From High School to College: Gender, Immigrant Generation, and Race-Ethnicity [1 ed.] 0871544180, 9780871544186

Citation preview

FROM HIGH SCHOOL TO COLLEGE

FROM HIGH SCHOOL TO COLLEGE GENDER, IMMIGRANT GENERATION, AND RACE-ETHNICITY

Charles Hirschman

Russell Sage Foundation  •  New York

The Russell Sage Foundation The Russell Sage Foundation, one of the oldest of America’s general purpose foundations, was established in 1907 by Mrs. Margaret Olivia Sage for “the improvement of social and living conditions in the United States.” The foundation seeks to fulfill this mandate by fostering the development and dissemination of knowledge about the country’s political, social, and economic problems. While the foundation endeavors to assure the accuracy and objectivity of each book it publishes, the conclusions and interpretations in Russell Sage Foundation publications are those of the authors and not of the foundation, its trustees, or its staff. Publication by Russell Sage, therefore, does not imply foundation endorsement. BOARD OF TRUSTEES Sara S. McLanahan, Chair Larry M. Bartels Karen S. Cook W. Bowman Cutter III Sheldon Danziger Kathryn Edin

Lawrence F. Katz David Laibson Nicholas Lemann Martha Minow Peter R. Orszag

Claude M. Steele Shelley E. Taylor Richard H. Thaler Hirokazu Yoshikawa

Library of Congress Cataloging-in-Publication Data Names: Hirschman, Charles, author. Title: From high school to college : gender, immigrant generation, and race-ethnicity / Charles Hirschman. Description: New York : Russell Sage Foundation, 2016. Identifiers: LCCN 2016006327| ISBN 9780871544186 (paperback) | ISBN 9781610448574 (ebook) Subjects: LCSH: Educational equalization--United States. | College preparation programs--United States. | BISAC: EDUCATION / Secondary. | SOCIAL SCIENCE / Ethnic Studies / General. | SOCIAL SCIENCE / Emigration & Immigration. Classification: LCC LC213.2 .H57 2016 | DDC 379.2/6--dc23 LC record available at http://cp.mcafee.com/d/1jWVIg43qb3z1EVuvsosdTdFTd7bVEVjuujd79JeVEVvd7 arNEVhodFTd7bVEVjuuphdEETsd7bzb3O9JnoSY-EazjifY01dNFF7-00C_juZ5Ys_Rhd7dXILetuVtdBZXCnT3hOqenzhPR4kRHFGTshVkffGhBrwqrhdECXYyUqenAkmkXL6XCOsVHkiP3XmxNZrshGpNeWbBXKcFzztGlr1hZrshGpdILK9CMnVClfgd40A6xoQg8rfjh0Xm9Ewd78VWXMV5Zxx5VMSOMrN_Sj Copyright © 2016 by Russell Sage Foundation. All rights reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Reproduction by the United States Government in whole or in part is permitted for any purpose. The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences—Permanence of Paper for Printed Library Materials. ANSI Z39.48-1992. Text design by Genna Patacsil. RUSSELL SAGE FOUNDATION 112 East 64th Street, New York, New York 10065 10 9 8 7 6 5 4 3 2 1

In Memory of Lynne Toshiye Taguchi 1963 to 2013

Contents

List of Illustrations About the Author

ix xvii

Preface xix Acknowledgments xxiii Chapter 1

The Role of Education in American Society: Expanding Opportunity and Persistent Inequality

1

Chapter 2

Recent Trends in College Graduation: The National Portrait

34

Chapter 3

The University of Washington-Beyond High School Project: Data and Description

60

Chapter 4

The College Pathways Model

116

Chapter 5

Social Origins and College-Pathway Transitions 153

Chapter 6

A Closer Look at the Role of Culture in Explaining Educational Transitions

194

Chapter 7

Work and Extracurricular Activities in the Lives of High School Seniors

239

Chapter 8

The Impact of Schools and the Promise of Scholarships on College-Pathway Transitions 273

viii  Contents

Chapter 9

Summing Up: Pathways to College Graduation

308

Notes 341 References 353 Index 371

List of Illustrations

Figure 2.1

Figure 2.2

Figure 2.3

Figure 2.4

Figure 2.5

Figure 2.6

Figure 2.7

Percent of High School Graduates and College Graduates in the U.S. Population, Aged Twenty-Five to Twenty-Nine, Based on the CPS Series from 1940 to 2013

35

Percent of College Graduates in the U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty), Based on the 1990 and 2000 Censuses and the 2010 American Community Survey

40

Percent of College Graduates in the Native-Born U.S. Population by Race-Ethnicity and Birth Cohort (Year Cohort Turned Age Twenty)

45

Percent of College Graduates and Educational-Transition Ratios for the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty)

49

Percent of High School Graduates in the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

51

Educational-Transition Ratios for the Transition from High School Graduation to College Enrollment in the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

52

Educational-Transition Ratios for the Transition from College Enrollment to College Completion in the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

54

x   List of Illustrations

Figure 3.1

Average Annual Number of High School Students Enrolled in School District 1, by Grade Level: 1997–1998 to 2004–2005

76

Figure 4.1

The College Pathways Model

121

Figure 4.2

The College Pathways Model of Educational-Transition Ratios with All Possible Paths

122

Figure 4.3

The Linear-Progression College Pathways Model of Educational-Transition Ratios

126

Figure 5.1

Hypothesized Model of Ascription, Social Origins, High School GPA, Encouragement, and College-Pathway Transitions

156

Logistic Regression of College Aspirations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors

163

Logistic Regression of College Expectations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Aspiring UW-BHS High School Seniors

169

Logistic Regression of College Preparedness on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Expecting UW-BHS High School Seniors

172

Logistic Regression of Enrollment in a Four-Year College (NSC Only) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of CollegePrepared UW-BHS High School Seniors

176

Logistic Regression of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors

177

Figure 5.2

Figure 5.3

Figure 5.4

Figure 5.5

Figure 5.6

List of Illustrations   xi

Figure 5.7

Logistic Regression of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Entrant UW-BHS High School Seniors

181

Logistic Regression of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors Who Did Not Enroll in a Four-Year College

187

Figure 6.1

A Model of Cultural Context, Cultural Orientations, and Cultural Expressions

200

Figure 6.2

A Revised College Pathways Model with Cultural Context, Cultural Orientations, and Cultural Expressions

216

Logistic Regression of the Probability of College Aspirations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors with Cultural Variables

222

Logistic Regression of the Conditional Probability of College Expectations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Aspiring UW-BHS High School Seniors with Cultural Variables

224

Logistic Regression of the Conditional Probability of College Preparedness on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Expecting UW-BHS High School Seniors with Cultural Variables

226

Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC Only) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors with Cultural Variables

229

Figure 5.8

Figure 6.3

Figure 6.4

Figure 6.5

Figure 6.6

xii   List of Illustrations

Figure 6.7

Figure 6.8

Figure 7.1

Figure 7.2

Figure 7.3

Figure 7.4

Figure 7.5

Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors with Cultural Variables

230

Logistic Regression of the Conditional Probability of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College Entrant UW-BHS High School Seniors with Cultural Variables

233

A Revised College Pathways Model of Ascription, Social Origins, Employment and Participation in Extracurricular Activities, Cultural Variables, and College-Pathway Transitions

243

Logistic Regression (Selected Models) of the Probability of College Aspirations on Student Employment and Participation Among UW-BHS High School Seniors

257

Logistic Regression (Selected Models) of the Conditional Probability of College Expectations on Student Employment and Participation Among College-Aspiring UW-BHS High School Seniors

260

Logistic Regression (Selected Models) of the Conditional Probability of College Preparedness on Student Employment and Participation Among College-Expecting UW-BHS High School Seniors

262

Logistic Regression (Selected Models) of the Conditional Probability of Enrollment in a Four-Year College (NSC Only) on Student Employment and Participation Among College-Prepared UW-BHS High School Seniors

265

List of Illustrations   xiii

Figure 7.6

Figure 7.7

Figure 8.1

Figure 8.2

Figure 8.3

Figure 8.4

Figure 8.5

Figure 8.6

Logistic Regression (Selected Models) of the Conditional Probability of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Student Employment and Participation Among College-Prepared UW-BHS High School Seniors

266

Logistic Regression (Selected Models) of the Conditional Probability of College Graduation (in Seven Years) on Student Employment and Participation Among College-Entrant UW-BHS High School Seniors

269

Logistic Regression of the Probability of College Aspirations on Type of High School Among UW-BHS High School Seniors

296

Logistic Regression of the Conditional Probability of College Expectations on Type of High School Among College-Aspiring UW-BHS High School Seniors

297

Logistic Regression of the Conditional Probability of College Preparedness on Type of High School Among College-Expecting UW-BHS High School Seniors

299

Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC Only) on Type of High School Among College-Prepared UW-BHS High School Seniors

301

Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Type of High School Among College-Prepared UW-BHS High School Seniors

302

Logistic Regression of the Conditional Probability of College Graduation (in Seven Years) on Type of High School Among College-Entrant UW-BHS High School Seniors

