding of the organizational and structural features of community colleges that are associated with higher rates of transfer can aid in the development of effective policies that may help colleges reduce the barriers to transfer faced by many of their students. Removing barriers to transfer should help to mediate racial/ethnic and socioeconomic disparities in baccalaureate attainment. Most research on transfer has focused on vertical transitions from community colleges to four-year institutions. Although this work on transfer has gone a long way in helping to improve our understanding of the role of community colleges in the educational attainment of students who first enter them after high school graduation, we know relatively little about the effects of community colleges on the outcomes of students who enter these institutions after attending other types of schools. Students who start in community college do not represent the full range of students who attend community colleges during their undergraduate histories (Adelman 2005). In fact, twenty-six percent of students from the high school class of 1992 who attended community colleges at some point in their postsecondary career started out in other types of institutions (Adelman 2005) and undergraduate reverse transfers constitute 25 percent of the universe of those who started elsewhere. These students who entered a four-year college and later transferred to a community college without having received a bachelor's degree constitute nearly 8% of all initial four-year students (Adelman 2005). This group is characterized by poor academic performance in four-year schools and high rates of non-continuous enrollment (Adelman 2005). 10 In the context of rapid higher education expansion and the drive for 'college for all' the community college has the potential to play a critical role as a postsecondary safety net for the growing number of initial four-year students who enter college with minimal or marginal academic qualifications and who require a great deal of remediation (ACT 2006; ACT 2007; Adelman 2004; Berkner, Chavez, and Carroll 1997). While past research has documented the ways in which the community college offers a second chance to students who have, for individual or structural reasons, not obtained strong academic preparation in secondary school, little work has considered how the community college serves as a second chance for those who began their postsecondary career at a four-year institution but failed to persist (cf. Goldrick-Rab and Pfeffer 2007; Townsend and Dever 1999). Study Overview In this dissertation I study pathways to and from community colleges both nationally and in California. First, I study the effectiveness of the transfer pathway as an alternative route to a baccalaureate degree among students who transfer. Second, I examine variation in transfer rates among colleges and seek to understand what characteristics distinguish schools that are relatively more effective in facilitating transfer among their students. Third, I analyze the transfer process and the outcomes of students who initially attend four-year colleges but later transfer to a community college prior to attaining a bachelor's degree. In chapter 1,1 examine whether community college students who transfer fare as well as their counterparts who initially enroll in a four-year institution. I compare rates of attrition from four-year institutions between community college transfers and students 11 who began their postsecondary education at a four-year college. Prior studies have provided inconclusive evidence as to whether there is a "penalty" for beginning at a community college rather than a four-year college in terms of educational attainment. Inconsistencies in prior work are largely driven by differences in sample composition and differences in methods for handling selection-bias. In this chapter I use a discrete-time hazard model to control for year by year differences in credit accumulation between transfers and initial four-year students and compare the results of alternative approaches to controlling for selection into first college type. My approach facilitates an analysis of the timing of dropout, degree completion, and transfer and permits identification of the periods in which students are most at risk of leaving college. The results from this study demonstrate that any disadvantage in the odds of attrition from a four-year institution experienced by community college transfers relative to four-year natives is explained entirely by differences in enrollment intensity, academic achievement, and socioeconomic background—not by any institutional features of community colleges that inhibit educational attainment. Community college students are indeed disadvantaged relative to four-year natives on a variety of factors associated with attrition from a fouryear institution. On average, they come from families with lower income and education, have lower levels of achievement in high school, work more hours per week, are more likely to delay and interrupt enrollment, and earn fewer credits each term that they are enrolled. After taking these differences into account, however, they do not experience any additional disadvantage simply by virtue of having attended a community college, as some critics have claimed. Therefore, the community college appears to be an effective route to a bachelor's degree for students who transfer. 12 In chapter 2,1 turn my attention to variation in the efficacy of community colleges in facilitating transfer to baccalaureate programs. I use data from the population of students who entered any of California's 107 community colleges between 1992 and 1997 to examine differences in transfer rates among institutions controlling for differences in the populations of students enrolled in different schools. I seek to understand whether there is 'value-added' by community colleges- or an effect of schools on transfer independent of the characteristics of their students. I find clear evidence that community colleges vary considerably in the types of students they enroll but variation in transfer rates is not explained entirely by student-level measures. Even within clusters of colleges with similar enrollment profiles I find variation in rates of transfer to four-year colleges. Because I control for a wide range of student-level variables and for measures of the academic and socioeconomic attributes of students' high schools I can be fairly confident that these differences in transfer result from variation in the organizational or structural attributes of community colleges rather than from uncontrolled attributes of the student bodies. Although I find wide variation in transfer rates even after accounting for student-level characteristics, I am unable to explain much of this variance with available college-level measures. Chapter 3 examines students who enter community colleges after attending a four-year college without earning a baccalaureate degree. Whereas prior studies of persistence in four-year institutions have primarily focused on dropout, I consider the full range of alternative outcomes for students who leave their initial four-year institution: dropout, lateral transfer to another four-year institution, and reverse transfer to a community college. The key questions for my purposes are, given the decision to 13 terminate enrollment in one's initial four-year institution, what differentiates those who transfer to other four-year colleges, those who transfer to a community college and those who dropout entirely? Are similar students better served by transferring to a community college or to another four-year college in terms of their prospects of attaining a degree, the total number of credits earned, and their labor market earnings? I find that disadvantaged students are significantly more likely to transfer to a community college or drop out of higher education entirely without completing a bachelor's degree as are students who enroll in less rigorous high school courses and those who struggle academically in their first school. I find that reverse transfers persist in higher education longer than four-year college dropouts and that, conditional on the number of undergraduate credits they attain, they do not suffer an earnings penalty as a result of attending a community college. Reverse transfers improve their grade point averages upon transitioning to a community college, 20% transfer back to a four-year college and earn a bachelors degree, and another 25% earn an associate degree. Although reverse transfers do not fare as well as students with exclusive four-year college enrollment, they do appear to have more favorable outcomes than otherwise similar students who drop out of postsecondary school altogether upon leaving their initial school. Conclusion As open-door institutions community colleges have the potential to play a powerful role in democratizing access to higher education, providing an alternative more cost efficient route to the baccalaureate by facilitating transfer to four-year institutions, and offering a safety net for four-year college entrants with poor academic preparation 14 who struggle academically in their initial school. For these reasons understanding the role of community colleges in social stratification is an important policy concern. However, prior research leaves many questions unanswered regarding the efficacy of community colleges in improving the educational and economic outcomes of their students. The results of this research suggest that the community college is an effective route to a baccalaureate degree for students who transfer. This is important news for policy makers who in some states are looking to cut costs by encouraging students to begin their college education at a two-year institution (Dougherty and Kienzl 2006). Transfer rates remain low, however, and although they vary considerably among different community colleges, my analysis provide no clear indications as to what policies or practices distinguish colleges that are relatively more successful in promoting transfer among their students. While it is clear from this study and other recent work (Lee, Mackie-Lewis, and Marks 1993; Melguizo 2008; Melguizo and Dowd 2008) that community college students who transfer are no less likely to complete a bachelor's degree than otherwise similar students who begin in a four-year college, students who do transfer appear to be a relatively advantaged group relative to the population of community college attendees (Dougherty and Kienzl 2006; Lee and Frank 1990). Therefore, further studying the factors that contribute to disparities in transfer for students from different social backgrounds will have important implications for our understanding of educational stratification. As an increasing number of students enter four-year institutions minimally prepared for college-level work (ACT 2006; ACT 2007; Berkner, Chavez, and Carroll 1997), the community college will also continue to play an increasingly important role as 15 a postsecondary safety net. While past research has shown that community colleges offer a second chance to students who have, for individual or structural reasons, not obtained strong academic preparation in secondary school, the results from this study make it clear that the community college also serves as a second chance for those who began their postsecondary career at a four-year institution but failed to persist. Students who are minimally prepared and those who struggle academically while in college are significantly more likely to reverse transfer or dropout from their initial four-year institution. However, upon enrolling in a community college those who reverse transfer are able to increase their GPAs, some transfer back to a four-year college and attain a bachelor's degree, and others earn an associate's degree. Some reverse transfers ultimately leave school altogether without earning a degree, but bear a significantly reduced financial burden for their final semester(s). If community colleges are available and more affordable, students who may have dropped out will be more likely to reverse transfer instead. These students will therefore persist in higher education for a longer period of time, earning more credits and enjoying greater initial success in the labor market. Therefore, the community college offers a second chance to earn additional postsecondary training for students who would have otherwise dropped out of college and does so at a discounted price. In sum, the results from this study suggest that community colleges offer an effective and economical route to a bachelor's degree among students who transfer and that they serve as a postsecondary safety net for students who enter four-year colleges with weak academic backgrounds. These findings are consistent with the democratization view of community colleges that contends that these institutions expand postsecondary 16 opportunities and decrease inequality in educational attainment. At the same time, however, the possibility that community colleges also play some diversionary role cannot be ruled out given relatively low rates of transfer and the fact that socioeconomically advantaged students are most likely to transfer. 17 References: ACT. 2006. "Reading Between the Lines: What the ACT Reveals About College Readiness in Reading." ACT, Iowa City, Iowa. —. 2007. "Rigor at Risk: Reaffirming Quality in the High School Core Curriculum." ACT, Iowa City, Iowa. Adelman, Clifford. 1999. "Answers in the Tool Box: Academic Intensity, Attendance Patterns, and Bachelor's Degree Attainment." Washington, D.C. —. 2004. "Principal Indicators of Student Academic Histories in Postsecondary Education, 1972-2000." National Center for Education Statistics, Washington D.C. —. 2005. "Moving into Town—and Moving On: The Community College in the Lives of Traditional-Age Students." Pp. 202: US Department of Education. Available from: ED Pubs. P.O. Box 1398, Jessup, MD 20794-1398. Tel: 877-433-7827; Fax: 301-470-1244; Web site: http://www.edpubs.org. Alba, Richard and David E. Lavin. 1981. "Community Colleges and Tracking in Higher Education." Sociology of Education 54:223-37. Alfonso, Mariana. 2006. "The Impact of Community College Attendance on Baccalaureate Attainment." Research in Higher Education 47:873-903. Alfonso, Mariana, Thomas Bailey, and Marc Scott. 2005. "The Educational Outcomes of Occupational Sub-Baccalaureate Students: Evidence from the 1990s." Economics of Education Review 24:197-212. Berkner, Lutz, Lisa Chavez, and C. Dennis Carroll. 1997. "Access to Postsecondary School for the 1992 High School Graduates." U.S. Department of Education, National Center for Education Statistics, Washington, D.C. Brint, Steven and Jerome Karabel. 1989. The Diverted Dream: Community College and the Promise of Educational Opportunity in America, 1900-1985. New York: Oxford University Press. Card, David and Thomas Lemieux. 2001. "Can Falling Supply Explain the Rising Return to College for Younger Men? A Cohort-Based Analysis." Quarterly Journal of Economics 116:705-46. Chapman, D and Ernest Pascarella. 1983. "Predictors of Academic and Social Integration of College Students." Research in Higher Education 19:295-322. Cheeseman Day, Jennifer and Eric C. Newburger. 