303

xiv   List of Illustrations

Table 2.1

Percent of College Graduates in the U.S. Population by Birth Cohort Based on Age Groups in 1990, 2000, and 2010: Based on the 1990 and 2000 Population Censuses and the 2010 American Community Survey

38

Racial, Ethnic, and Nativity Composition of the U.S. Population by Birth Cohorts from 1920–1924 to 1980–1984

42

Decomposition of Intercohort Change in College Graduation Rates Attributable to Educational-Transition Ratios for High School Graduation, Transition to College, and College Completion in the Native-Born U.S. Population by Race-Ethnicity

56

Core Respondents in the UW-BHS Baseline Survey and in the One-Year Follow-Up Survey: 2000 to 2005

67

Coverage of the UW-BHS Survey of High School Seniors by Enrollment Status and Yearbook Listing: Class of 2000 in District 1

70

Coverage Rates and Composition of the Total Sample of UW-BHS High School Seniors by Enrollment Status and Yearbook Listing: 2000 to 2005

74

Comparison of College Enrollment One Year After High School Between UW-BHS Follow-Up Survey and Administrative Records from the National Student Clearinghouse (NSC)

84

Table 3.5

Alternative Measures of Race and Ethnicity Among UW-BHS High School Seniors

90

Table 3.6

Decomposition of Intergroup Gaps in GPA (Actual and Self-Reported) by Washington Assessment of Student Learning (WASL) Test Scores and the Residual (Effort) Among UW-BHS High School Seniors in District 1

112

Percent of All High School Seniors at Each Stage of the College Pathways Model and Educational-Transition Ratios by Gender, Race-Ethnicity, and Immigrant Generation: UW-BHS High School Seniors

127

Table 2.2

Table 2.3

Table 3.1

Table 3.2

Table 3.3

Table 3.4

Table 4.1

List of Illustrations   xv

Table 4.2

Table 4.3

Table 4.4

Table 5.1

Alternative Estimates of Enrollment in a Four-Year College by NSC and UW-BHS Follow-Up Survey Data

141

Percent of College Graduates Among UW-BHS High School Seniors via Linear Progression and Alternative Pathways

143

Decomposition of Intergroup (Gender, Race-Ethnicity, and Immigrant-Generation) Gaps in College Graduation Rates by Stages in the College Pathways Model

147

Logistic Regression of the Probability of College Aspirations for UW-BHS High School Seniors

161

Table 5.2

Alternative Pathways to College Graduation of UW-BHS Seniors Who Did Not Enroll in a Four-Year College Right After High School, by Gender, Race-Ethnicity, and Immigrant Generation 184

Table 6.1

Correlations Among Cultural Variables (Context, Orientations, and Expressions) for UW-BHS High School Seniors

204

Mean Values of Cultural-Context, Cultural-Orientations, and Cultural-Expressions Variables and Correlations with Gender, Race-Ethnicity, and Immigrant Generation Among UW-BHS High School Seniors

208

Logistic Regression of the Probability of College Aspirations Among UW-BHS High School Seniors with Cultural Variables

218

Paid Employment, Participation in Sports Activities, and Participation in Extracurricular (Nonsports) Activities by Gender, Race-Ethnicity, and Immigrant Generation of UW-BHS High School Seniors

245

Logistic Regression of the Probability of College Aspirations for UW-BHS High School Seniors with Employment and Participation Variables

253

Number of UW-BHS Schools and Respondents by Type of School and Year

280

Table 6.2

Table 6.3

Table 7.1

Table 7.2

Table 8.1

xvi   List of Illustrations

Table 8.2

Table 8.3

Table 8.4

Characteristics of UW-BHS Respondents by Type of School and Period: 2000 and 2002–2005

283

College-Pathway Transitions by Type of School for UW-BHS High School Seniors

286

Logistic Regression of College-Pathway Transitions by Type of School and Additional Covariates for UW-BHS High School Seniors

292

About the Author

Charles Hirschman is Boeing International Professor in the Department of Sociology and the Daniel J. Evans School of Public Policy and Governance at the University of Washington.

Preface

T

his book is the product of the University of Washington-Beyond High School (UW-BHS) project, a collaborative research program that I directed for more than a decade. My initial plans were modest— it was to be an “add-on” project lasting only a few years. But the thrill of fieldwork, interesting new research questions, and funding from external grants soon moved the UW-BHS project to center stage on my research agenda. The data-collection component grew incrementally, from five schools for one year to twelve schools for five years, and larger research budgets meant more management of staff and students, more obligations to our collaborating schools and other agencies, and a more ambitious research agenda. At times, the project seemed more in charge of me than I of it. Although I often wish that I could have wrapped up the UW-BHS study sooner, the long life of the project has had some benefits. Most importantly, the results and interpretations presented in this work offer a more comprehensive and substantial contribution than what I could have produced in the early years of the project. My preliminary analyses showed familiar patterns of inequality that had been found time and time again in the literature, and my efforts to interpret and explain the research findings seemed remarkably pedestrian. For some, inspiration and insight may arrive in a blinding flash, but for me, it was more of a gradual process based on rereading classical studies and endlessly revising papers that seemed incomplete and unsatisfactory. Now that it is done, I believe that this study holds the promise of a new and potentially cumulative direction for future research on educational inequality and stratification. This rather immodest claim requires some context and justification. Few topics have been more intensively studied than American education. Policy-oriented studies typically find fault with the performance of schools, teachers, students, and even the entire system. Social-science researchers (my tribe) generally document the pervasive inequality in

xx  Preface

educational opportunities and outcomes, but they rarely offer unambiguous prescriptions for policy or even for understanding the complex mechanics of school processes and outcomes. Most empirical studies report that the educational performance and attainment of students are highly correlated with every measurable background variable, including socioeconomic status, family structure, and race and ethnicity—whether measured at the individual level or in terms of school and neighborhood composition. Longitudinal research often shows that academic performance at one level of schooling is highly correlated with subsequent performance, likelihood of dropping out, and graduation. Moving from describing these relationships to explaining them, however, has been a daunting challenge. With everything related to everything else, attention is drawn to long lists of variables, questions of measurement, and predictive power rather than to a conceptual framework for the field. Further compounding the problem is the oversized role ideology plays in educational research. To some degree this is driven by the real need for educational reform and the assumption that there must be a simple “fix” for schools that will minimize, if not eliminate, educational disparities. For example, many policy prescriptions begin with claims that schooling would be greatly improved by changing teacher supply, training, and incentives. Some of these proposed strategies may well be effective, and all are deserving of empirical investigation, but I fear that many are promoted because of hunches, a priori assumptions, and ideology rather than clear evidence of their effectiveness. Another obstacle to cumulative research on educational inequality has been the frequent assumption that contemporary American schools only serve to replicate intergenerational socioeconomic status and to discipline students to conform to society’s normative culture. American schools are, however, multifaceted institutions that serve a variety of functions, some of which are contradictory. In addition to reproducing inequality—“status maintenance”—schools also create opportunities for many students from modest means and immigrant origins to rise far above their socioeconomic origins. The American belief in equal educational opportunity, the redistributive fiscal model of public schooling, and the meritocratic process of student evaluation—however imperfectly these ideals are realized in practice—does offer a challenge to the seemingly ubiquitous critiques of American education. Too often, we compare social outcomes with ideals rather than observed variations across time and space. My objective in writing this book is to nudge American educational research away from ideological debate and toward the development of analytical models that allow for cumulative empirical research—the comparative advantage of social science. The experimental method is the gold standard in all research, but it is often impossible to design

Preface  xxi

experiments to measure the many important dimensions of educational inequality. Statistical models are the alternative, but they invariably require making assumptions about causality in a simplified model of the temporal process of schooling. To paraphrase the renowned statistician George Box, simplifying assumptions and models may be wrong (incomplete), but it can be useful. The first key assumption of the UW-BHS study was that the associations between our three primary measures of ascription—gender, race-ethnicity, and immigrant generation—and educational outcomes were not causal but needed to be explained through their joint corre­ lations with other background variables having direct effects on education. The rationale for this assumption was the lack of any a priori or logical reason to posit that gender, race-ethnicity, or immigrant generation “cause” inequality in educational outcomes. In contrast, families with greater social and economic advantages have the resources to train, coach, and subsidize their children’s education relative to poorer families. This assumption allows for the analysis and interpretation of the proximate reasons for disparities in educational outcomes for the three key ascriptive background characteristics. The second assumption is that educational disparities between ascriptive groups that are mediated by academic performance are a function of family coaching and encouragement (and not of between-group differences in ability). This shifts attention from the direct individual-level correlation between grade point average (GPA) and educational attainment (highly significant, but theoretically uninteresting) to the question of how “learned” academic performance mediates educational disparities between ascriptive groups. These assumptions and their implications are presented at greater length in chapter 1 and throughout the book. This work also presents a conceptual (and measurable) framework for analyzing the longitudinal “process” of educational attainment. Unlike many other behavioral outcomes, education is not an event but the cumulative product of schooling that can extend over twelve, sixteen, twenty, or more years. The roots of college failure may not occur during college but many years earlier. To address this complex issue, I propose a primitive sequential model of key turning points along the road to college graduation. The College Pathways Model identifies five steps or stages: (1) college aspirations, (2) college expectations, (3) college preparation, (4) college enrollment (in a four-year college), and (5) college completion (with a baccalaureate degree). Each stage is conceptualized as a transition and measured as a conditional probability from those who have completed the prior stage. Like all models, the College Pathways Model is a simplification of the complexities of schooling, but it provides a framework for assessing the impact of inequality at each stage on gender, race-ethnicity, and immigrant-generation disparities

xxii  Preface

in college graduation rates. The results provide a preliminary guide to where resources might be most effectively applied in order to reduce ascriptive inequality in college graduation rates. The value of this approach might well be its replacement by superior models in future research. The availability of the UW-BHS data archive at the Inter-university Consortium for Political and Social Research (ICPSR) data library at the University of Michigan allows the research community to reanalyze the research questions posed in the study (as well as new questions) and to extend, modify, or discard the conclusions and inter­pretations offered in this volume. Better science and scientifically informed policy depend on such progress.

Acknowledgments

A

I am the author of this book, I have been surrounded by an army of colleagues, graduate and undergraduate students, and staff who worked on the UW-BHS project at the University of Washington. They collaborated with me on almost every phase of the project, from the study design and development of ques­tionnaires to the coordination with our participating schools and sub­sequent data analyses. For this reason, I use the first-person plural—we—in the text. I have written the words and take responsibility for all errors of commission and omission, but much of the credit for the ideas, measurement techniques, and analytical methods goes to my collaborators and students. The following paragraphs are a small, but very sincere, expression of my gratitude. In the early stages Susan Brown, now professor of sociology at University of California-Irvine, was the graduate-student research assistant on the project, but her actual role was more that of a colleague than a student. Over the years, a number of other University of Washington graduate students have also played central roles in the UW-BHS project, including Amon Emeka, Lynne Taguchi, Irina Voloshin, Karen Brooks, Maya Magarati, Liz Ackert, and Deleena Patton. Tony Perez worked with me as a postdoctoral researcher for several years on a related project, but his fingerprints are on many aspects of this study, especially the creation of the UW-BHS race-ethnicity classification. The person with the longest attachment to the project is Nikolas PharrisCiurej. Nick joined the project as a first-year graduate student and was part of the field-research team administering surveys in schools. Over the years, Nick became the statistical guru on the project and taught me various analytical methods, while I nominally served as his academic advisor. After completing his PhD, Nick continued as a postdoctoral researcher on the project, and we initially planned for this book to be a collaborative project. His career, however, took him to full-time employment in “the other” Washington (D.C.), and our collaboration slowed. Nick read the lthough

xxiv  Acknowledgments

preliminary drafts of many of the chapters and corrected many of my errors. His joint authorship of chapters 2 and 3 is a partial recognition of his imprint on many of the ideas and analyses that are reported in the volume. For many years, Patty Glynn’s job title was computer programmer on the UW-BHS project. In reality, Patty did everything, including creating and managing the complex UW-BHS data archive, training graduate students, and advising me on all matters, large and small. One indicator of Patty’s many contributions is that the UW-BHS data are so well organized that I am able to conduct analyses on my own. Patty also managed the complex undertaking of preparing the UW-BHS data for archiving at the Inter-university Consortium for Political and Social Research (ICPSR) data library at the University of Michigan. The project has also benefited from superb administrative and organizational support by Carolyn Liebler, Julie Miller, and Michele Hanzeli. Faculty colleagues have also been invaluable sources of support, encouragement, and advice over the years. During the initial stages of the UW-BHS project, Richard Morrill, Robert Plotnick, Rob Warren, and Paul LePore served as coinvestigators. As the project expanded in the early 2000s, several University of Washington colleagues joined the UW-BHS team, including Gunnar Almgren, Jerald Herting, Stew Tolnay, and Paula Nurius. The early design of the project was shaped in large measure by a distinguished advisory board consisting of Adam Gamoran, Robert Hauser, and Alejandro Portes. A number of graduate students and visiting scholars contributed as research assistants or analyzed UW-BHS data (or both), including Jennifer Holsinger, Christine Fountain, Jennifer Lee, Jeanine Schmitz, Phuong Nguyen, Peter Graham, Sang Thanh Lee, Liz Mogford, Rosanna Lee, Brian Hammer, Geoffrey Palmer, Tim Lum, Patrick Ishuzuki, Stephanie Ewert, Andrew Overmyer, Allison Dunne, and Dung Thi Kieu Vu. Undergraduate research assistants made major contributions to the success of the project, including Alisa Vuong, Jason Thomas, Pam Hageman, Marilyn Sinkewicz, Julie Warden, Garth Almgren, Michele Hanzeli, Ashley Isaksen, Kia Sorensen, Oluwatope Fashola, Cara Biddlecom, Laila Thomkinson, Mindy Szeto, Anthony Poon, Duc Ngo, Gioia Rizzo, Richard Wang, Stephanie Carson, Vincente Cordova, Kelsi Cuniff, Kristen Hess, Naomi Lam, Helen Mak, Letizia Malone, Shannon Murphy, Elinor Nicolson, Olivia Thai, and Sione Lister. Regina Shaw worked as an undergraduate research assistant on the project during the last year. She revised all the graphics, organized the references, and did almost everything else short of writing the text. All of this work could not have been accomplished without generous research support from both external and internal sponsors. The UW-BHS project began with two grants from the Andrew W. Mellon Foundation,

Acknowledgments  xxv

“The Impact of Changes in Affirmative Action Policy on the Transition from High School to College in Washington State” in 1999 (19900692) and “Explaining Race and Ethnic Inequality in the Transition from High School to College in Washington State: The Impact of I-200 and Beyond” in 2001 (10100685). I will be forever grateful to Harriet Zuckerman, vice president at the Mellon Foundation, for her encouragement and steadfast support. In 2003 the UW-BHS project data collection expanded to include two new school districts and three private schools with a grant from the Bill and Melinda Gates Foundation (22954) on “The Effects of Scholarships and School Reform on the Transition from High School to College in Washington State.” My meetings with Deborah Wilds (then senior program officer at the Gates Foundation) and her colleagues, including William Gates, Sr., inspired me to undertake more ambitious goals for the project. We were also generously supported by multiple grants from the Population Dynamics Branch (formerly the Demographic and Behavioral Sciences Branch) of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), including “Concepts and Measures of Race and Ethnic Identities” (NICHD 1R01HD04728901A2) and two supplemental grants, “Research Supplement to Promote Diversity in Health-Related Research for Concepts and Measures of Race and Ethnic Identities” (NICHD 3R01HD047289-02S1) and “ARRA Supplement for Concepts and Measures of Race and Ethnic Identities” (NICHD 3R01 HD047289-05S1). My home institution, the University of Washington, and the Dep­art­ ment of Sociology and the Center for Studies in Demography and Ecology (CSDE) in particular were enormously supportive. The Department of Sociology generously provided office and research space for project staff and students. CSDE provided an excellent computing environment and administrative support for the project, thanks to the NICHD research infrastructure grant R24 HD042828 and a training grant, T32 HD007543. Matt Weatherford, CSDE computing director, was a source of advice and technical support for me and the CSDE research team. The late Fred Nick, the founding director of the University of Washington’s Center for Social Science Computation and Research, was a fount of wisdom on computers, computing, and data management. Fred’s counsel was invariably wise and helpful. The research reported in chapter 2 is based on national census and American Community Survey (ACS) data that are processed and distributed by the Integrated Public Use Microdata Series (IPUMS-USA) project at the University of Minnesota.1 Over the years I have presented preliminary results from the UW-BHS project at countless seminars, conferences, and events at the University of Washington and many other venues. The questions, comments, and criticisms from colleagues and students reassured me that I was on the

xxvi  Acknowledgments

right track, even though my analyses and interpretations were often found wanting. I dare not name names because their numbers are legion. Michael White, a distinguished professor at Brown University, read the manuscript with a fine-tooth comb and saved me from countless errors, large and small. Good friends are rare, and good friends who are also serious critics are rarer yet. Michael is both. As a reader and an author, I consider the Russell Sage Foundation to be my favorite publisher. The foundation publishes much of the best work by leading researchers in the social sciences. Books published by the foundation are carefully reviewed and edited, attractively designed, actively marketed, and kept in print for decades. Suzanne Nichols, director of publications, keeps the wheels turning with measured words of support and encouragement, but she does not hesitate to scold delinquent authors. Both techniques were used adroitly to help me finish this work. Suzanne also manages to find experts who write thorough and detailed constructive reviews of book manuscripts. My reviewers caught more than a few errors, identified weaknesses in my arguments, and recommended major reorganization of some materials (and chapters). Although I alone am responsible for all remaining errors in this book, the foundation’s review process greatly improved the final product. The research was only possible because numerous school administrators and teachers allowed my colleagues and me to administer the baseline UW-BHS senior survey in twelve high schools in three school districts from 2000 to 2005. Some of my introductory letters and phone calls to school offices went unanswered, and the first responses were often skeptical of the value of academic research. The workdays of school administrators and teachers are overstretched with overwhelming demands and limited resources, and visits from “ivory tower” academics are sometimes seen as another burden. However, after a few meetings, the university research team and our colleagues in the public and private schools in our sample (the school administrators and teachers) developed considerable mutual respect and even close personal bonds. Our project team was even enlisted to conduct additional research on the needs of the schools. Our sincere gratitude goes to all our collaborators in the public and private schools, and especially to Joe Willhoft, Robin Munson, and Patrick Cummings. Our deepest debt goes to the almost ten thousand high school seniors who completed the baseline survey and then responded to follow-up surveys. Almost all students took the survey seriously and responded to detailed questions about their post–high school plans and social background. Many even wrote nuanced responses in the margins of the questionnaires and wrote additional comments in a space on the last page. Reading these comments and having conversations with students during the data collection reminded me of the pathos in the lives of so many