2002. "The Big Payoff: Educational Attainment and Synthetic Estimates of Work-Life Earnings." U.S. Census Bureau. Chen, Xianglei and C. Dennis Carroll. 2007. "Part-Time Undergraduates in Postsecondary Education: 2003-04." National Center for Education Statistics Washington, D.C. Choy, Susan P. 2002. "Access & Persistence: Findings from 10 Years of Longitudinal Research on Students." American Council on Education, Washington, D.C. Clark, Burton R. 1960. "The 'Cooling-Out' Function in Higher Education." American Journal of Sociology 65:569-576. Cohen, Arthur M. 1989. The American Community College. San Francisco: Jossey-Bass. Dougherty, Kevin J. 1987. "The Effects of Community Colleges: Aid or Hindrance to Socioeconomic Attainment." Sociology of Education 60:86-103. 18 —. 1992. "Community Colleges and Baccalaureate Attainment." The Journal of Higher Education 63:188-214. —. 2002. "The Evolving Role of the Community College: Policy Issues and Research Questions." in Higher Education: Handbook of Theory and Research, edited by J. S. a. W. G. Tierney. Dordrecht, Netherlands: Kluwer Academic. Dougherty, Kevin J. and Gregory S. 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." Teachers College Record 108:452-487. Goldin, Claudia and Lawrence F. Katz. 2007. "The Race between Education and Technology: The Evolution of U.S. Educational Wage Differentials, 1890 to 2005." NBER Working Paper. Goldrick-Rab, Sara and Fabian T. Pfeffer. 2007. "Second Chances: Student Mobility, Institutional Differentiation, and Stratification in College Completion." in Annual Meetings of the American Sociological Association. New York, NY. Grubb, Norton W. 1991. "The Decline of Community College Transfer Rates: Evidence from National Longitudinal Surveys." The Journal of Higher Education 62:194222. Horn, Laura, Emily Forest Cataldi, and Anna Sikora. 2005. "Waiting to Attend College: Undergraduates Who Delay their Postsecondary Enrollment." National Center for Education Statistics, Washington, D.C. Johnson, Hans P. and Deborah Reed. 2007. "Can California Import Enough College Graduates to Meet Workforce Needs." California Counts: Population Trends and Profiles 8 May. Kane, Thomas J. and Cecilia Elena Rouse. 1999. "The Community College: Educating Students at the Margin between College and Work." Journal of Economic Perspectives 13:63-84. Karabel, Jerome. 1972a. "Community Colleges and Social Stratification." Harvard Educational Review 42:521-562. —. 1972b. "Community Colleges and Social Stratification: Submerged Class Conflict in American Higher Education." Harvard Educational Review 42:521-562. Katz, Lawrence F. and David H. Autor. 1999. "Changes in the Wage Structure and Earnings Inequality." Pp. 1463-1555 in Handbook of Labor Economics, vol. 3A, edited by O. Ashenfelter and D. Card. New York: Elsevier Science. Lee, Valerie E. and Kenneth A. Frank. 1990. "Students' Characteristics that Facilitate the Transfer from Two-Year to Four-Year Colleges." Sociology of Education 63:178193. Lee, Valerie E., Christopher Mackie-Lewis, and Helen M. Marks. 1993. "Persistence to the Baccalaureate Degree for Students Who Transfer from Community College." American Journal of Education 102:80-114. McCormick, Alexander C. 1997. "Transfer Behavior Among Beginning Postsecondary Students: 1989-94." National Center for Education Statistics, Washington D.C. Melguizo, Tatiana. 2008. "Are Community Colleges an Alternative Path For Hispanic Students to Attain a Bachelor's Degree?" Teachers College Record Forthcoming. Melguizo, Tatiana and Alicia C. Dowd. 2008. "Baccalaureate Success of Transfers and Rising Four-Year College Juniors." Teachers College Record Forthcoming. 19 National Center for Education Statistics. 2005. "Digest of Education Statistics." National Center for Education Statistics, Washington, D.C.. —. 2007. "The Condition of Education." Pp. Transition to College. Pascarella, Ernest T. and D.W. Chapman. 1983. "Validation of a Theoretical Model of College Withdrawal." Research in Higher Education 19:25-48. Reynolds, John, Mike Stewart, Ryan MacDonald, and Lacey Sischo. 2006. "Have Adolescents Become Too Ambitious? High School Seniors' Educational and Occupational Plans, 1976 to 2000." Social Problems 53:186-206. Richardson, Jeanita W. 2005. "Who Shall be Educated?: The Case of Restricting Remediation at the City University of New York." Education and Urban Society 37:174-192. Rouse, Cecilia Elena. 1995. "Democratization or Diversion? The Effect of Community Colleges on Educational Attainment." Journal of Business & Economic Statistics 13:217-224. Sturrock, Carrie. 2003. "Cal State Begins to Enforce Math, English RemediationFreshmen Told Not to Return Until They Complete Community College Classes." Pp. aOl in Contra Costa Times (Walnut Creek, CA). Taber, Christopher. 2001. "The Rising College Premium in the Eighties: Return to College or Return to Unobserved Ability?" The Review of Economic Studies 68:665-691. Tinto, Vincent. 1993. Leaving College: Rethinking the Causes and Cures of Student Attrition. Chicago, IL: The University of Chicago Press. Townsend, Barabara and John T. Dever. 1999. "What do We Know About Reverse Transfer Students?" New Directions for the Community College 106:5-14. Velez, William. 1985. "Finishing College: The Effects of College Type." Sociology of Education 58:191-200. Wellman, Jane. 2002. "State Policy and Community College-Baccalaureate Transfer." National Center for Public Policy and Higher Education, San Jose. Zumeta, William and Deborah Frankle. 2007. "California Community Colleges: Making them Stronger and More Affordable." The National Center for Public Policy and Higher Education. 20 CHAPTER 2: Community College Transfer and Baccalaureate Attainment 21 Abstract: By facilitating transfer to four-year institutions, community colleges have the potential to provide an important alternative pathway to the baccalaureate degree, especially for disadvantaged youth. However, questions remain about whether community college transfers are equally as likely to persist to the baccalaureate compared to their peers who initially enroll in four-year institutions. In this paper I compare the hazard of attrition from a four-year college among community college transfer students and initial four-year students. I use a discrete-time hazard model to control for year by year credit accumulation and compare the results of alternative approaches to controlling for selection into first college type. The results from this study demonstrate that any disadvantage in the odds of attrition from a four-year institution experienced by community college transfers is explained entirely by differences in enrollment intensity, academic achievement, and socioeconomic background—not by any institutional features of community colleges that inhibit educational attainment. Therefore, the community college appears to be an effective route to a bachelor's degree for students who transfer. The effect of community colleges on the attainment of students who fail to transfer, however, remains unclear. More than half of the students who enter college shortly after high school graduation will begin their postsecondary education at a community college (Adelman 2005). Students from disadvantaged social and economic backgrounds are even more likely to be found in these institutions than are their more affluent peers and their representation has increased substantially over time. Although students enter community colleges with varying ambitions, the transfer function of these institutions remains very important and has become increasingly important in recent years for a variety of reasons. For example, some states are looking to cut costs by encouraging students to begin their college education at a two-year institution (Dougherty 2002; Mercer 1992). States where affirmative action has been abolished, such as California and Texas, are increasingly looking to community college transfer as a way of maintaining a diverse student body (Hebel 2000). Finally, those concerned with social stratification have emphasized the importance of transfer, especially for students from disadvantaged social and economic backgrounds who are increasingly relying on community colleges for access to the baccalaureate because of rising four-year college tuition, stagnating need-based student aid, declining incomes for disadvantaged families, and the reduction of remedial education in public four-year colleges (Dougherty 2002; Dougherty and Kienzl 2006; Richardson 2005; Sturrock 2003; Wellman 2002). Given the growing importance of a baccalaureate degree in today's economy and the role of community colleges in providing access to the baccalaureate by facilitating transfer to four-year institutions, the question remains whether community college transfers fare as well as their counterparts who initially enroll in a four-year institution. Researchers have long debated whether community colleges democratize higher 23 education by expanding enrollment opportunities or whether they divert disadvantaged students from attaining bachelors degrees (Alfonso 2006; Cohen and Brawer 2003; Gonzalez and Hilmer 2006 ; Karabel 1972b; Leigh and Gill 2003). On the one hand, some have argued that community colleges serve a democratizing role by increasing access to higher education for disadvantage d students given their open door admissions policies, low costs, and their geographic dispersion (Cohen and Brawer 2003; Rouse 1995). On the other hand, some have argued that community colleges have failed to live up to this democratizing ideal. These critics maintain that community colleges aid in social reproduction, ensuring that disadvantaged children inherit their parents' social class positions (Brint and Karabel 1989; Karabel 1972a) or that they simply serve a "cooling out" function working to level the unrealistic aspirations of low-achieving youth (Clark 1960). Rather than democratizing higher education, these critics argue that community colleges are an important component of a class-based tracking system in higher education that diverts disadvantaged students away from four-year colleges and from attaining a bachelor's degree (Alba and Lavin 1981; Karabel 1972a; Rouse 1995). Studies that have evaluated these two perspectives have provided contradictory results but, in general, they suggest that community colleges may increase the educational attainment of students who would have otherwise not have attended college (Rouse 1995) but that they may decrease years of education attained for students who would have otherwise attended four-year institutions in their absence (Alba and Lavin 1981; Alfonso 2006). The potential of community colleges to ameliorate inequalities in higher education hinges not only on their role in increasing access to higher education but also 24 on their role in providing access to the baccalaureate by facilitating transfer to four-year institutions. The purpose of this study is to compare rates of attrition from four-year institutions between community college transfers and students who began their postsecondary education at a four-year college (hereafter "four-year natives"). Rather than comparing the bachelor's degree attainment of all students who begin in a community college with all students who begin in a four-year college as several prior studies have done (Alfonso 2006; Rouse 1995), I restrict the sample of community college students to those who successfully transition to a four-year institution. Limiting my sample to transfer students helps me avoid the difficulty of attempting to identify which community college entrants really aspire to transfer and earn a bachelor's degree and helps ensure that I am making a fair comparison. Prior research shows us that the size of the community college effect depends, in part, on how researchers define the counterfactual (Alfonso 2006; Lee, Mackie-Lewis, and Marks 1993; Melguizo and Dowd 2008; Rouse 1995; Velez 1985). Focusing only on transfer students permits an appropriate comparison with initial four-year students without confounding the results by the presence of community college students with uncertain educational ambitions (Melguizo and Dowd 2008). In this study I also make two important methodological improvements over prior work. First, I use a variety of approaches to handle self-selection into a community college. I compare results adjusted on an extensive set of observable student attributes to results from selection models using propensity scores and instrumental variables. To consistently estimate the effect of attending a community college on bachelor's degree 25 attainment, it is necessary to account for the observable and unobservable factors that lead students to first enroll in a community college rather than a four-year institution. Failure to account for both components of the selection process can result in biased estimates of the community college effect, since characteristics that determine college selection might also predict the probability of attaining a bachelor's degree. Second, I utilize information from each year of student enrollment in postsecondary school to adjust for differences in credits earned between community college transfers and four-year natives. While some studies have compared rates of baccalaureate attainment between transfers and four-year natives (Lee, Mackie-Lewis, and Marks 1993; Melguizo and Dowd 2008), none have adequately controlled for differences in part- versus full-time enrollment status. By standardizing on credits earned, my approach permits a comparison of initial two- and four-year students who are not only similar on a host of covariates but who have also made similar progress toward their bachelor's degree. This is of critical importance because few transfer students actually transfer with an associate's degree or have junior status upon initial enrollment at their four-year institution (McCormick 1997). Therefore, two years of attendance at a community college, even as a full-time student, may not yield as many credits as two years of attendance at a four-year college. My approach facilitates an analysis of the timing of dropout, degree completion, and transfer and permits identification of the periods in which students are most at risk of leaving college. Prior Research Early research on the effects of community colleges fueled critics' contentions that these institutions play a diversionary role working to hinder the attainment of disadvantaged students. These studies found high attrition from community colleges, low rates of transfer to four-year institutions, and low rates of bachelor's degree attainment among successful transfers (Adelman 2004; Anderson 1981; Brint and Karabel 1989; Dougherty 1992; Holmstrom and Bisconti 1974; Velez 1985). More recent studies have used more robust methods to account for observable and unobservable differences between those who begin in a two-year versus a four-year college after high school graduation. The bulk of these recent studies suggest that, on average, community colleges increase years of educational attainment among students who would not have otherwise attended college and, while there is still some evidence of a diversion effect, recent research suggests that this effect is smaller than suggested by early work (Alfonso 2006; Gonzalez and Hilmer 2006 ; Grubb 1989; Leigh and Gill 2003; Melguizo and Dowd 2008; Rouse 1995). Most studies that have examined the effect of community colleges compare rates of bachelor's degree attainment or years of education earned by students who first attended a community college and those who initially attended a four-year college. The goal of this research is to determine whether there is a penalty to beginning at a community college rather than a four-year college after accounting for differences in the attributes of initial two- and four-year students. There are a variety of theories as to why community college students may experience less favorable educational outcomes relative to their four-year counterparts. For example, community college students tend to have limited social and academic integration and weak connections to the institution, factors that are associated with dropout (Chapman and Pascarella 1983; Dougherty 1992; Tinto 1993; Velez 1985). Community colleges are also unselective and this lack of selectivity leads community college entrants to often be surrounded by peers who are not interested in or good at academic work (Dougherty 1992). Community colleges have also been critiqued for failing to provide sufficient advice and counseling for students hoping to transfer and for tracking some students with baccalaureate aspirations into vocational programs (Brint and Karabel 1989; Clark 1960). Students who transfer from a community to a four-year college may also face significant challenges such as adjusting to increases in tuition and living expenses accompanying the move to a four-year college, losing credits in their transition to a four-year college, and being poorly prepared for the kind of work they will encounter at a four-year college, which makes them prone to failure after they transfer (Cohen 1989; Dougherty 1992; Pascarella and Chapman 1983; Zumeta and Frankle 2007). A lack of strong articulation agreements between community colleges and four-year colleges in some states coupled with inadequate information regarding the prerequisites for academic success and timely degree completion may also create additional challenges for transfer students (Keith 1996). A significant challenge with this line of research is fully controlling for all differences between initial two- and four-year students. Compared to four-year college students, community college students come from less advantaged social class backgrounds, are more likely to be racial/ethnic minorities, have lower levels of secondary school academic achievement and educational and occupational aspirations, are less likely to attend college full-time, and are more likely to delay post-secondary enrollment after high school graduation (Adelman 2005; Chen and Carroll 2007; Dougherty 1992; Horn, Cataldi, and Sikora 2005). These factors are all negatively associated with bachelor's degree attainment (Adelman 1999; Choy 2002). Failure to 28 adequately adjust for the many differences between initial two- and four-year Students may lead to biased estimates of the community college effect since uncontrolled factors associated with beginning postsecondary school at a community college instead of a fouryear college may also be associated with degree attainment (Alba and Lavin 1981; Alfonso 2006; Rouse 1995). Though most studies control for differences in family background and academic achievement, this approach is based on the assumption that, conditional on the included covariates, there are no additional factors related to both attending a community college and attaining a bachelor's degree. Although conditioning on family background and achievement goes a long way toward addressing the selection problem, there are likely unobserved measures of ability and aspirations that are not fully captured by such measures (Rouse 1995). Several prior studies have attempted to use more robust quasi-experimental methods to account for this potential selection bias. Alba and Lavin (1981) attempted to control for self-selection in their study of applicants to four-year schools in the City University of New York system. They used a natural experiment to compare the educational outcomes of two similar groups of students who applied for entry to CUNYs and met the open-admissions category for entry to the senior colleges: one group was placed in them, while the other group was placed in two-year schools. They argue that the two groups are very similar academically because acceptance or rejection was determined by factors weakly related to subsequent academic achievement and because two-year students' baccalaureate aspirations were confirmed in a later survey (Alba and Lavin 1981). They find that students assigned to a two-year college had lower educational attainment than the four-year students and conclude that community colleges deter students from attaining their educational aspirations. There is evidence, however, that assignment to two and four year schools in this case were not entirely random as the two groups differed in average levels of high school achievement. This unobserved heterogeneity may undermine their finding of lower educational attainment among initial two-year students (Rouse 1995).2 Several other studies have used instrumental variables and/or propensity score matching techniques to estimate the effect of community colleges on educational attainment. Rouse (1995) used distance between high schools students attended and public two- and four-year colleges and average in-state tuition at public two- and fouryear colleges as instruments with which to identify selection into a community college. Her findings suggest that community college students are less likely than four-year students to attain a bachelor's degree prior to adjusting for selection. However, after adjusting for selection in her instrumental variables models, only small non-significant differences in educational attainment between two- and four-year students remain. A similar instrumental variables approach was taken by Alfonso (2006), although she found a significant negative community college effect ~ enrolling at a community college, instead of a four-year institution, reduced students' chances of attaining a bachelor's degree by between 21% and 33%. The difference in these findings is possibly due to differences in data used (High School and Beyond class of 1980 and National Education Longitudinal Study class of 1992) and to differences in the set of control variables included in the models. 1 Students assigned to a two-year college were about 20% less likely to attain a bachelor's degree, earned substantially fewer credits, and had lower second to third and third to fourth year persistence than students assigned to four-year institutions, net of high school achievement and college performance (Alba and Lavin 1981). 2 For an additional critique of Alba and Lavin's (1981) study see Lau (1984). 30 Long and Kurlaender (2007) used propensity score matching and instrumental variable strategies to evaluate the effect of community college attendance in the state of Ohio. Their findings suggest that community college students are 14 percent less likely to obtain a bachelor's degree in six years and that they complete 18 fewer credits than students who initially attend a four-year college. Finally, Melguizo et al. (2008) compared rates of bachelor's degree attainment between junior level four-year students and community college transfers using propensity score matching to control for selection on observables. They found no differences in the outcomes of transfers and four-year college juniors and conclude that there are no penalties to degree completion and credits earned for the small proportion of community college entrants who transfer to a four-year college. While instrumental variables have been used with some success in prior research, this method can present some problems. Finding instruments that are strongly correlated with the endogenous explanatory variable but conditionally independent of the outcome can be difficult. Using instruments that are only weakly correlated with the endogenous variable can produce inconsistent estimates with large standard errors (Bound, Jaeger, and Baker 1995; Cameron and Trivedi 2005). The measures of college accessibility that are typically used to identify selection into a community college (proximity to community and four-year colleges and average tuition of community and four-year colleges in respondents' states), while strongly associated with community college entrance, have been critiqued as being potentially endogenous. If, for example, families who value education choose to live in close proximity to colleges or to live in states with low college tuition then the instruments are not truly exogenous which can create biased 31 estimates (Rouse 1995). The strength of instrumental variables is that, in theory, they allow the analyst to adjust for selection on unobservables. However, given the potential problems with using these models, I will compare the results of an IV model with alternate models adjusted on an extensive set of observable student attributes and to a model based on propensity scores. Another significant challenge in studying whether there is a penalty to attending a community college involves determining which groups of two- and four-year students to compare. Three different approaches have been taken in prior research: comparing all initial two-year to all initial four-year students, comparing community college transfers to all initial four-year students, and comparing community college transfers to initial fouryear students who persist until their junior year. Each of these approaches presents some problems. First, studies that compare all initial two- and four-year students are problematic given that many community college students enroll without any intention of transferring and attaining a bachelor's degree (Cohen and Brawer 2003). Although some studies control for students' stated educational aspirations (e.g., Leigh and Gill 2003), this remains an imperfect solution given other studies that have shown that students' aspirations tend to be universally (and often unrealistically) high and that the connection between ambitions and attainments has become increasingly weak over time (Reynolds, Stewart, MacDonald, and Sischo 2006). Studies that have compared all initial two- and four-year students have found that community college entrants earn fewer years of education than four-year college entrants, although the size of this effect varies in 32 different studies given variation in control variables and methods for handling selection bias (Alfonso 2006; Anderson 1981; Leigh and Gill 2003; Rouse 1995; Velez 1985). Also, although this method may be useful for identifying the total community college effect, it is somewhat problematic because it offers no clear policy implications. Putting aside issues of selection for a moment, there are at least three stages where community college students may fall behind initial four-year students in their progress toward a bachelor's degree: high rates of attrition from community colleges, failure to transfer to a four-year college, and high rates of attrition from four-year institutions among community college transfers. The results of these studies, however, provide no indication of which of these stages contribute most to the negative community college effects identified. In order to develop effective policies, it would be helpful to know if all or most of the negative community college effect is concentrated within one of these three stages. By restricting my analyses to community college students who transfer to a four-year institution, I am able to evaluate the effectiveness of the transfer pathway as an alternative route to a bachelor's degree among those who do transfer and to gauge whether and when transfers fall behind four-year natives en route to the baccalaureate. Second, studies that compare community college transfers with all initial fouryear students are potentially biased because comparisons are usually made between students who have not made equitable progress toward their degrees. Since a substantial share of attrition from college occurs during the freshman and sophomore years (Pascarella and Terenzini 1991), this method biases the comparison of completion rates in favor of transfer students. The students whom I observe transferring have already persisted in college for one or two years and are therefore more likely to complete while 33 the comparison group includes all initial four-year students, even those who drop out in the first year or two. Studies that have used this approach have also found a negative effect of beginning at a community college, although they likely underestimate this negative effect given potential bias in sample composition (Ganderton and Santos 1995). Finally, some studies have compared rates of bachelor's degree completion between transfer students and initial four-year students who persist until their junior year or who remain enrolled two years after initial entry (Holmstrom and Bisconti 1974; Lee, Mackie-Lewis, and Marks 1993; Melguizo 2008; Melguizo and Dowd 2008). Though a significant improvement, this approach is also flawed if differences in credits earned are not taken into account. Few transfer students actually transfer with an associate's degree or have junior status upon initial enrollment at their four-year institution (McCormick 1997). Therefore, if credit differences are not considered, these comparisons are biased in favor of four-year natives since, on average, four-year natives who are juniors will have made more progress toward their degree than transfer students. Results using this method of comparison are mixed. Studies of students in the 1960s and 1970s found a persistent disadvantage for community college transfer students en route to the baccalaureate compared to initial four-year students (see Dougherty [1987,1991] for reviews of this literature). However, studies using more recent data from the 1980s and 1990s have found that after controlling for achievement and family background measures there are no differences in the probability that transfer students receive a bachelor's degree compared to initial four-year students (Lee, Mackie-Lewis, and Marks 1993; Melguizo and Dowd 2008; Melguizo, Kienzl, and Alfonso 2008). The authors of these studies conclude, therefore, that any negative community college effect is exerted within 34 the community college through its effect on transfer. Community college students who transfer and make the transition to a four-year institution are a select group whose odds of graduation are not inhibited by their attendance at a community college (Lee, MackieLewis, and Marks 1993). To overcome the potential bias present in these prior studies, I use an alternate approach that permits a comparison of initial two- and four-year students who have made similar progress toward their degree by controlling for differences in credits earned. This is especially important because community college students are more likely to interrupt enrollment, attend part-time, work while enrolled, and delay enrollment (Chen and Carroll 2007; Horn, Cataldi, and Sikora 2005) all of which are negatively related to bachelor's degree completion (Adelman 1999; Choy 2002). While some studies control for student reports of their enrollment status (i.e., full versus part time), none take into account that this may vary over the course of