Acknowledgments  xxvii

young people. Survey questions about the characteristics of fathers, or father figures, and the college plans of friends presume that students have a family and peers with whom they are engaged. More than a few high school seniors had neither. The educational challenges faced by students often reflect even deeper problems experienced by adolescents preparing for adulthood with too little economic and social support. Writing this book, and indeed my entire career, has only been possible because of the love and support I receive from my family. Jo is my wife, my best friend, and my favorite traveling companion. Our children, Andrew and Sarah, have grown up, but they and their families, including Jen, Edmund, Charlie, Coleman, Judah, and Josie, are an active presence in our lives and a reminder that there is more to life than research. This book is dedicated to the memory of Lynne Taguchi, a brilliant graduate student and wonderful colleague who did not live to see the completion of this project. Lynne and I frequently discussed her interest in the educational attainment of the third generation (grandchildren of immigrants). Do they continue to experience the “second-generation edge,” or does the impact of immigrant ancestry fade away? Inspired by Lynne’s ideas and example, I will try to honor her memory by pursuing her research question on the third generation in the coming years.

Chapter 1 The Role of Education in American Society: Expanding Opportunity and Persistent Inequality

A

college degree does not guarantee success. Nothing really does—there is too much happenstance in life to say what is best for every person. But if we had to choose one attribute that would carry greater weight than any other, it would be educational attainment. The economic fault line between high school and college graduates in the United States is wider than ever—college graduates have, on average, earnings 75 percent higher than those of high school graduates.1 Higher education is also highly correlated with a lower probability of divorce, better health, and higher satisfaction with work and family life. These facts are not lost on young people. National surveys show that the overwhelming majority of high school students desire to attend and graduate from college.2 The reality is, however, that most of these students will not achieve their dreams. At the end of formal schooling, only about one-third of young adults attain a bachelor’s degree. The aim of this book is to explain the inequality in educational outcomes—college graduation, in particular—by three fundamental ascriptive characteristics: gender, race-ethnicity, and immigrant generation. Sociologists use the term ascription to refer to characteristics that are generally fixed at birth and rarely change over the life course. In a fair world, such “accidents of birth”—not subject to choice—should not limit how far one can go in life. The democratic theory of equality of opportunity suggests that all children should have the right and freedom to go as far as their abilities and efforts allow them. Despite widespread belief in the theory of equal opportunity, the reality is that ascription is highly correlated with what happens later in life.

1

2   From High School to College

Scholars and laypersons alike have ideas and hunches about the “real” reasons for the strong association between ascription and educational attainment. In the University of Washington-Beyond High School (UW-BHS) study, we attempted to empirically test the leading hypotheses for educational inequality between men and women, racial and ethnic groups, and immigrant generations. At the top of the list, for the last two categories, was the hypothesis that the social class or socioeconomic status (SES) of families in which students are reared explains much of the educational disparities by ascription. (In this work, we use the terms social class and socioeconomic status interchangeably.) Other questions pursued were the presumed role of socialization and the transmission of cultural influences by families, peers, and communities. We also explored the role of student employment, participation in extracurricular activities, and the characteristics of schools as mediators of educational inequality. Our aim was not to “explain away” the impact of ascriptive characteristics but to clarify why ascription continues to matter in schooling, which is generally considered one of the most meritocratic of all institutions. Our focus was on access to and completion of higher education, that is, earning a bachelor’s degree (BA or BS). High school graduation remains a serious problem in American society. The conventional estimate of a 90 percent high school graduation rate is clearly inflated; most researchers put the figure at 70 to 75 percent.3 But increasingly, college is what matters. The majority of high school graduates—upwards of 65 to 70 percent—begin postsecondary schooling, but only half of those who enroll in college earn a bachelor’s degree. Successful completion of college largely depends on where you begin. Community colleges and vocational schools offer educational opportunities to many students, most of whom do not have opportunities to attend a traditional four-year college.4 However, the likelihood of completing college with a baccalaureate degree is more than five times higher for students who begin college in a four-year institution than for those who begin in a two-year institution.5 Our study was based on the assumption that college graduation is a process that extends from adolescence to the early years of adulthood. The process is illustrated with the metaphor that the path to college graduation is akin to grabbing on to a fast moving train that begins in high school. College graduation—reaching the final destination—is most likely for students who hang on to the train despite some fast turns and bumps along the way. Our conceptualization of this process is the College Pathways Model, which is described and explained in greater detail in chapter 4. According to the model, most college graduates begin with high aspirations to complete college, have confidence that their aspirations will be realized, prepare for college while in high school, enroll in a four-year college right after high school, and persevere in completing college. Some students who fall off the college-bound train manage to climb back on,

The Role of Education   3

but the odds are against it. We also include an analysis of those who pursue “alternative pathways” to college graduation. The College Pathways Model is a useful framework for analyzing how each step in the process contributes to disparities in college graduation rates by gender, raceethnicity, and immigrant generation. In chapters 1 and 2, we set the stage with an overview of the role of education in American society and recent trends in college graduation. In chapter 3, we explain the origins of the UW-BHS project and describe the data and methods used to conduct the research and interpret the results. The project was a longitudinal study of about ten thousand high school seniors from twelve high schools in the Pacific Northwest. Despite the geographical and temporal limits of these data, we suggest that the results speak to national issues and concerns. Chapters 5 through 8 present the analyses of how socioeconomic origins, child rearing and culture, student employment and participation, and type of high school affect each stage of the College Pathways Model by gender, race-ethnicity, and immigrant generation. In chapter 9, we present a summary of the major findings of the study and their implications for the public debate over educational opportunity and persistence of inequality.

Why Education Matters One of the dominant images of American history is the little red schoolhouse. The one-room building with students of all ages was a symbol of the nineteenth-century American value that formal schooling was a priority even in frontier communities. By 1840, perhaps as many as 40 percent of youths aged five to nineteen years were enrolled in some form of schooling in the United States.6 Mass education did not become common in most other countries until well into the twentieth century. Education continues to be a priority, perhaps the number-one priority, in contemporary American society. But the one-room schoolhouse is long gone. Perhaps the modern equivalent might be a sprawling, sub­ urban high school campus with multimedia classrooms, varied educational programs and counseling services, and expansive athletic fields. With half or more of high school students going to some sort of postsecondary schooling, the American landscape is dotted with institutions of higher education from small picturesque undergraduate colleges to mega-universities that enroll tens of thousands of students. Some universities have more students than the population of medium-sized cities. Even more numerous than traditional baccalaureate degree–granting colleges and universities are the thousands of community colleges, technical schools, and other institutions that lead to two-year (associate) degrees, certification, and vocational training.

4   From High School to College

The number of students, teachers, administrators, support staff, researchers, and related personnel involved in the American educational sector is staggering. In the fall of 2013, eighty-five million persons, more than one in four Americans, were either students or employees in an educational institution.7 The largest component of this number is the fifty-five million students enrolled in elementary or secondary schools. In addition, twenty-three million students are enrolled in postsecondary degree–granting institutions, and about ten million persons are teachers, faculty, or professional, administrative, and support staff employed by educational institutions. If we define educational services more broadly to include technical and trade colleges and other schools, thirteen million workers were employed in educational services in 2013—over 9 percent of the American workforce.8 There is also a huge monetary investment in schooling in the United States. The National Center for Education Statistics (NCES) reports that over one trillion dollars was spent on elementary, secondary, and tertiary (only degree-granting institutions) education in 2012—about 7 percent of the total U.S. gross domestic product (GDP).9 Direct general expenditures for public schooling constitute almost one-third of the budgets of all state and local governments in the United States.10 The value of education is even greater than the sum of its current expenditures. Schooling is literally an investment in the future. Although individuals pursue education for many reasons, including the love of learning, the dominant motive is to prepare for a remunerative and productive career. Indeed, attending school is virtually synonymous with growing up, and leaving school is generally associated with entering the workforce. Children are expected to enter the age-graded schooling system at age five (increasingly at age three or four) and are generally required to continue until age sixteen. Many persons continue to enroll in school well into their mid-twenties or later. The objectives of elementary and secondary schooling—literacy, numeracy, and basic knowledge of science, history, literature, and other fields—are considered to be essential skills for employment in the modern economy and for active citizenship in society. The importance of education extends far beyond the provision of basic skills taught in primary and secondary schools. Colleges and universities train students in advanced knowledge and specialized skills in liberal arts, science, engineering, medicine, business management, and many other fields. The continuous production of new graduates is vital to the functioning and advancement of the many complex and technologically sophisticated sectors of the modern economy. With rapidly changing technology, universities equip students with the skills necessary to continue learning throughout their careers. Colleges and universities, however, do much more than transfer knowledge across generations; they also