students' postsecondary careers. Using a time frame of one academic year, Chen and Carroll (2007) and McCormick, Geis, and Vergun (1995), for example, found that 14-16% of students change their enrollment status at least once. Students may also enroll for a full term of credit but if they fail or withdraw from one or more courses they may fail to earn a full term of credit, which could also create bias in student reports of their enrollment status. Therefore, an accurate examination of differences in graduation rates among transfer students and four-year students requires that the analyst create a credit standardized measure of enrollment duration to take the timing of transfer and differences in the rate of credit accumulation into account, which has not been done in prior research. Data 35 I use data from the National Educational Longitudinal Study (NELS), including postsecondary transcript data, to compare the hazard of four-year college attrition among community college transfer students and initial four-year students. NELS follows a nationally representative cohort of 1988 8th graders for 12 years, with follow-ups in 1990, 1992, 1994, and 2000. The postsecondary transcripts were collected in 2000 (8.5 years after high school graduation for most sample members) for 8,900 NELS sample members who reported obtaining any postsecondary education and include information on coursetaking behavior, degrees obtained, and characteristics of the postsecondary institutions these students attended (Adelman, Daniel, Berkovits, and Owings 2003). The postsecondary transcript data can be linked with information from students' high school transcripts as well as with student and parent surveys. My analytic sample is restricted to those students with complete postsecondary records who participated in the high school transcript study. These restrictions are necessary since I am examining college credit accumulation (and thus require students' complete records) and given prior findings of the importance of high school achievement and course taking in determining first college type and eventual degree attainment (Adelman 1999; National Center for Education Statistics 2004). My sample is further restricted to only include students who attended a public high school since the geographic identifiers necessary to estimate the distance students must travel to attend a community college or four-year college are not available for students who attend private high schools. Eighty percent of sample members attend public high schools, including 90% of initial community college students and 78% of initial four-year students. The key outcome variables in my analyses are students' first college type, their transfer status, and their persistence in a four-year institution. The first institution type variable appears as the "reference" institution on the postsecondary transcript file. It excludes any institution the student attended exclusively prior to their high school graduation date, any institution the student attended only in the summer between the high school graduation date and the beginning of fall semester, and cases of "false starts" (i.e., when the student enrolls, then withdraws without earning credit within one academic term). Students are classified as either initial four-year students or initial two-year students. Students with no postsecondary education and those who attend for-profit institutions are excluded from the analyses. Students are considered community college transfers if their first institution (i.e., their reference institution) was a public two-year college and they later transferred to a four-year institution. I place no restrictions on the time between leaving the community college and enrolling in the four-year college.3 Finally, persistence is measured by whether the respondent dropped out of a four-year institution prior to earning a bachelor's degree. Students are considered to be dropouts if they ceased enrollment at a four-year institution prior to receiving a bachelor's degree and did not re-enroll before the end of the study period. My main concern is in comparing time-varying rates of attrition from a four-year institution between community 3 Some transfers do delay enrollment in a four-year institution after ceasing enrollment in community college, especially students who transfer after earning few credits in a community college. The mean of these delays is about 6.5 months, slightly higher than what we would expect if a student ceased enrollment in a community college in the spring of one academic year and immediately entered a four-year college the following fall (in which case we would expect an enrollment lapse of about 4 months over the summer). I account for these delays by including a measure of whether students were continuously enrolled (i.e., did not have a break in enrollment of one-year or more) in postsecondary school. This measure captures lapses in enrollment for transfer students and also accounts for non-continuous enrollment among four-year natives. 37 college transfers and four-year natives and in identifying the periods when students are at the highest risk of leaving college. To compare community college transfer students and four-year college natives who are at similar stages of their degree programs, I use a measure of the cumulative number of credits each student has earned. This measure is constructed by taking the course level postsecondary transcript data and collapsing it to the year level, taking the sum of the credits earned for each year. Next, the cumulative number of credits earned is computed by summing the credits earned in the current year and in all prior years. The credit measure is standardized to a semester scale so that schools on different calendars are directly comparable (i.e., standardized credits earned during a quarter are determined as: (credits*2)/3). Using this detailed measure of credits earned allows me to account for full versus part time status in a much more nuanced way than has been done in prior studies. One consideration with using credits as a detailed measure of full- versus parttime enrollment status is whether credits attempted or credits earned should be used. Credits earned is more appropriate for my purposes (i.e., comparing transfers and fouryear natives who have made similar progress toward their degrees) but it is also important to account for the fact that students may enroll in courses that they fail to earn credit from (i.e., if they earn a non-passing grade for a course or if they withdraw from a course). Therefore, I also include a measure that reflects the proportion of credits attempted that were earned in each year. I also include a time-varying measure of whether the student ever enrolled part-time (i.e., in fewer than 12 units) either in or prior to a given year. 38 Unfortunately, there is no way to determine whether credit for community college courses was accepted by the four-year institution, which creates a problem when computing the number of credits earned by transfer students. There is no ideal solution to this problem. The approach I take in my analyses is to generate a variable that flags courses as either academic or vocational. The coding for this flag is based on the 1998 revision to the secondary school taxonomy (Bradby and Hoachlander 1999). For the credit variables used in my analyses, the only credits that are included are those which were earned in academic courses. This is true for both community college transfers and four-year natives, regardless of whether the credit was earned at a two- or four-year institution.4 About 70% of all credits earned were from academic courses. NELS also includes a rich array of variables that I will use to control for differences in the individual attributes of community college transfers and native fouryear students. These variables include measures of student socioeconomic background, high school and college academic performance, and measures of non-traditional enrollment pathways, all of which are associated with both beginning one's postsecondary education at a two-year institution and the likelihood of completing a degree (Adelman 1999; Kane and Rouse 1999) . The measures of students' social and economic background include: gender, race/ethnicity, parent reports of parental education and family income, region of residence, the number of hours worked while in college, and whether the respondent had a child by 1994. Using total credits earned for all courses instead of total academic credits does not change the substantive results. To measure high school achievement I use high school grade point average (from high school transcripts), math and reading achievement test scores5, whether the student was in the college preparatory track (from high school transcripts), whether the student took algebra 2 or a higher math class (from high school transcripts), and whether the student graduated from high school on time. To control for college academic achievement I include four-year college grade point average. To control for nontraditional enrollment pathways I include measures of whether the student delayed college entry by more than eight months after high school graduation, whether the respondent was enrolled continuously, and the measure of part-time enrollment status mentioned above. Students are considered to have been continuously enrolled if, within the span of their first and last enrolled terms, they do not have a break of a year or more. For the instrumental variable models I use three instruments to identify selection into a community college: distance to the closest community college, distance to the closest four-year college, and the difference in the average mandatory tuition and fees charged by two- and four-year colleges in the respondent's state for each year. I also include non-linear transformations of these variables. The tuition measure, coded in real (2000) dollars, is obtained from the Annual Survey of Colleges for the year that the respondent graduated from high school. I measure miles to the closest community and four-year colleges by estimating the distance between the zip code centroid of the respondent's high school and the zip code centroid of each community college included in IPEDS. Zip codes for respondent's high schools are obtained by linking the NELS data to the Common Core of Data- a survey of the universe of public schools in the United 5 Test scores are standardized to have a mean of zero and a standard deviation of 1. Students were assessed in reading and math in the 8th, 10th, and 12th grades. The mean of the standardized test scores from all available assessment years are used in the models. 40 States. I also take the complex sample design of NELS into account in all of my analyses by using appropriate weights.6 Methods To provide a complete view of the process of bachelor's degree attainment for community college transfers it is necessary to study three phases of the attainment process. First, there is the decision to attend a community college versus a four-year college and the timing of college entry. Second, there is the decision to transfer to a fouryear institution and the timing of transfer. All students are at risk of enrolling in a community college as their initial institution, but only community college students are at risk of transferring to a four-year college. Third, there is persistence toward the bachelor's degree for initial community college students who transfer to a four-year institution. Initial community college students are not at risk of attaining a bachelor's degree until they transition to a four-year college. I estimate three sets of models to provide a complete examination of the attainment process of transfer students: 1) the probability of entering a community college as a function of family background, high school achievement, and college accessibility in the student's area of residence; 2) the hazard of transferring to a four-year college among initial community college students as a function of family background, high school and college achievement, and credits earned; and 3) the hazard of attrition from a four-year institution prior to receiving a bachelor's degree among transfer students The f4f2hp3w weight is used in my analyses, which applies to 12 grade freshened panel members with high school transcripts and complete postsecondary transcript records. The target population for this weight is students who were 12th graders in spring 1992 and who subsequently participated in some post-secondary education. 41 and four-year natives as a function of the same covariates included in the transfer equation. I use three alternative approaches to controlling for selection into first college type. First, I estimate logistic regression models of attrition that control for a wide range of student-level characteristics including family social and economic background, high school and college achievement, and credits earned using standard logistic regression techniques. Although conditioning on family background and achievement goes a long way toward addressing the selection problem, there are likely unobserved measures of ability and aspirations that are not fully captured by such measures (Rouse 1995). Second, I estimate models where I instrument for first college type using regional variation in college accessibility based on distance to the closest two- and four-year institutions and the difference in average tuition between these institutions in each state. Prior studies have found that students who reside in closer proximity to a community or four-year college are more likely to attend postsecondary school (Anderson, Bowman, and Tinto 1972; Rouse 1995). Also, the majority of undergraduate students attend college in the same state where they attended high school and most do so at public colleges (Alfonso 2006; Rouse 1994). Studies have also shown that college attendance decisions are sensitive to college costs (Angrist 1993; Dynarski 2003; Kane 1994) and I hypothesize that community colleges become more attractive to students as the average cost of public four-year colleges increases relative to the average cost of public two-year colleges. There is considerable variation between states in the tuition charged by public two- and four-year colleges and in the extent to which tuition at a community college is a 42 "bargain" compared to tuition at a four-year college (National Center for Public Policy and Higher Education 2006). To the extent that these measures of distance and differences in mean tuition are orthogonal to the characteristics of students that lead them to choose to attend a community college or that constrain their postsecondary alternatives to the community college, the expected probability of attending a community college based on these exogenous predictors will capture the exogenous components of variation in community college attendance. Using instrumental variables should improve upon results that simply control for variables within a standard regression model if there are unobserved factors associated with both the decision to begin postsecondary school at a community college and degree attainment. The third approach I take, based on propensity scores, is a compromise between the limitations of covariate adjustment on the one hand and the limitations of instrumental variables based on weak instruments on the other. Where the IV measure is a predicted probability based on attributes exogenous to the selection process, the propensity is based on all the covariates thought to lead to the endogeneity of community college attendance. By including the propensity and a dummy indicator for whether or not the student began his or her postsecondary career at a community college, I hope to purge the community college indicator of its endogeneity. Where the IV approach ideally captures only a fraction of the exogenous variance in community college attendance, the propensity measure captures only endogenous variance in community college attendance, leaving all of the remaining (exogenous) variance in the community college indicator (Winship and Morgan 1999). 