The Role of Education   5

produce new knowledge. Many, if not most, of the important technological innovations of recent years are based on research conducted at universities.11 While it is still possible for creative individuals to produce new inventions in a “garage,” most contemporary discoveries and technological innovations are derived from basic scientific research, the corpus of knowledge produced by researchers in universities or by university-trained researchers in laboratories, industries, or government agencies. The huge investments in education, both individually and collectively, in American society are premised on the role of educational institutions in creating and transmitting knowledge across generations. Economic historians estimate that the growth of human capital—the knowledge produced by schooling—accounts for at least 25 percent of economic growth, especially in modern developed countries.12 Samuel Preston attributes most of the gains in world life expectancy from the 1930s to the 1960s to the increased knowledge that led to the understanding and control of disease.13 Although educational expenditures must compete with other national priorities, the widespread belief that schooling and basic research are investments for future well-being has contributed to the growth of the education sector in the United States. Public investment in education is also a response to the popular demand for schooling. Individuals may not be certain about the precise economic payoff, but they know that, in general, obtaining more education leads to better jobs, higher incomes, and more rewarding lives. Research by social scientists strongly supports the conventional view that more schooling is the key predictor of individual socioeconomic attainment.14 The Census Bureau estimates that over the course of a fortyyear career, a worker with a bachelor’s degree will earn a million dollars more than a high school graduate.15 The finding of a high economic rate of return to years of schooling is robust across a broad range of empirical studies.16 Human capital theory posits that the high correlation between educational attainment and income is causal: more years of schooling represent skills and knowledge that will lead to higher economic productivity that is rewarded directly (to entrepreneurs) or indirectly by firms that offer higher wages to workers who contribute to productivity. Another interpretation is that education is just a mechanism for the transmission of status, wealth, and power across generations. These are not mutually exclusive interpretations. Families do indeed seek to pass along socio­economic advantages to their children, and providing educational opportunities is a means of doing so. The current evidence indicates that education plays a dual role; it is both a means of passing along highstatus positions across generations and a channel for upward mobility for students from disadvantaged backgrounds.17

6   From High School to College

The benefits of schooling are not limited to labor-market returns. A 2011 review of the nonpecuniary benefits of schooling with data from the NORC at the University of Chicago General Social Survey found that persons with higher levels of schooling enjoy advantages in health, happiness, family stability, and a broad array of social and psychological outcomes.18 Not all of the associations between schooling and positive outcomes are causal; other factors may be the joint determinant of both. However, most of the positive effects of schooling held up in multivariate models with statistical controls for socioeconomic background and other covariates. One of the strongest findings from recent research is that more highly educated people are healthier and live longer than persons with less schooling.19 In addition to a general association with SES and improved access to health care, education appears to be associated with a greater sense of control over behaviors and habits that contribute to health and longevity. Individuals with more education tend to smoke less, engage in fewer risky behaviors, experience less stress, exercise more, and have healthier diets. Highly educated women were once less likely to marry, but this tendency has reversed in recent decades. Since 1970, women, both white and African American, with college degrees have become more likely to marry and less likely to divorce.20 Cohabitation has become a normative feature of American family formation, but it is more prevalent among those with a high school diploma (or less) than among persons who have attended or graduated from college.21 These trends are accentuated by a growing social divide by educational attainment. Since 1970, college graduates have become much more likely to marry other college graduates than persons with lower educational attainments.22 These familyformation patterns have effects on the next generation. The children of less-educated parents are more likely to experience a parental divorce or have a parent with a cohabiting partner, or both.23 In general, highly educated Americans also report higher levels of happiness, life satisfaction, and social and civic engagement.24 The belief that these effects are “caused” by higher education is reinforced by evidence that persons whose backgrounds would not have predicted they would go to college experience the most positive effects of a college education.25 Although the positive effects of education, collectively and individually, are interpreted as an American phenomenon, the lesson is increasingly universal. In recent years, the growth of enrollments in higher education in other countries has exceeded that of the United States. Understanding how this has happened requires a look backward to the history of schooling in the United States.

The Role of Education   7

A Historical Perspective on American Schooling With industrialization and political modernization transforming societies during the second half of the nineteenth and early twentieth centuries, the functions and structure of educational institutions also began to change. In modernizing societies, schools, generally financed by public authorities, began to assume some of the responsibilities of families by training children and adolescents for careers. The transformation of the occupational structure meant that children were less likely to follow the vocations of their parents, especially in agriculture. Workers needed to learn new skills of literacy and numeracy, as well as discipline and punctuality, to work in factories and other urban pursuits. Initially, schools in industrializing countries reflected the customs and prejudices of their feudal past. Schools were segregated by social class and mission. The schools for the children of workers and peasants were much more vocational than academic. There was relatively little emphasis on schools as channels of social mobility.26 Gradually, the mission of education changed from reproducing inherited privilege among elites to sorting students by aptitude and achievement for roles in an increasingly complex economy.27 These changes were most evident in the post–World War II era, as modern states assumed responsibility for managing their economies and insuring the social welfare of their populations. Modern industries needed highly skilled, flexible, and motivated workers. The role of schooling in modern societies was expanded to provide almost universal primary and secondary education and to give an increasing share of students the opportunity to go to college. In their magisterial history of American schooling, Claudia Goldin and Lawrence Katz claim that the United States had a modern school system right from the start.28 They identify three distinct periods in the expansion of American schooling: the spread of the common school in the nineteenth century, the “high school movement” in the early twentieth century, and the expansion of public universities and colleges in the early post–World War II era. Several distinctive organizational features radically differentiated American schooling from that of most other industrializing countries. The American tradition was typically a decentralized system of largely public, secular schools that were open to all, forgiving of failure, and gender neutral. These were not absolutes—there were many religious institutions; some schools were very elitist; and female students were not always welcome. But the overwhelming majority of students who attended common schools (one-room schools that emphasized primary education), flocked to secondary schools in the years before World

8   From High School to College

War II, and filled college classrooms in the 1950s and 1960s were participants in a mass educational experiment that had few, if any, precedents. The non-elitist structure and culture of American education were quite unusual in the nineteenth and twentieth centuries. By its very nature, schooling creates a hierarchy based on competition among students, who are generally ranked by their performance. High-achieving students and their schools are often lauded while others are labeled as second- or thirdrate. Educators and administrators tend to be drawn from the “winners” of educational competition, and they generally favor a system with a small number of students at the highest level that rewards people like themselves. The elitist model of education is also reinforced by cost considerations. It is very rare for the costs of schooling to be borne completely by students and their families. Even elite educational systems tend to be heavily subsidized, both directly through state revenues and indirectly through grants and tax exemptions. As discussed earlier, education is an investment that leads to higher incomes among graduates and higher rates of economic growth for society. But these are long-term benefits, and in the short run, the costs of salaries (for teachers, administrators, and support staff) and the construction and maintenance of facilities (buildings, libraries, laboratories, stadiums) must be paid for with taxpayer funds. The total educational budget is less when educational opportunities are rationed within an elite system of schooling. The key attribute of elite educational systems is competitive exams that restrict enrollments to relatively small numbers in secondary schools and even fewer at the tertiary level. A smaller educational system can generally afford to subsidize those at the higher rungs. For example, an elite higher-education system can be staffed with a small number of master teachers or senior professors, assisted by acolytes. There are also fewer demands for accountability or practicality in an elitist system of higher education. In fact, the content of classical elite schooling was generally focused on philosophical, literary, religious, and other abstract topics. The elitist character of national educational systems in most countries, including advanced European countries, persisted until fairly recent times, well into the twentieth century. In contrast to the standard model of an elitist educational structure, the American educational system evolved with an inclusive democratic ethos. Goldin and Katz argue that the exceptionalism of American public schooling is largely due to decentralized states and localities, which taxed themselves to provide common schools in the nineteenth century and (nearly) universal high schools in the early decades of the twentieth century. Although there are some signs that the practical and economic values of schooling were motivating forces (such as the creation of the landgrant college system that emphasized agriculture in the 1860s), American secondary and tertiary schooling generally had an academic rather than

The Role of Education   9

vocational curriculum. An interesting feature of American secondary schools is that enrollments were often higher in rural areas and smaller towns than in large industrial cities. Broad-based support for American schools, based on local and state taxes, was reinforced with the openness to all youth. An elite educational system with high failure rates would probably not have generated popular support and public funding. The exclusion of African American students in the South (and Asian Americans in some West Coast localities) was a notable exception to the overall pattern of mass schooling that generally accepted women and newly arrived immigrant groups from Europe. Mass schooling at the tertiary level is a more recent phenomenon in the United States. During much of the nineteenth century, only about 1 percent of college-age youth attended institutions of higher education, and the figure was still below 5 percent in the 1920s.29 There was a group of elite colleges in the nineteenth century, primarily in the Northeast. In other regions, especially in the Midwest, most colleges displayed the features of openness that characterized American primary and secondary schools, with subsidized tuition, flexible schedules, and minimal entrance requirements.30 Many colleges were church based, but the majority of college students attended public institutions. Teacher training and other practical pursuits, including agriculture, engineering, and business, became central missions of publicly funded colleges in the early twentieth century.31 The takeoff in college enrollment—the beginning of mass tertiary education—started in the late 1940s. In the mid-1940s, less than 10 percent of American men and 5 percent of women were college graduates.32 By the mid-1970s, a quarter of young men and one-fifth of young women were college graduates. College attendance rates were much higher, generally hovering around 50 percent. A significant contributor to the jump in college enrollment in the late 1940s was the enrollment of more than two million veterans, many of them from working class origins, with benefits from the GI Bill.33 The surge in college enrollment in the three decades after World War II was, however, much broader and systemic with the expansion of publicly funded state colleges and universities in the 1950s and 1960s. Goldin and Katz credit the creation of a mass secondary-school system during the first half the twentieth century with being an essential prerequisite for the expansion of college enrollments.34 But perhaps more important was the growing role of scientific knowledge and training in a modern society and the quintessential American belief in the value of public education. President Harry Truman’s Commission on Higher Education recommended dramatically expanding government support and scholarships for students to attend community colleges, liberal arts colleges, and graduate schools. In a prescient statement, the Commission reported that “higher education is an investment, not a cost” in a democratic society.35 It also boldly called

10   From High School to College

for abolition of the restrictions and quotas that limited participation of African Americans and Jews in higher education. The growth in college-student enrollment, as well the renown of American universities, in the post–World War II era, was primarily due to greater investment in public (state-supported) institutions. Before this point, the most highly reputed colleges in the United States were private institutions, and the elite private universities (the Ivy League, Johns Hopkins, and Stanford) were generally at the top of national rankings of prestige and quality. Over the course of the twentieth century and especially following World War II, public universities rose to the top. At present, 75 percent of all undergraduate students are enrolled in public institutions,36 and the majority of major research universities are public institutions.37 The rise of public colleges and universities can be most clearly understood by examining budgets. In 1949 the expenditures of all postsecondary degree-granting institutions in the United States was only 0.8 of 1 percent of the national GDP, a ratio that was unchanged from 1939 and barely higher than the 0.6 percent in 1929.38 This figure rose dramatically to 2.3 percent of the GDP in 1975, during an era of robust economic growth. Although the budgets of college and universities continued to expand in absolute terms over the next few decades, the relative share of spending on higher education stagnated at 2.2 to 2.5 percent from the 1970s to the 1990s. The relative share of GDP spent on higher education did rise to over 3 percent by 2010. The rise and subsequent slowdown of public investment in higher education were also evident in the affordability of attending a public university. Real tuition at public institutions was stable during the 1940s and 1950s and only rose modestly in the 1960s—averaging about $1,100 (in constant 1982–1984 dollars) in 1970, or about 4 percent of median family income in that year.39 From 1970 to the early 2000s, tuition at public institutions tripled in constant dollars and rose to almost 10 percent of median family income.40 The major reason for rising tuition at public colleges and universities was the reduction of support from state governments.41 In addition to the slowdown of investment in public higher education and rising tuition levels, other potential reasons are mentioned in the research literature for the stagnation in college graduation rates since the 1970s. In addition to traditional factors that influenced the relative levels of college enrollment of men and women, public policy related to military benefits and the draft may have widened gender differences in higher education. For example, veterans who served during World War II and the Korean War received significant financial support to attend college. There was also a deferral of military service for college students before the creation of an all-volunteer force in 1973. These policies probably increased men’s (relative to women’s) college attendance during the 1950s and 1960s.

The Role of Education   11

In their final chapter, which addresses the decline of American education in the late twentieth century, Goldin and Katz speculate that the virtues of American education—its decentralization and open, forgiving character—may have run their course and become impediments to improving educational quality and the preparation of students for postsecondary schooling.42 Decentralization and competition between states and localities worked well to finance universal primary and secondary schooling and to build public colleges and universities after World War II. However, decentralization made it difficult to formulate national policies to address systemic needs with federal funding. The openness and forgiving nature of American education allowed many points of access, including second chances to students to recover from early failure. The open-admissions policies of public universities may work reasonably well if most secondary schools prepare students for college-level work, but the emerging problem may be an increasing number of high school graduates who are very poorly prepared for college. An assessment by William Bowen, Martin Kurzweil, and Eugene Tobin of the stagnation in college enrollment and graduation from the 1970s to the present is similar to that of Goldin and Katz.43 They note that the slowdown in college growth varies by gender and race-ethnicity. White women (but not men) and Asian Americans have continued to increase their college attendance in recent decades. Bowen, Kurzweil, and Tobin conclude that public-sector disinvestment in the quality of high schools has left many disadvantaged American students, disproportionately racial and ethnic minorities, poorly prepared for college. The decentralized structure of public schools, which was once an asset to increasing public funding and access, now penalizes students who live in low-income areas with inferior schools. In their 2009 work, William Bowen, Matthew Chingos, and Michael McPherson also emphasize the widening class divide in access to, and completion of, higher education.44 Based on national longitudinal surveys of high school graduates in 1982 and 1992, Bowen and colleagues find stagnation or regression in college enrollment rates and college completion rates (of college entrants) of students in the lower-income quartile compared to students in the highest income quartile. Students in the middle-income quartiles are more likely to enroll in a four-year college but less likely to earn a degree.45 The persistent and widening gaps in educational success by socioeconomic and minority status are rooted, at least in part, in the organization of higher education. Net of all background variables, students who attend flagship public universities (which are better funded, more prestigious, and have higher student academic profiles) are considerably more likely to earn a college degree than students who begin at other state (and community) colleges. These organizational features of higher education, including the expansion of public universities in the 1950s and 1960s and the more recent rise

12   From High School to College

in costs, have conditioned trends and inequality in access by ascriptive characteristics. But most of the research has focused more on microlevel variables and family background.

Equal Educational Opportunity The commitment to equal educational opportunity in modern societies provides an additional impetus for investment in educational expansion. James Coleman argues that the concepts of equal opportunity, and more specifically equal educational opportunity, only arise in modern democratic societies where individuals can imagine themselves as equal to others, that is, as “exchanging places with anyone else in the system.”46 In traditional societies, there was little social mobility, and the function of schools was to impart the ideals of tradition and authority. In contrast, the expectation of social mobility has become part of the social and cultural fabric in modern societies. This democratic ethos is embodied in the popular belief that a child from a humble background can be just as likely as those from more privileged backgrounds to advance through the educational hierarchy and qualify for a higher status occupation. The underlying assumptions of the meritocratic model are that all children have equal access to schooling, that educational competition is based on ability and effort, and that educational credentials serve as a primary sorting mechanism in the competition for good jobs and rewarding careers. The meritocratic model is an “ideal type” and not an accurate description of American (or any other) society. But so, too, is the image of the United States as a class-ridden society with education merely serving to reproduce the social hierarchy across generations. The reality seems to lie somewhere in between these abstract models. There is a fairly high degree of intergenerational “status maintenance,” with correlations between social origin (the socioeconomic standing of the family of origin) and occupational and economic destination of about .3 or .4.47 A recent study of intergenerational economic mobility with direct measures of income (based on tax records) reported higher correlations of .52 for men and .47 for women between parental income and adult income of persons in their prime working years.48 Even with some degree of uncertainty, it is important to see both sides of the coin. Parental SES does have a strong impact on a child’s occupational and economic attainments. But there is also considerable evidence of substantial intergenerational social mobility, primarily because of educational opportunity.49 Education functions as the critical intervening variable between social origins and adult socioeconomic attainment. Families can bequeath positions and wealth, but most intergenerational advantages are conveyed indirectly through education. However, there is also a moderately high direct effect of education on occupation and income that is

The Role of Education   13

independent of social origins.50 This means that many individuals rise and fall in the socioeconomic ladder based on their “earned” educational attainment. As such, the degree of social mobility in the contemporary United States might be characterized as a constrained (or partial) meritocracy.