43 First Stage Models The first step I take to examine whether transfer students experience inferior educational outcomes relative to their four-year counterparts is to estimate the probability of entering a community college versus a four-year college using both propensity score and instrumental variables models. This first stage of estimation is important for two reasons. First, it is substantively important to know what factors (i.e., family background, high school achievement, aspirations, and measures of college accessibility) are associated with the type of postsecondary school students first attend so I know what types of students are more likely to enter community colleges. Second, since the decision to attend a community college versus a four-year college is a potentially endogenous outcome associated with bachelor's degree attainment, the first stage estimates can be used to control for selection into a community college in later models. The propensity score in this instance is simply the predicted probability that an individual with a set of observed characteristics Z, will begin their postsecondary education at a community college instead of a four-year college, or P(Z ; ) = Pr(C = l | Z ; ) (1) where C equals 1 for students who enter a community college. Since my primary concern is with identifying the factors associated with first college type (i.e., two- versus four-year) rather than the factors associated with attending college in general, students with no postsecondary experience are excluded from these models. If choice of first college type is purely a function of the observed Zs (i.e., selection is only on the observables), then conditional on the Zs, choice of first college type is random with respect to outcomes (Winship and Morgan 1999). I account for selection using propensity scores with a two-stage estimation strategy. In the first stage, I estimate the propensity of entering a community college versus entering a four-year college as a function of socioeconomic status (parent reports of income and education), race/ethnicity, gender, high school academic achievement (high school grade point average, standardized test scores), highest math course taken in high school, educational aspirations, and whether the respondent was an on-time high school graduate. I also include fixed effects for state of residence given substantial between state differences in the scope of community college systems (Dougherty and Kienzl 2006; Lee, Mackie-Lewis, and Marks 1993). In the second stage, I estimate the hazard of leaving college prior to earning a bachelor's degree including the propensity score as a control variable. Next, as an alternative approach I instrument for first college type using proximity (in miles) from the respondents' high school to the closest community college, miles to the closest four-year college, and the difference in tuition between public two- and fouryear colleges in the respondent's state. The means presented in Table 2.1 suggest some preliminary evidence for the quality of my instruments given that relative to four-year natives, community college transfers live in closer proximity to community colleges and reside in states where the average difference in tuition between public four-year and twoyear colleges is larger. The instrumental variables models will be estimated in two stages. In the first stage, the probability of entering a community college rather than a four-year institution will be predicted as a function of the instruments. In the second stage, I estimate the hazard of attrition from a four-year college and replace the observed values of first college type with the predicted values from the first stage. Modeling Community College Transfer and Four-Year Attrition 45 I use a discrete time hazard model to estimate a) the hazard of transfer from a community college to a four-year institution among community college entrants and b) the hazard of attrition from a four-year institution among four-year college natives and community college transfers. While the ultimate goal of the analysis is to compare rates of attrition from a four-year institution among transfer students and four-year natives, it is also important to know what factors are associated with transfer from a community college so that the selectivity of transfers can be gauged. Similar to prior research that has studied college attrition and degree completion longitudinally, I conceptualize time in my hazard models by using years of enrollment (DesJardins 2002a; DesJardins 1999; DesJardins 2002b; Ishitani 2006). Years, however, are not based on calendar years but instead are based on the number of terms a student has enrolled. Two terms of enrollment is equivalent to one year for students who attend schools on semester calendars while three terms of enrollment is equivalent to one year for students who attend schools on quarter calendars. For students who attend continuously, years of enrollment based on the number of terms will be the same as calendar years elapsed since initial enrollment whereas these values will differ for students who interrupt enrollment. The data are structured such that each student contributes one observation for each year that they are enrolled (i.e., one observation for each period that a student is at risk of transfer/dropout) (Singer and Willet 2003).7 The years that transfers were enrolled in community college are not included in the four-year attrition models; however, their cumulative number of credits earned after their first year of enrollment in a four-year 7 The standard errors in my estimates are calculated using Huber-White robust standard errors that are adjusted for correlations of error terms across observations. 46 college includes the number of credits they earned in community college. For example, a transfer student who earned 30 credits at a community college and 20 credits in their first year of enrollment in a four-year college will have a cumulative total of 50 credits earned after their first year of enrollment in a four-year college. The discrete-time hazard of transfer/attrition for person / at yeary, /z,7, can be interpreted as the conditional probability of the event occurring in time period j , given that the event did not occur prior to yeary, and is defined as: hj = Pr[r = j I T > j] (2) where Tis a discrete random variable that indicates the timing of the event (Singer and Willet 1993; Singer and Willet 2003). In the transfer analyses, the dependent variable retains a value of 0 until the student's last year of enrollment at a community college where it changes to a 1 for students who transfer and retains a value of 0 for students who do not transfer. Students who fail to transfer are censored at their last year of enrollment. For the four-year attrition models, the dependent variable retains a value of 0 until the student's last year of enrollment where it changes to a 1 for students who dropout and remains at 0 for students who receive a bachelor's degree and for those who are still enrolled. Students who receive a bachelor's degree are censored at the time of graduation. Students who are still enrolled in postsecondary school and working on a bachelor's degree are censored at the time of the last survey folio wup. The hazard of transfer and attrition are estimated using the following equation: f In h,j \\-hijj > = [aiDuj + aiDiu + ... + ajDjij] + [fipXptj] (3) 47 where Djj are dummy variables representing the year of enrollment for person / in yeary and Xpy represents a vector of student-level variables (those discussed in the previous section) for student i in yeary. For the transfer models, I begin by estimating the hazard of transfer as a function of social origins. In subsequent models I add measures of high school and college achievement to gauge which factors mediate the association between socioeconomic background and transfer. For the four-year attrition models, I begin by comparing the hazard of attrition between transfers and four-year natives solely as a function of years of enrollment and credits earned. Then I enter measures of social origins, high school achievement, and college achievement/attendance pathways to ascertain which of these factors help to explain differences in attrition between transfers and four-year natives. I estimate this same series of models for each of my three methods for controlling for selection into first college type (i.e., covariate adjustment, instrumental variables, and propensity score control). Results Descriptive statistics for the variables used in the multivariate models are presented in Table 2.1. The figures are presented separately for four-year natives, community college transfers, and community college entrants who failed to transfer to a four-year institution. There are significant differences between community college entrants who transfer and those who do not. Community college students are most likely to transfer to a four-year institution when they have higher levels of high school achievement, come from more advantaged socioeconomic backgrounds, have higher educational aspirations, attend full time, work fewer hours, enroll continuously, are male, and do not have any children. Not surprisingly, there are also many differences between community college transfer students and four-year natives. Transfer students have lower levels of high school achievement, have taken less academically rigorous high school courses, have lower levels of family income and parent education, and they work more hours per week, on average. First Stage Models I present the results of the first-stage instrumental variables and propensity score models in Tables 2.2 and 2.3. The instrumental variables model in Table 2.2 shows that, as expected, the probability of entering a community college declines as the distance to the closest community college increases. Similarly, the probability of entering a community college increases as the distance to the closest four-year college increases, but does so at a declining rate. Finally, a greater difference in average tuition between community colleges and four-year institutions in the respondent's state increases the likelihood that a student will enter a community college, but also does so at a declining rate. These results are also presented graphically in Figures 2.1 and 2.2, which further display the curvilinear nature of some of these relationships. Students who obtain some postsecondary school and reside closer to community colleges and further from four-year colleges are more likely to enter a community college as their first institution. Students who reside in states where the tuition of public four-year institutions is relatively more expensive than the tuition charged by community colleges are also more likely to enter a community college. The results show that most of the instruments are significantly associated with the decision to attend a two- versus a four-year college suggesting that the model fits the data fairly well. A Hosmer-Lemeshow chi-square test also suggests 49 that there are no significant differences between the observed values of the dependent variables and the values predicted by the model.8 The substantive findings from this model are consistent with earlier work by Rouse (1995) and Anderson et al. (1972) on the influence of college accessibility on college entry. The second column of Table 2.2 shows the effect of college accessibility on the probability of dropping out of a four-year institution, net of first college type. A key assumption behind the IV estimator is that the instruments have no association with the dependent variable net of the endogenous regressor. Therefore, in this case, measures of accessibility should be unrelated to attainment after controlling for first college type. The results from this model show that none of the instruments have significant or large effects on attrition from a four-year institution, net of first college type. This suggests that the key assumption of the instrumental variables model has not been violated. The results of the first-stage propensity score model (Table 2.3) show that, among students with some postsecondary experience, students with higher achievement in high school, higher family socioeconomic background, and higher educational aspirations are less likely to enter a community college than a four-year college. High school achievement mediates much of the negative association between parent income and education and community college attendance but advantaged students remain substantially less likely to enter a community college even after taking achievement differences into account. This is consistent with prior work that suggests that socioeconomically disadvantaged students and those with low high school achievement 8 Breaking subjects into deciles based on predicted probabilities and computing the difference between observed and expected frequencies (with the cutoff set at .5) yields a chi-square of 12.57 with 8 degrees of freedom and a p-value of. 13. Therefore, the null hypothesis that the observed and predicted values are the same cannot be rejected suggesting that the model fits the data at an acceptable level. 50 are more likely to attend community colleges than their more advantaged counterparts (Adelman 2005; Dougherty 1987; Kane and Rouse 1999; Kurlaender 2006; Lee and Frank 1990). Females are also substantially less likely than males to enter a community college but this gap is entirely explained by gender differences in high school achievement. Interestingly, black and Hispanic students who attended any college are significantly more likely than white students to enter a four-year college than a community college, but only after controlling for both social background and achievement measures. The results of these first-stage models are used to control for selection in the four-year attrition models. Transfer Models Next, I examine the relationship between credits earned and transfer to a four-year institution among community college entrants. It is important to keep in mind that only 35% of community college entrants in the sample eventually transferred to a four-year institution. If I just consider community college students who aspire to a bachelor's degree, the transfer rate is 42%. More than half of community college entrants who aspire to earn a bachelor's degree fail to make the transition to a four-year institution. The predicted probabilities of transfer are plotted across the academic credit distribution in Figure 2.3. The probability of transfer increases over time as students accumulate more credits and reaches a high of about 0.6 among students who earn 60 or more credits. As shown in Table 2.1, transfer students spend about three years enrolled in a community college and earn an average of 37 academic credits prior to transferring. The results from models that estimate the hazard of transfer are presented in Table 2.4. Consistent with Figure 2.3, the hazard of transfer increases as students earn more 51 credits. The odds of transfer increase by about 50% for each additional 10 credits earned. The hazard of transfer is higher once students reach their second year of enrollment in the community college but declines if students remain enrolled four or more years. The results also indicate that community college students are most likely to transfer to a fouryear institution when they have a college educated parent, have higher educational aspirations, attend full time, and are male. High school and college achievement do little to mediate the association between parental education and transfer as students from the most educated families remain significantly more likely to transfer after taking these factors into account. Thus, even after controlling for academic achievement having a college educated parent continues to exert a significant independent effect on transfer. Also, although there are small mean differences in proximity to a four-year college between community college entrants who transfer and those who do not (shown in Table 2.1), proximity to a four-year institution is unrelated to transfer net of the social background and achievement measures included in the models. Four-Year College Attrition Next, I descriptively examine the association between community college attendance, credit accumulation and attrition from a four-year institution among transfer students and four-year natives. The results indicate both that community college students accumulate fewer credits than four-year natives and that the odds of attrition decline as the number of credits earned increases. First, four-year natives and community college transfers differ substantially in their rate of credit accumulation. As shown in Table 2.1, after two years of enrollment four-year natives have earned an average of 42 academic credits while community 52 college transfers have only earned 31 academic credits. Figure 2.4 shows that differences in the accumulation of academic credits emerge early in students' postsecondary careers and grow slightly larger over time. The graph shows that across all years of enrollment community college transfers have earned fewer academic credits, on average, than fouryear college natives. For example, while it takes four-year natives roughly three years to earn 60 credits, it takes the average transfer student four years to do so. The credits gap at 3 years (around the time when the average community college student transfers) is roughly ten credits. A similar result is found in Table 2.5 where I show the mean number of terms taken to earn various numbers of credits. Whereas it takes four-year natives roughly four terms to earn 30-40 credits, for example, it takes transfers roughly five terms to do so. These results lend support to my claim that it is inappropriate to study differences in degree attainment by comparing community college transfers and four-year natives who persist until their junior year as several prior studies have done (Lee, Mackie-Lewis, and Marks 1993; Melguizo and Dowd 2008). Upon their initial enrollment in a four-year institution, transfer students have not made as much progress towards their degree as four-year natives. Next, I examine the association between credits earned and attrition from a fouryear institution. In Figure 2.5 I plot the predicted probability of four-year college attrition at different points in the credit distribution. The graph only includes terms in which students are enrolled in a four-year institution. However, transfer students' credits in their first term of enrollment in a four-year school reflect the number of academic credits they earned in a community college. The median number of credits that community college students transfer with is 40, denoted on the graph by a vertical line. For both groups, the 9 This graph looks similar if I use all credits earned rather than only academic credits earned. 53 probability of attrition declines as students accumulate more credits. Comparing students who have earned a similar number of credits suggests that students who transfer early— with below the median number of credits—are substantially more likely to drop out of their four-year institution than are four-year natives. The gap is smaller for students who transfer with 40 or more credits; however, these students still remain more likely to drop out of their four-year school than are their four-year native counterparts at all points in the credit distribution. This finding is consistent with Adelman (2004) who finds that students who start in a community college but earn 10 or fewer credits from the community college before moving on to a four-year school have comparatively low rates of bachelor's degree completion compared to students who start in a community college and earn more than 10 credits before transfer. Similar results are found in Figure 2.6, where I show that the predicted probability of attrition is higher for transfers than fouryear natives across all years of enrollment. For both groups, the probability of attrition declines with each year of enrollment but increases slightly if students remained enrolled five years or more. Table 2.6 presents results from a discrete-time hazard model of attrition from a four-year institution using covariate adjustment to control for selection. The first model includes a variable flagging if the student was a community college transfer (compared to a four-year native) and dummy variables for the year of enrollment. In the second, third and fourth models I add credits earned, measures of social origins, and high school and college achievement, respectively. Turning first to the first column of Table 2.6, the results suggest that when controlling only for the year of enrollment, the marginal odds of attrition are about 44% higher for transfers than for four-year natives (e('365) = 1.44). 54 Similar to Figure 2.6, the results from this model also show that the risk of attrition is highest early in students' postsecondary careers but declines substantially over time as students persist through the first four years of enrollment. For example, the odds of attrition are about 45% lower in the third year of enrollment compared to the first two years of enrollment. Adding the number of credits earned to the model completely mediates the community college effect as the transfer coefficient becomes slightly negative and close to zero. This suggests that transfers are more likely to drop out because they accumulate credits more slowly than four-year natives. The findings from the full model in Table 2.6 suggest that the odds of attrition decline by about 20% for each additional 10 credits earned. Students who earned higher grades in college and those enrolled in a college preparatory curriculum in high school are substantially less likely to drop out of their four-year institution prior to receiving a bachelor's degree. The risk of attrition is also higher following years where students earn a smaller proportion of the credits they attempt either due to late withdrawal from a course or receipt of a non-passing grade. Lastly, students who work more hours per week, those who have a child, and those who do not enroll continuously are also more likely to drop out prior to receiving their degree, which is consistent with prior work (Horn and Carroll 1996). There are no differences in the odds of attrition for students with some part-time enrollment, which is likely because we also control for credits in these models.10 10 In models not shown, I interact the part-time indicator with the year dummies to gauge whether the relationship between years of enrollment and attrition varies by full- versus part-time enrollment status. The interaction terms are quite small relative to their standard errors and suggest no consistent variation in the relationship between years of enrollment and attrition for students who attend full and part-time. The only exception is the interaction between year 4 and part-time enrollment which suggests that the odds of attrition for full-time students may be lower in year 4 while they are roughly zero or slightly positive for students with some part-time attendance. 55 The community college coefficients from the IV and propensity score models are presented in Table 2.7. u The IV results tell a similar story as the results discussed thus far. The only difference in these estimates and those yielded by simply adjusting for various covariates is that the instrumental variables models produce no significant marginal differences in the hazard of attrition. Although the standard errors in the IV models are large, all of the coefficients for community college attendance are very small and close to 0, even prior to controlling for credits earned, family background and high school and college achievement. The propensity score models produce slightly different results, yielding negative community college transfer coefficients across all models. When controlling for the propensity to enter a community college, year of enrollment, and credits earned the hazard of attrition is about 44% lower for community college transfers than four-year natives. This coefficient remains negative but is reduced in magnitude after entering other covariates into the models. Lastly, in results not shown I also test to see whether the relationship between transfer status and attrition varies by race/ethnicity and by parent education. 2 I find little evidence to suggest that the magnitude of the relationship between transfer status and attrition is different for students of different racial/ethnic groups or for students from different socioeconomic backgrounds as none of the interaction terms reach statistical significance. Therefore, among students who transfer, the community college appears to be an effective route to the baccalaureate for students from all racial/ethnic groups and for students whose parents have obtained different levels of education. Discussion 1 2 The full estimates from these models are available upon request. Results from these interaction models are available upon request. 56 The purpose of this study was to evaluate whether community college transfer students experience a penalty to their odds of attaining a bachelor's degree by virtue of having first attended a community college rather than a four-year institution. I improve upon prior work by comparing results using multiple strategies for controlling for selfselection into first college type and by including a much more detailed account of differences in enrollment pathways and credit accumulation between transfers and fouryear natives. Moreover, I also break my analyses into various stages of the attainment process (i.e., selection into first college type, transfer to a four-year institution for community college entrants, and degree attainment within four-year institutions) and am better able to identify the stages where community college students may fall behind initial four-year students en route to a bachelor's degree. Three important conclusions emerge from this study. First, descriptively I find that community college transfer students and initial four-year students differ considerably in the rate at which they accumulate credits. Since I only focus on successful community college transfers (who tend to earn more credits than their community college counterparts who do not transfer), my estimates of the credits gap would likely be much larger had I compared all community college entrants with all four-year entrants. Transfers have earned fewer credits than four-year natives upon initially enrolling in a four-year institution and these differences persist over the course of their enrollment in a four-year college. Because the odds of drop out decline as the number of credits earned increases, it is clear that these credit differences between transfers and four-year natives must be accounted for in order to provide an unbiased comparison of rates of attrition. 57 I also show that many students vary their enrollment intensity over the course of their postsecondary careers which makes traditional controls for full- versus part-time enrollment status problematic. Using a minimum of 12 credits earned per term to capture full-time enrollment, I find that about 80% of transfers have a mix of full and part time enrollment as do 60% of four-year natives (shown in Table 2.1). This provides further evidence that studying attrition longitudinally is preferable, especially in my case given the importance of capturing the substantial differences in enrollment intensity between transfers and four-year natives. Second, the results from the four-year attrition models show that community college students who transfer experience no disadvantage en route to a bachelor's degree compared to their counterparts who initially attend a four-year college after controlling for credits earned. This finding is consistent with other recent work (Lee, Mackie-Lewis, and Marks 1993; Melguizo and Dowd 2008; Melguizo, Kienzl, and Alfonso 2008). Community college students are indeed disadvantaged relative to four-year natives on a variety of factors associated with attrition from a four-year institution. On average, they come from families with lower income and education, have lower levels of achievement in high school, work more hours per week, are more likely to delay and interrupt enrollment, and earn fewer credits each term that they are enrolled. After taking these differences into account, however, they do not experience any additional disadvantage simply by virtue of having attended a community college, as some critics have claimed. These substantive results are similar across alternative selection models and there is even some evidence from propensity score models that transfers are less likely to dropout conditional on the propensity to enter a community college. 