The Role of Models in Research on Educational Opportunity Empirical research can serve to adjudicate between competing claims of fluid social mobility, based on a meritocracy or on an ossified social class structure, but only if there is a clear theoretical and testable model of stratification processes. Facts do not always speak for themselves because they can be expressed in a variety of ways to provide support for quite different interpretations. Expressed in statistical terms, outcomes (dependent variables) are often correlated with a number of potential causes (independent variables) that are correlated with one another. For example, many successful students who attend and graduate from college are ambitious, and their high expectations are considered a primary determinant of their educational attainment. But there is also a high correlation between socioeconomic origins and educational attainment. These findings provide support for rather different interpretations of the causes of unequal educational attainment. The association between ambition and educational attainment is consistent with the meritocratic model, while the high correlation between parental status and offspring’s educational attainment could be interpreted as rigidity in the opportunity structure. The nub of the problem is that socioeconomic origins are correlated with educational aspirations—successful parents socialize their children to be ambitious. The interpretation of the shared explanatory power of competing independent variables depends on assumptions about causal priority. More data and more complex statistical methods might be useful, but techniques cannot substitute for a model that specifies how and why independent variables are correlated with each other and their presumed impact on the outcome to be explained. Much of the research in the field of social stratification, including the UW-BHS study, is guided by the model of the socioeconomic life cycle formulated by Otis Dudley Duncan and his colleagues in the 1960s.51 According to the socioeconomic life-cycle model, causal priority among correlated background variables is assumed to follow the temporal order of events over the life course. Since children are exposed to parental influences before schooling, and education generally precedes work careers, the socioeconomic life-cycle model posits that educational attainment is endogenous to social (parental) origins, and that their shared variance (of social origins and education) in explaining occupational attainment is

14   From High School to College

due to the causal impact of social origins that is mediated by educational attainment. Thus, part of the impact of education is status maintenance (social immobility), and the remaining impact of education (net of social origins) represents social mobility or the role of education in loosening the ties between origins and destinations. The logic and method of estimating the effects of exogenous variables on outcomes via direct and indirect pathways is illustrated in Duane Alwin and Robert Hauser’s classic 1975 exposition of the decomposition of effects in path analysis.52 Despite the advances in research based on the socioeconomic life cycle, there are many ambiguous relationships in which temporal order or logic does not clearly imply a one-way causal order. For example, many ascriptive influences, such as the characteristics of the family of origin (race and ethnicity, SES, geographic residence, family structure, and others) have no clear causal priority. It is tempting to think that temporal order could be known with more detailed data, but such fine distinctions are rarely available. Moreover, temporal order assumes that events are more significant than processes. For example, the timing of parental divorce or separation may be intertwined as both a cause and consequence of economic problems. If causal order cannot be clearly established, the nature of a research question determines how variables are specified in a model and how the results are interpreted. In this book, we address the impact of three ascriptive variables—gender, race-ethnicity, and immigrant generation—on educational outcomes. Considerable educational inequality exists along all three dimensions, and our objective is to explain these disparities in terms of their joint variation with other background variables. The first set of covariates is labeled “social origins,” which includes a large number of measures of SES and family background. There is no unambiguous causal order between the three ascriptive variables and social origins—they might be considered jointly determined. However, our research question asks to what extent educational disparities by race-ethnicity and immigrant generation could be a function of social origins if the latter is considered causally prior. The logic of this approach is well established in the research literature and will be discussed in more detail. The more complex, and perhaps contentious, issue in our research design is how to address the relationship between ascriptive variables and the assumption of the meritocratic model, which posits that ability and effort are the primary determinants of educational attainments. Our goal of making assumptions explicit can best be clarified with a digression on various approaches in the prior research literature. One of the most frequently debated issues is the role of measured ability (IQ) as a determinant of educational attainment. There is little disagreement over the question of whether “ability” (based on test scores) has a significant impact on academic achievement among individuals.

The Role of Education   15

Virtually every study shows that test scores have a strong association with, and net direct effects on, all educational outcomes from high school to college. However, a number of related issues involve problematic measurements and dubious assumptions. The first issue is the heritability of ability (assumed to be measured by IQ tests), which is often extrapolated to even more contentious issues over the role of inherited IQ in intergenerational stratification. However, the most thorough and careful review of evidence concludes that the role of heredity in intergenerational stratification is relatively modest.53 Individuals do differ in predispositions to learn, and some elements of these predispositions may be “hard wired” and heritable, just as potential athletic prowess may have an innate component. However, many varied predispositions influence student test scores. Just because we use one summary measure of academic achievement—grade point average (GPA) or test scores—does not imply an underlying single dimension.54 Even if we assume that IQ is only one dimension, the correlation between the IQs of parents and those of their children cannot be assumed to be solely due to heritability. Nature (biological potentialities) interacts with experience and can even shape environmental influences. Tall children are more likely to be recruited (formally and informally) to play basketball. Active and inquiring infants can stimulate parents and caregivers to provide more nurturing learning environments. Since nature and nurture are interdependent in human development, all efforts to quantitatively apportion outcomes to one or the other are based on erroneous assumptions.55 Several major flaws can also be found in the claim that the heritability of intelligence has a major role in intergenerational stratification. Not only is there a social feedback loop from nature to nurture, but advantaged parents do everything in their power to boost the potential learned skills of their children. This issue will be discussed at greater length later in this chapter, but the key point is that all measured tests of ability are strongly influenced by the motivation and resources of families. The recurrent claim that the hereditability of intelligence is the major reason for the persistence of racial and socioeconomic inequality has been largely discredited in classic studies by Christopher Jencks et al., Claude Fischer et al., and Stephen Jay Gould.56 A related flaw in the hereditarianism argument is the leap from the assumed intergenerational association between test scores at the individual level to a between-group interpretation. This leap is a logical fallacy. This may seem counterintuitive to laypersons unfamiliar with statistics, but an example might illustrate the issue. Most children of exceptionally gifted parents rarely inherit the same talents (athletic, intellectual, or artistic). At the same time, exceptionally talented individuals often arise from families of very modest circumstances. This pattern, “regression to the mean” (the random redistribution of many biological traits across

16   From High School to College

generations), loosens the individual-level correlations for many traits that are partially heritable. Since the correlations between social characteristics and biological traits are much looser to begin with, and they are mostly shaped by social patterns rather than heredity, it is virtually impossible to claim theoretical or empirical support for claims that heritable differences in abilities exist between groups defined by social class, race, religion, or national origin. Our focus in this work is on the interpretation of measures of academic ability and cultural factors as potential explanations (our preferred phrase is mediation variables) of between-group (defined by gender, raceethnicity, and immigrant generation) disparities in educational outcomes. In terms of statistical models, we make explicit our assumptions about the shared variance between background variables to explain educational outcomes. Our approach draws on prior research, including several generations of research from the “Wisconsin school” of educational stratification. One of the most famous articles in the Wisconsin tradition is the classic study of the relative impact of SES and measured intelligence on college plans, attendance, and graduation by sociologists William Sewell and Vimal Shah.57 Their research was based on a longitudinal study of Wisconsin high school seniors in 1957, matched with school administrative records, including the Henmon-Nelson Test of Mental Maturity that was a widely used “intelligence” test at the time.58 In order to have an impartial estimate of the effect of SES, independent of intelligence, Sewell and Shah analyzed the bivariate and net effects of SES on college graduation. Male high school seniors from the highest SES quartile were three to four times more likely to graduate from college than were students in the lowest SES quartile, holding constant measured intelligence. The SES advantages were even greater for female students. The findings were replicated in a multivariate path regression analysis. Although Sewell and Shah are agnostic about the causal order of SES and measured intelligence (neither is assumed to be causally prior to the other), their analytical approach, in our judgment, underestimates the true impact of SES. Although a predisposition to learn (rooted in biological differences among individuals) undoubtedly has a strong impact on measured intelligence, we claim that there is no reason to assume that these dispositions differ at birth by parental SES. At a minimum, we claim there is no scientific evidence for biological differences in the ability to learn among newborns by SES. However, we have every reason to believe that differences in measured ability by SES emerge during infancy and childhood because of differential training and coaching of children in families of different SES. Parents are strongly motivated to pass along socioeconomic advantages to their offspring. Some families can bequeath property and wealth to the children and also offer employment and rapid promotion in family

The Role of Education   17

firms. However, most bequests are fairly modest, and few families own or control firms. Advantaged families can also promote the prospects of their children though marriage and contacts with well-to-do and influential persons in their social circle. This, however, is likely less common than it once was. Instead, the primary means of families in modern societies to advance the SES of their children is by providing them with opportunities for educational advancement. Education, especially in the form of a college degree, is the ticket to a better occupation and higher income. Colleges and universities can also serve as marriage markets and venues for making valued social contacts. Given the central role of education in the American stratification system, many families, especially those who have benefited from education themselves, are highly motivated to do whatever it takes to prepare their children to succeed in school. Even before formal schooling begins, many parents and other relatives talk and read to children, seek out stimulating learning experiences, and buy educational toys to prepare children for preschool and kindergarten. Once children begin school, differences in resources and SES among families can lead to wide variations in resources and environments, such as private schools, additional tutoring, and the availability of books and computers at home. Children from disadvantaged families face deficits in each of these areas. As long as children are reared by families, there will not be full equality of educational opportunity. This claim is not based on the assumption of hereditability of ability. Rather, it is grounded in the universal desire of parents to help their children and the differential resources among families. Equality of educational opportunity is one of the most perplexing phrases in social-science terminology. For many, any inequality in educational outcomes (test scores, enrollment rates, attainment) is prima facie evidence of unequal opportunity. This may be a reasonable conclusion in highly stratified caste societies but not in modern industrial societies with universal public schools that are infused with the cultural ideal of meritocratic competition. Some degree of inequality of educational outcomes (the correlation between parental status and a child’s education) may be inevitable in a democratic society composed of families with different resources and values. However, this does not imply that all of the actual inequality in educational outcomes is inevitable, socially desirable, and immutable to social change. The technical question is whether ability (IQ) and cultural characteristics should be considered endogenous or exogenous in measuring the relationship between SES (and other ascriptive characteristics) and educational outcomes. The issue is often framed as differences between observed relationships and “true” effects. Controlling for as many variables as possible is assumed to allow for a better estimate of the true—more conservatively estimated—effect. This is correct if the additional variables