58 When taken together these first two findings point to the importance of studying the timing of transfer and degree completion when comparing the outcomes of transfer students and four-year natives. Upon transitioning to a four-year institution, transfer students have not made as much progress toward their degree as their four-year counterparts and, not surprisingly, also take longer to complete their bachelor's degree than do four-year natives. While it took an average of 4.4 years for four-year natives to complete their degree, it took an average of 5.4 years for transfers to earn their degrees. Transfer students may take longer to complete their degrees but they do eventually finish at a similar rate as their four-year counterparts. Third, the results of this study point to the importance of disaggregating the attainment process into multiple stages when studying the community college effect, rather than comparing the outcomes of all initial two- and four-year students. This approach provides a clearer understanding of at which point, if any, community college students fall behind four-year natives en route to the baccalaureate. I show that any disadvantage experienced by transfer students in the hazard of attrition from a four-year institution is explained entirely by differences in enrollment intensity—not by any institutional features of community colleges than inhibit educational attainment. Therefore, the community college appears to be an effective route to a bachelor's degree for students who transfer and seems to be similar in effectiveness for students from different racial/ethnic groups as well as for students from different socioeconomic backgrounds. I show that the most advantaged students (measured by parent income and education) tend to avoid community colleges altogether and those that do enter 59 community colleges are more likely to transfer. Whereas 56% of the community college entrants in my sample with a college educated parent transfer to a four-year institution, only 22% of students whose parents obtained a high school degree or less do so. This transfer advantage for students with a college educated parent remains after accounting for other social background differences, differences in high school achievement, college achievement, aspirations, non-traditional enrollment pathways, and measures of external demands (i.e., hours worked, having a child). Though these factors mediate some of the association between parental education and transfer, the odds of transfer for students with a college educated parent remain about two times as high compared to their counterparts whose parents only obtained some college. Since community college students who transfer to a four-year institution are no more likely than four-year natives to drop out prior to attaining a bachelor's degree, further studying the factors that contribute to disparities in transfer for students from different social backgrounds will have important implications for our understanding of educational stratification. My findings also have implications for my understanding of effect of community colleges on racial/ethnic inequality in education. Although it has been widely documented that Latinos are significantly more likely than whites to start at a community college (Adelman 2005; Kurlaender 2006; Shulock and Moore 2005), I find that these differences are entirely accounted for by differences in family socioeconomic background. In fact, conditional on family background and high school achievement, blacks and Hispanics who obtain some college education are significantly less likely than whites to enter a community college. Moreover, if they do enter a community college they are equally as likely as whites to transfer after controlling for credits earned and 60 family background. This finding is consistent with recent work by Dougherty and Kienzl (2006) who also find no significant differences in transfer rates for blacks and Hispanics compared to whites. If there are indeed institutional features of community colleges that impede attainment that are not artifacts of the selection process it is clear that these effects are exerted within community colleges themselves through their effect on transfer and not are experienced by students who transition to a four-year institution. With this in mind we should turn our attention to improving our understanding of the factors that facilitate transfer among students who aspire to earn a bachelor's degree and to ways in which the transfer gap between advantaged and disadvantaged students can be closed. 61 References: Adelman, Clifford. 1999. "Answers in the Tool Box: Academic Intensity, Attendance Patterns, and Bachelor's Degree Attainment." Washington, D.C. —. 2004. "Principal Indicators of Student Academic Histories in Postsecondary Education, 1972-2000." National Center for Education Statistics, Washington D.C. —. 2005. "Moving into Town—and Moving On: The Community College in the Lives of Traditional-Age Students." Pp. 202: US Department of Education. Available from: ED Pubs. P.O. Box 1398, Jessup, MD 20794-1398. Tel: 877-433-7827; Fax: 301-470-1244; Web site: http://www.edpubs.org. Adelman, Clifford, Bruce Daniel, Ilona Berkovits, and Jeffrey Owings. 2003. "Postsecondary Attainment, Attendance, Curriculum, and Performance: Selected Results from the NELS: 88/2000 Postsecondary Education Transcript Study (PETS), 2000." United States Department of Education, Washington, DC. Alba, Richard and David E. Lavin. 1981. "Community Colleges and Tracking in Higher Education." Sociology of Education 54:223-37. Alfonso, Mariana. 2006. "The Impact of Community College Attendance on Baccalaureate Attainment." Research in Higher Education 47:873-903. Anderson, C.A., M.J. Bowman, and Vincent Tinto. 1972. Where Colleges are and Who Attends: Effects of Accessibility on College Attendance. New York: McGraw-Hill. Anderson, Kristine L. 1981. "Post-High School Experiences and College Attrition." Sociology of Education 54:1-15. Angrist, Joshua. 1993. "The Effect of Veterans Benefits on Education and Earnings." Industrial and Labor Relations Review 46:637-52. Bound, John, David Jaeger, and Regina Baker. 1995. "Problems with Instrumental Variables Estimation When the Correlation Between the Instrments and the Endogeneous Explanatory Variable is Weak." Journal of the American Statistical Association 90:443-450. Brint, Steven and Jerome Karabel. 1989. The Diverted Dream: Community College and the Promise of Educational Opportunity in America, 1900-1985. New York: Oxford University Press. Cameron, A. Colin and Pravkin K. Trivedi. 2005. Microeconomics: Methods and Applications. Cambridge. Chapman, D and Ernest Pascarella. 1983. "Predictors of Academic and Social Integration of College Students." Research in Higher Education 19:295-322. Chen, Xianglei and C. Dennis Carroll. 2007. "Part-Time Undergraduates in Postsecondary Education: 2003-04." National Center for Education Statistics Washington, D.C. Choy, Susan P. 2002. "Access & Persistence: Findings from 10 Years of Longitudinal Research on Students." American Council on Education, Washington, D.C. Clark, Burton R. 1960. "The 'Cooling-Out' Function in Higher Education." American Journal of Sociology 65:569-576. Cohen, Arthur M. 1989. The American Community College. San Francisco: Jossey-Bass. Cohen, Arthur M. and Florence B. Brawer. 2003. The American Community College. San Francisco: Jossey-Bass. 62 DesJardins, Stephen L, Brian P. McCall, Dennis A. Ahlburg, and Melinda J. Moye. 2002a. "Adding a Timing Light to the "Tool Box"." Research in Higher Education 43:83-114. DesJardins, Stephen L., Dennis A. Ahlburg, and Brian P. McCall. 1999. "An Event History Model of Student Departure." Economics of Education Review 18:375390. —. 2002b. "A Temporal Investigation of Factors Related to Timely Degree Completion." The Journal of Higher Education 73:555-581. Dougherty, Kevin J. 1987. "The Effects of Community Colleges: Aid or Hindrance to Socioeconomic Attainment." Sociology of Education 60:86-103. —. 1992. "Community Colleges and Baccalaureate Attainment." The Journal of Higher Education 63:188-214. —. 2002. "The Evolving Role of the Community College: Policy Issues and Research Questions." in Higher Education: Handbook of Theory and Research, edited by J. S. a. W. G. Tierney. Dordrecht, Netherlands: Kluwer Academic. Dougherty, Kevin J. and Gregory S. 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." Teachers College Record 108:452-487. Dynarski, Susan. 2003. "Does Aid Matter? Measuring the Effect of Student Aid on College Attendance and Completion." The American Economic Review 93:279288. Ganderton, Philip and Richard Santos. 1995. "Hispanic College Attendance and Completion: Evidence from the High School and Beyond Surveys." Economics of Education Review 14:35-46. Gonzalez, Arturo and Michael J. Hilmer. 2006 "The Role of Two-Year Colleges in Improving the Situation of Hispanic Postsecondary Education." Economic of Education Review 25:249-257. Grubb, Norton W. 1989. "The Effects of Differentiation on Educational Attainment: The Case of Community Colleges." Review of Higher Education 12:349-74. Hebel, Sara. 2000. "States without Affirmative Action Focus on Community College Transfers. ." Pp. A35-37 in Chronicle of Higher Education, vol. 46(38). Holmstrom, E. and A. Bisconti. 1974. "Transfers from Junior to Senior Colleges." American Council on Education, Washington, D.C. Horn, Laura, Emily Forest Cataldi, and Anna Sikora. 2005. "Waiting to Attend College: Undergraduates Who Delay their Postsecondary Enrollment." National Center for Education Statistics, Washington, D.C. Ishitani, Terry. 2006. "Studying Attrition and Degree Completion Behavior among FirstGeneration College Students in the United States." The Journal of Higher Education 77:861-885. Kane, Thomas J. 1994. "College Entry by Blacks Since 1970: The Role of College Costs, Family Background, and the Returns to Education." Journal of Political Economy 102:878-911. Kane, Thomas J. and Cecilia Elena Rouse. 1999. "The Community College: Educating Students at the Margin between College and Work." Journal of Economic Perspectives 13:63-84. 63 Karabel, Jerome. 1972a. "Community Colleges and Social Stratification." Harvard Educational Review 42:521-562. —. 1972b. "Community Colleges and Social Stratification: Submerged Class Conflict in American Higher Education." Harvard Educational Review 42:521-562. Kurlaender, Michal. 2006. "Choosing Community College: Factors Affecting Latino College Choice " New Directions for Community Colleges 133:7-16. Lee, Valerie E. and Kenneth A. Frank. 1990. "Students' Characteristics that Facilitate the Transfer from Two-Year to Four-Year Colleges." Sociology of Education 63:178193. Lee, Valerie E., Christopher Mackie-Lewis, and Helen M. Marks. 1993. "Persistence to the Baccalaureate Degree for Students Who Transfer from Community College." American Journal of Education 102:80-114. Leigh, Duane E and Andrew M. Gill. 2003. "Do Community Colleges Really Divert Students from Earning Bachelor's Degrees?" Economics of Education Review 22:23-30. McCormick, Alexander C. 1997. "Transfer Behavior Among Beginning Postsecondary Students: 1989-94." National Center for Education Statistics, Washington D.C. Melguizo, Tatiana. 2008. "Are Community Colleges an Alternative Path For Hispanic Students to Attain a Bachelor's Degree?" Teachers College Record Forthcoming. Melguizo, Tatiana and Alicia C. Dowd. 2008. "Baccalaureate Success of Transfers and Rising Four-Year College Juniors." Teachers College RecordForthcoming. Melguizo, Tatiana, Gregory S. Kienzl, and Mariana Alfonso. 2008. "Comparing the Educational Attainment of Community College Transfer Students and Four-Year Rising Juniors Using Propensity Score Matching Methods." Mercer, Joye. 1992. "States Turn to Community Colleges as Route to Bachelor's Degree as 4-Year Campuses Face Tight Budgets and Overcrowding." in Chronicle of Higher Education. National Center for Education Statistics. 2004. "National Education Longitudinal Study: Base Year Through Fourth Follow-Up, 1988-2000: Data File User's Manual." Washington D.C. National Center for Public Policy and Higher Education. 2006. "Measuring Up: The National Report Card on Higher Education." National Center for Public Policy and Higher Education. Pascarella, Ernest T and Patrick T. Terenzini. 1991. How College Affects Students: Findings and Insights from Twenty Years of Research: Jossey-Bass. Pascarella, Ernest T. and D.W. Chapman. 1983. "Validation of a Theoretical Model of College Withdrawal." Research in Higher Education 19:25-48. Reynolds, John, Mike Stewart, Ryan MacDonald, and Lacey Sischo. 2006. "Have Adolescents Become Too Ambitious? High School Seniors' Educational and Occupational Plans, 1976 to 2000." Social Problems 53:186-206. Richardson, Jeanita W. 2005. "Who Shall be Educated?: The Case of Restricting Remediation at the City University of New York." Education and Urban Society 37:174-192. Rouse, Cecilia. 1994. "What to do After High School: The Two-Year Versus Four-Year College Enrollment Decision." Pp. 59-85 in Choices and Consequences: 64 Contemporary Policy Issues in Education, edited by R. G. Ehrenberg. Ithica, New York: ILR. Rouse, Cecilia Elena. 1995. "Democratization or Diversion? The Effect of Community Colleges on Educational Attainment." Journal of Business & Economic Statistics 13:217-224. Shulock, Nancy and Colleen Moore. 2005. "Diminished Access to the Baccalaureate for Low-Income and Minority Students in California: The Impact of Budget and Capacity Constraints on the Transfer Function " Educational Policy 19:418-442. Singer, Judith D. and John B. Willet. 2003. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press. Sturrock, Carrie. 2003. "Cal State Begins to Enforce Math, English RemediationFreshmen Told Not to Return Until They Complete Community College Classes." Pp. aOl in Contra Costa Times (Walnut Creek, CA). Tinto, Vincent. 1993. Leaving College: Rethinking the Causes and Cures of Student Attrition. Chicago, IL: The University of Chicago Press. Velez, William. 1985. "Finishing College: The Effects of College Type." Sociology of Education 58:191-200. Wellman, Jane. 2002. "State Policy and Community College-Baccalaureate Transfer." National Center for Public Policy and Higher Education, San Jose. Winship, Christopher and Stephen L. Morgan. 1999. "The Estimation of Causal Effects from Observational Data." Annual Review of Sociology 25:659-706. Zumeta, William and Deborah Frankle. 2007. "California Community Colleges: Making them Stronger and More Affordable." The National Center for Public Policy and Higher Education. Student Background White (Omitted) Black Hispanic Other Race Female Parent Education: 21.34 24.05 20.78 0.25 % Enrolled in Chem or Physics in Prior Year 14.07 15.84 13.71 0.25 High school zip code characteristics 25.41 % 25+ College or More 29.49 24.57 0.25 2.30 2.59 2.24 0.25 Median Housing Value (100000s) 13.16 11.39 13.52 Median HH Income (1000s) 0.25 6.32 % Below Poverty 7.16 7.34 0.25 % Unemployment 50.87 55.33 49.95 0.25 Mean Number of Observations Per CC, Per Cohort 950 Observations 604,607 102,828 501,779 Note: 15% of the students missing high school and high school zip code information attended a high school that was private, alternative, outside of California, or outside of the United States. High school characteristics Enrolled Full Time Ever Took a Basic Skills Course Student Took ESL Class or Received ESL Services Age (17-21) Other Race Hispanic Black Demographic Background Male Entered in 1997 Entered in 1996 Entered in 1995 Entered in 1994 Cohort (1992 Omitted) Entered in 1993 -0.121** (0.007) -0.467** (0.018) -0.409** (0.011) 0.495** (0.009) -0.339** (0.004) -0.091** (0.022) -0.512** (0.007) 0.704** (0.008) 0.134** (0.013) 0.049** (0.013) -0.085** (0.013) 0.285** (0.013) 0.310** (0.013) -0.088** (0.010) -0.637** (0.039) -0.292** (0.021) 0.482** (0.019) -0.340** (0.004) -0.104** (0.022) -0.504** (0.011) 0.703** (0.008) 0.134** (0.013) 0.050** (0.013) -0.085** (0.013) 0.285** (0.013) 0.310** (0.013) Table 3.2: Student-Level Log it Model Predicting Transfer Within 6 Years. Used to Recover Predicted Probability of Transfer Student Student+ Measures Interactions -0.091** (0.010) -0.660** (0.040) -0.300** (0.022) 0.477** (0.019) -0.335** (0.004) -0.126** (0.023) -0.531** (0.011) 0.715** (0.009) 0.130** (0.013) 0.052** (0.013) -0.097** (0.014) 0.275** (0.013) 0.303** (0.013) Student+ CC Fixed Effects Black*Basic Skills Black*Free Lunch Interactions Black*Male Median Household Income, in 1000s % Unemployed % in Poverty Median Housing Value Owner Occupied Units, in 100,000s High school zip code characteristics % in Pop 25+ With College or More % Enrolled in Chemistry or Physics in Prior Year % Enrolled in Advanced Math (Above Algebra II) in Prior Year % Grads UC/CSU Eligible in Prior Year % Dropouts in Prior Year % Free Lunch Eligible in Prior Year % Hispanic in Prior Year % Black in Prior Year Student-Teacher Ratio 0.006** (0.000) 0.024** (0.005) 0.004** (0.001) -0.004* (0.002) 0.002** (0.000) 0.004** (0.001) -0.003** (0.000) -0.002** (0.000) -0.001** (0.000) -0.023** (0.001) 0.003** (0.000) 0.003** (0.000) 0.004** (0.001) 0.229** (0.033) 0.002 (0.001) 0.019 (0.033) 0.006** (0.000) 0.023** (0.005) 0.004** (0.001) -0.004* (0.002) 0.002** (0.000) 0.004** (0.001) -0.003** (0.000) -0.001** (0.000) -0.001 ** (0.000) -0.023** (0.001) 0.004** (0.000) 0.003** (0.000) 0.004** (0.001) 0.227** (0.033) 0.002* (0.001) 0.048 (0.033) 0.004** (0.001) 0.010 (0.007) 0.003* (0.001) -0.007** (0.002) 0.001 (0.001) 0.014** (0.001) -0.005** (0.001) -0.002** (0.000) 0.000 (0.000) -0.021** (0.001) 0.004** (0.000) 0.003** (0.000) 0.006** (0.001) 44754.39 183.43** 604607 44570.96 604607 .83 35 .83 26 47942.40 3188.01** 604607 .83 141 -0.098** -0.098** (0.019) (0.019) -0.001** -0.001* (0.001) (0.001) -0.037 -0.048* (0.020) (0.020) -0.102** -0.103** (0.017) (0.017) 0.002** 0.003** (0.001) (0.001) 0.014 0.001 (0.018) (0.018) 3.797** 4.085** 4.078** (0.094) (0.090) (0.091) -7.542e+06 -7.54E+06 -7.542e+06 White is the omitted racial/ethnic group. 1992 is the omitted cohort. BIC Proportion Correctly Classified df Chi-squared Liklihood-ratio test Observations Standard errors in parentheses * significant at 5%; "significant at 1% Constant Other Race*Basic Skills Other Race*Free Lunch Other Race*Male Hispanic*Basic Skills Hispanic*Free Lunch Hispanic*Male Table 3.3: Descriptives by Clusters of Community Colleges. Clusters Formed based on Attributes of Student Bodies Cluster 1 Cluster 2 9 19 Frequency Student Predicted Probability of Transfer 0.09 0.13 Observed % of Students Transferring 0.10 0.15 0.56 0.34 Proportion of Students in Quintile 1 Proportion of Students in Quintile 2 0.24 0.26 Proportion of Students in Quintile 3 0.11 0.19 Proportion of Students in Quintile 4 0.05 0.13 Proportion of Students in Quintile 5 0.03 0.08 College Attributes Miles to Closest CSU 23.69 43.61 Miles to Closest UC 20.65 6.93 Urban 0.21 0.56 0.26 Rural 0.11 Percent Black 0.28 0.10 0.43 Percent Hispanic 0.58 UC Acceptance Rate 38.41 60.07 61.42 62.91 Percent Transferable Courses Percent Basic Skills Courses 10.27 15.68 Percent Part-Time Faculty 63.74 53.15 High School Characteristics 16.91 8.40 Percent Black 65.27 48.22 Percent Hispanic Percent Free Lunch Eligible 49.32 34.50 3.85 Percent Dropout 6.65 27.85 Percent UC/CSU Eligible 37.76 Percent Enrolled in Advanced Math 17.74 21.30 12.58 Percent Enrolled in Chem or Physics 12.99 High School Zip Code Characteristics Percent Ages 25+ College Graduates 16.26 16.89 Median Housing Value, in 100000s 1.88 1.66 Percent in Poverty 21.50 16.54 13.42 11.71 0.67 0.08 64.39 18.98 30.73 29.25 3.66 39.02 22.22 15.38 35.04 2.83 14.62 6.58 31.08 25.38 2.93 27.52 19.04 12.08 18.23 1.58 15.27 8.41 61.51 8.12 65.60 58.38 67.67 0.41 0.07 0.27 0.18 0.25 42.40 68.28 12 0.17 0.22 0.20 0.20 0.20 0.19 0.21 22 0.15 0.16 0.19 0.24 0.24 0.20 0.13 28.24 0.23 Cluster 4 Cluster 3 20.90 13.61 0.11 0.05 0.05 0.16 46.69 76.51 5.68 63.03 49.84 20.73 26.27 2.26 11.24 5.00 20.27 19.81 2.41 34.53 23.04 14.43 0.16 59.97 68.20 6.77 68.02 0.19 0.27 0.05 38.55 3.70 7.84 5.40 20.39 15.73 2.05 38.85 26.76 18.28 Cluster 6 19 0.22 0.25 0.07 0.13 0.18 0.25 0.36 26 0.19 0.20 0.11 0.18 0.23 0.25 0.23 Cluster 5 Percent Unemployed Median Household Income, in 1000s 10.42 37.64 9.16 42.80 8.81 44.93 6.53 49.63 6.20 51.86 4.51 66.95 Enrollment Urban Rural Miles to UC Number UCs in 30 Miles Miles to CSU Number CSUs in 30 Miles Av Acceptance Rate, UC % Courses Offered that were Transferable % Courses Offered that were Remedial % Part-time Faculty Ratio of FTE Counselors/Advisors Per Stu (92-93) (n) Expenditures on Transfer Center (2001-02) Other Transfer Expenditures (2001-02) Instructional Expenditures (2001-02) Admissions Expenditures (2001-02) Counseling Expenditures (2001-02) Financial Aid Expenditures (2001-02) Articulation Services Survey (2002-03) Number of Full Time Equivalent Articulation Staff College has Articulation Plan/Goals for Art. Articulation Goals Coordinated with Transfer Center Plan Articulation Process is Well Coordinated or Seem less Sufficient Course to Course Articulation with UCs Sufficient Course to Course Articulation with CSUs Sufficient Articulation for Major Prep with UCs Sufficient Articulation for Major Prep with CSUs Transfer Center Survey (2002-2003) Transfer Agreements with More than 3 CSUs (n) Bot20 1.12 0.47 0.68 0.47 0.53 0.53 0.42 0.37 0.25 (4) 1.53 0.64 0.71 0.51 0.61 0.71 0.55 0.48 0.13(44) 0.97 0.70 0.75 0.40 0.70 0.70 0.60 0.55 0.11 (27) 0.15(13) 1.84 0.67 0.70 0.55 0.60 0.76 0.57 0.49 10917 10775 11622 8699 0.27 0.21 0.20 0.55 0.22 0.10 0.18 0.50 28.84 40.30 40.75 50.26 0.75 0.70 0.79 0.65 18.82 13.28 15.91 34.13 2.21 2.00 2.00 2.33 53.52 56.99 51.31 53.14 68.07 70.13 68.50 64.56 7.75 8.41 9.55 8.27 63.90 63.51 63.98 64.01 1473(64) 1352(11) 1450(38) 1667(13) 21.52 20.50 22.60 18.48 10.54 9.12 9.60 15.03 2747 2674 2900 2723 114.54 93.79 120.88 114.02 266.92 291.81 250.26 297.86 73.35 64.48 70.40 92.64 Table 3.4: Comparison of Means for Community College Characteristics for High and Low Ranking Schools. Rankings based on Magnitude of Coefficients of Colleges' Independent Effect on Transfer All Top20 Middle 0.43 (44) 0.64 (45) 0.69 (45) 0.75 (45) 0.42 (45) 0.75 (4) 0.60 (5) 0.80 (5) 1.00(5) 1.00(5) 0.40 (27) 0.74 (27) 0.67 (27) 0.85 (27) 0.44 (27) 0.38(13) 0.46(13) 0.69(13) 0.46(13) 0.15(13) missing for colleges, the numbers in parentheses reflect the number of observations used to compute the mean. Notes: Expenditure measures reflect expenditures per FTE student. In cases where a significant amount of data are Transfer Agreements with More than 3 Ucs (n) Goal Met: Transfer Support Services to Students (n) Goal Met: Ensure the Academic Planning for Transfer (n) Goal Met: Students receive transfer information (n) Goal Met: Monitor Student Progress until Transfer (n) Table 3.5: OLS Regression of Coefficients for Community Colleges from Fixed Effects Model on CC Characteristics Enrollment, in 1000s -0.002 (0.004) Urban Location 0.077 (0.058) Rural Location -0.147 (0.090) Miles to CSU -0.004* (0.002) Miles to UC 0.001 (0.001) Acceptance Rate of Closest UC 0.002 (0.002) Percent Transferable Courses 0.004 (0.002) Percent Basic Skills Courses 0.008 (0.006) Percent Part-Time Faculty 0.002 (0.002) Constant -0.495 (0.261) Observations 107 R-squared 0.18 Adjusted R-squared 0.10 BIC -468.188 Standard errors in parentheses * significant at 5%; ** significant at 1 % 24.43 5.49 18.54 16.84 1.59 42.88 28.62 18.88 36.23 3.12 9.02 5.15 62.87 24.49 5.68 19.78 18.01 1.70 41.45 27.73 18.28 34.83 2.99 9.43 5.35 61.12 Fifth Quintile No TranTransfer 0.45 0.46 0.01 0.01 0.03 0.04 0.41 0.41 17.63 17.63 0.02 0.02 0.20 0.16 0.46 0.51 0.23 0.10 *: 0.15 0.15 * 0.21 0.21 *' 0.32 0.34 Predicted Probability of Transfer 0.16 0.05 0.06 * 0.10 n 110983 12021 105248 17745 97390 25600 82013 40977 511905 103068 116271 6725 critical t 4.400 4.209 4.209 4.209 4.209 4.209 *strong evidence for sig difference in mean between those who transfer & do not transfer- using BIC; critical values for test statistic calculated as sqrt(ln(n)+6) following Raftery (1995) year and interactions between race and year, %free lunch, gender, and remedial course were also used to predict the propensity score. Appendix 3.1: Means of Independent Variables by Transfer Status & Quintiles of Predicted Probability of Transfer Prior to Strat First Quintile Second Quintile Third Quintile Fourth Quintile No TranTransfer IMo TranTransfei' No TranTransfer No TranTransfer No TranTransfer 0.50 0.52 0.51 0.48 * 0.52 0.50 0.51 0.50 0.48 0.48 Male 0.08 0.16 0.12 0.07 0.02 Black 0.05 * 0.18 * 0.13 0.06 0.03 0.67 * 0.44 0.44 0.22 0.32 0.17 * 0.68 0.22 0.10 0.10 Hispanic 0.17 Other Race 0.27 * 0.02 0.03 * 0.07 0.09 * : 0.16 0.19 * 0.26 0.26 18.21 17.91 * 18.93 18.81 * 18.32 18.25 *' 18.09 18.05 * 17.87 17.87 Age (17-21) 0.04 0.04 0.04 0.03 * 0.05 0.05 * 0.04 0.03 0.03 Limited English Proficient 0.03 Ever took Remedial Course 0.53 0.38 * 0.81 0.82 0.66 0.68 0.51 0.53 * 0.36 0.35 0.27 * 0.03 Full-Time Attendance 0.15 0.03 * 0.07 0.07 0.10 0.11 0.17 0.18 HS Characteristics 24.79 24.63 * 25.11 25.14 Student-Teacher Ratio 24.88 24.84 24.75 24.77 24.63 24.63 7.98 6.72 * 10.82 10.72 8.81 8.69 7.24 7.30 6.37 6.29 % Black 29.77 30.09 24.58 24.36 % Hispanic 34.29 26.47 * 53.18 52.09 * 38.00 38.31 27.55 22.41 * 39.51 38.96 * 29.95 30.22 24.76 25.10 21.58 21.44 % Free Lunch Eligible 2.55 * 7.29 6.39 * 3.86 3.79 2.88 2.90 2.24 2.24 % Dropouts 3.78 32.68 36.75 * 27.01 26.87 29.99 30.05 32.28 32.38 35.56 35.69 % Grads UC/CSU Eligible 20.76 24.06 * 15.47 15.95 * 18.78 19.04 20.93 21.05 23.29 23.33 % in Advanced Math 13.72 13.81 15.28 15.26 % in Chem or Physics 13.73 15.86 * 10.51 10.91 * 12.38 12.59 HS Zip Code Characteristics % 25+ College or More 24.57 29.47 * 16.37 16.42 21.64 21.51 24.99 25.08 28.62 28.85 2.47 1.79 2.02 2.21 2.21 2.46 Med Housing Value , in 100000s 2.25 2.59 * 1.79 2.02 12.53 12.59 10.95 10.92 % Below Poverty 13.50 11.34 * 18.38 18.00 * 14.55 14.61 6.29 * 9.44 7.87 7.97 6.93 6.92 6.15 6.08 % Unemployment 7.30 9.40 49.99 55.40 * 40.92 41.47 * 46.70 46.60 50.61 50.62 54.54 54.55 Median HH Income, in 1000s 134 Appendix 3.2: Transfer Rates using Alternative Definitions of those at-risk Transfer % who Transfer Rate but do not meet criteria 0.17 0 One Transfer Course 0.24 0.15 10+Transfer Units 0.28 0.23 20+ Transfer Units 0.31 0.28 30+ Transfer Units 0.34 0.33 40+ Transfer Units 0.36 0.39 50+ Transfer Units 0.37 0.52 60+ Transfer Units Figure 3.1. Proportion of Variance Explained by Various Cluster Solutions 1 t -a 0.94> | 0-8- • CL x 0.7 - • V * •• -^-Q1 -*--Q2 s Q3 x Q4 o 0.6 (0 / fe 0.5 > o 0.4 - f f | 0.3 o §• 0.2£ 0.1- * 0 -r 0 4 6 8 10 12 Number of Clusters Figure 3.2. Kernel 2Density Plot of Observed Institutional Transfer Rates Kernal Density Plot of Observed Transfer Rates : • « * ' • liCsi-H .05 .1 .15 .2 .25 Proportion of Students Observed Transferring Figure 3.3. Distributions of Proportion of Students each College Enrolled in each Quintile of the Predicted Probability of Transfer iiipislflisip •fe'cJ'N mm .;.:;.;.0:;- :: ^v_ X: ; '':.:'K?.:: : *"" •Kjstfl:;* , . . ' : . ' « . : _ ; • . • A s.i/.feW::'.N'8'»*• • • • / • ,.-,,:;!p .,,, .15 .2 .25 .3 % Transferred Cluster 5 : \ \ / ':.... : ( s j.. .1 V / ,,:. .15 .2 .25 .3 i Transferred Cluster 4 \ ...>15:::,.„2:':.';;j.25:..::...:.:3,..:.. 35 ::::«;:%;TrarisfefmdCluster6 :::: Clusters of community colleges were formed based on tfie attfiiutes of the students they enroll Figure 3.6. Coefficients and Confidence Intervals for the 20 Schools that Have the Most Positive Impact on Transfer, Net of Student Attributes Coeffs/CIs for 20 Top Schools O b o o 0[...]... College- Baccalaureate Transfer. " National Center for Public Policy and Higher Education, San Jose Zumeta, William and Deborah Frankle 2007 "California Community Colleges: Making them Stronger and More Affordable." The National Center for Public Policy and Higher Education 20 CHAPTER 2: Community College Transfer and Baccalaureate Attainment 21 Abstract: By facilitating transfer to four-year institutions, community colleges... community college students aspire to a bachelor's degree, most do not receive any type of degree, few transfer to a four-year college and their rates of bachelor's degree attainment trail far behind those of students who initially enter four-year colleges (Brint and Karabel 1989; Dougherty 1987; Dougherty and Kienzl 2006; Karabel 1972a; Lee and Frank 1990) However, the types of students who enter community. .. baccalaureate degree among students who transfer Second, I examine variation in transfer rates among colleges and seek to understand what characteristics distinguish schools that are relatively more effective in facilitating transfer among their students Third, I analyze the transfer process and the outcomes of students who initially attend four-year colleges but later transfer to a community college prior... four-year college Understanding the role of community colleges in social stratification is an important policy concern yet questions still remain about the efficacy of community colleges in improving the educational and economic outcomes of their students In this project I study pathways to and from community colleges both nationally and in California I study student and institution-level predictors of transfer. .. lateral transfer to another four-year institution, and reverse transfer to a community college The key questions for my purposes are, given the decision to 13 terminate enrollment in one's initial four-year institution, what differentiates those who transfer to other four-year colleges, those who transfer to a community college and those who dropout entirely? Are similar students better served by transferring... Mackie-Lewis, and Marks 1993; Melguizo 2008; Melguizo and Dowd 2008) that community college students who transfer are no less likely to complete a bachelor's degree than otherwise similar students who begin in a four-year college, students who do transfer appear to be a relatively advantaged group relative to the population of community college attendees (Dougherty and Kienzl 2006; Lee and Frank 1990)... struggle academically while in college are significantly more likely to reverse transfer or dropout from their initial four-year institution However, upon enrolling in a community college those who reverse transfer are able to increase their GPAs, some transfer back to a four-year college and attain a bachelor's degree, and others earn an associate's degree Some reverse transfers ultimately leave school... Arthur M 1989 The American Community College San Francisco: Jossey-Bass Dougherty, Kevin J 1987 "The Effects of Community Colleges: Aid or Hindrance to Socioeconomic Attainment. " Sociology of Education 60:86-103 18 — 1992 "Community Colleges and Baccalaureate Attainment. " The Journal of Higher Education 63:188-214 — 2002 "The Evolving Role of the Community College: Policy Issues and Research Questions."... results and shed new light on the impact community colleges have on the educational outcomes of their students Literature Review Community colleges offer a wide variety of degree programs and have diverse missions that range from the provision of transfer education to the provision of adult, remedial, and vocational education (Dougherty 2002) The ambitions of community college attendees are diverse and. .. background—not by any institutional features of community colleges that inhibit educational attainment Therefore, the community college appears to be an effective route to a bachelor's degree for students who transfer The effect of community colleges on the attainment of students who fail to transfer, however, remains unclear More than half of the students who enter college shortly after high school graduation