18   From High School to College

(that are held constant) are truly exogenous and represent confounding influences. Our argument is that measured ability and cultural characteristics should be considered endogenous between social origins and progress through schooling. The underlying assumption is that child-rearing and socialization are key mediators that can be measured through the attitudes, traits, and other attributes of adolescents as they pass through educational institutions. These alternative approaches and assumptions can be illustrated with a reconsideration of the study by Sewell and Shah.59 After controlling for IQ, they conclude that about half of the observed relationship between parental SES and college graduation is really due to inequality (unequal chances) and not due to richer people having “smarter” children. Our working assumption is that ability (innate predisposition to learn) varies across individuals, but the distribution of abilities at birth is roughly similar across all socioeconomic strata. The reason for the observed association between parental SES and measured IQ is that parents with higher SES coach their children to do well in school and on tests. Accordingly, our “revised” interpretation of Sewell and Shah’s results is that half of the effect of parental SES on college graduation is mediated by family influences, child-rearing, and other experiences that allow youth of higher SES origins to compete more effectively in schools. The other half of the effect of parental SES on college graduation is direct, which means that it must be explained by other advantages that are not measured by variables in the study. This approach is consistent with subsequent (and more comprehensive) models of educational stratification from the Wisconsin school that posited both test scores and GPA as endogenous to parental SES and other measures of ascription.60 Our objective is to understand the reasons for disparities in educational outcomes. We begin by describing the observed differences in outcomes by ascribed characteristics. These are “social facts” that can be measured with simple bivariate statistics. We then seek to account for these differences in terms of other characteristics, such as family background, childrearing practices, social influences, and student attributes. The objective is not to “explain away” disparities but rather to identify the factors that “explain why” the observed disparities exist. Our tools are multivariate statistical analyses informed by explicit models that posit how and why groups differ in their educational outcomes because of socioeconomic background, socialization, and other early life-course experiences. In other words, our objective is to describe and explain educational disparities by three ascriptive dimensions: gender, race-ethnicity, and immigrant generation. Noticeably absent from this list is “social origins” (SES and other attributes of family background). Family SES is, of course, an ascriptive characteristic, but we argue that socioeconomic origin is the underlying factor that potentially explains other forms of inequality. To be

The Role of Education   19

specific, “testing the SES hypothesis”—that is, determining whether family SES can account for other forms of ascriptive inequality—will be the first order of business in each of our multivariate analyses of educational attainment. As discussed earlier, we can make an obvious and compelling argument that SES contributes to educational attainment. Families with more resources are better able to prepare their children for schooling and to subsidize the costs of higher-quality schooling than poorer families.

Hypotheses of Gender Inequality in Education Historically, the differences in educational attainment between men and women were considered unremarkable—they were relatively modest in magnitude and thought to be explicable in terms of gender differences in career orientations. In general, women received higher grades in high school and were more likely to graduate from high school than men, but were less likely to enroll in and graduate from college.61 Interestingly, the male-female gap in college attendance and graduation widened in the middle decades of the twentieth century as opportunities for higher education grew. The lack of females in higher education was typically explained (“rationalized”) in a variety of ways—differential gender socialization, lower returns to college education, highly feminized occupations not requiring a college degree, and differential family formation patterns (younger age at marriage and child-rearing obligations for women). These interpretations are not unrelated to discrimination against women. Many male employers and male-dominated institutions may have preferred to hire and promote men rather than women; cultural beliefs often stressed marriage prospects over careers for women; and many women may have set educational expectations lower than their abilities warranted. From the 1960s to the 1980s, trends in gender relations began to change. Expectations of lifelong marriage began to wane, tolerance of extramarital sexual relations grew, and prevalence of cohabitation began to rise. These societal trends—often collectively called the second demographic revolution—changed the career orientations of young women.62 A couple of generations ago, about half of young women reported their only career goal was to be a wife and mother. Among cohorts coming of age in the 1980s, the goal of being a homemaker had all but disappeared, and women’s career aspirations were becoming more like those of men.63 By the turn of the twenty-first century, roughly 80 percent of twelfthgrade females expected to graduate from college. The ambitions of men have also changed, but the gender gap has been turned upside down. In the 1970s women were less likely to expect to graduate from college than men, but after 1990 women were ten percentage points more likely to have this expectation.64

20   From High School to College

Since parents do not (generally) choose the sex of their children, the distribution of social (family) origins of men and women are the same. The implication is that the standard models of social stratification, including family background covariates in multivariate statistical equations, has little relevance for explaining gender inequality in educational attainment. If differences exist in parental support for daughters and sons by social origins, however, compositional changes may be relevant for explaining trends in gender inequality in educational attainment.

Hypotheses of Race-Ethnicity and Immigrant-Generation Inequality in Education In contrast to the narrowing and reversal of the gender gap in higher education, disparities in college attendance and graduation by race and ethnicity have been resistant to change. African American, American Indian, and Hispanic youth are much less likely to enroll in and graduate from college than are white youth.65 However, not all racial and ethnic minorities are educationally disadvantaged. Asian American students are more likely to attend college than any other group, and many new immigrants (and the children of immigrants) have above-average levels of educational enrollment and achievement.66 In this book, we seek to understand the sources of racial and ethnic disparities in higher education. Two standard sociological hypotheses have been proposed to “explain” racial and ethnic differences in educational attainment in American society. The first is that racial and ethnic differences are simply a function of SES or social class. Under this interpretation—the social-class hypothesis— race and ethnicity are correlated with educational outcomes, but the “real” cause is socioeconomic differences between racial and ethnic minorities and the majority white population. Minority youth are disadvantaged because they are more likely to be reared in families with fewer resources (such as books, computers, and educational toys) and learning opportunities (travel, educational programs) to prepare them for formal schooling. Other dimensions associated with poorer home environments are unstable family structure and less-educated parents. Children reared in stable, twoparent families and with highly educated parents are much more likely to do well in school and pursue higher education. Low income and family instability often translate into a social and cultural environment that deemphasizes long-term investment and deferred gratification. The social-class hypothesis also links poorer families with poorer neighborhoods, lowerquality schools, and negative peer-group influences that discourage schooling and upward mobility. Testing the social-class hypothesis was one of the primary objectives of the UW-BHS study (see chapter 5).

The Role of Education   21

The second interpretation—the cultural hypothesis—posits that ambition and motivation can be transmitted through intergenerational socialization, irrespective of social class. While the social-class hypothesis considers socialization and cultural traits as important mediating variables that are endogenous to economic and environmental conditions, the cultural hypothesis suggests that some groups are able to sponsor the educational attainment of their children despite having fewer socio­economic resources. The cultural hypothesis is particularly important in understanding the educational attainment of the children of immigrants—popularly known as the second generation. During the first half of the twentieth century, second-generation immigrants in northern cities made remarkable progress in leap-frogging over the educational levels of African Americans and then approaching and often surpassing those of the long-resident (native born of native parentage, or third generation and higher) whites.67 The educational gains of the second-generation immigrants—primarily from southern and eastern Europe—were due in part to the selectivity and economic progress of their immigrant parents, but also to the positive educational climate in northeastern and midwestern cities. A more recent version of the cultural hypothesis is the immigrantoptimism hypothesis.68 Immigrant optimism refers to the “selective” traits of determination and persistence among immigrants and their belief that hard work will eventually pay off in American society. Immigrant parents often work long hours in small businesses or in menial jobs with relatively low pay and opportunities for advancement. These conditions are accepted by immigrant parents as a necessary sacrifice in order to support their children’s education and chances for success. A considerable body of research supports this interpretation, with studies showing that the children of immigrants do better educationally than would be expected on the basis of their socioeconomic origins.69 This boost— the immigrant-optimism effect—is hypothesized to be a response to parental and community pressures to succeed. The children of immigrants are often reminded, in verbal admonitions as well as by observing the sacrifices of their parents, that they must do well in school to climb the economic ladder. Another version of the cultural hypothesis posits that the overrepresentation of some immigrant communities in small businesses promotes the social mobility of the second generation. This interpretation, which originated as the middleman-minority hypothesis, posits that some immigrant groups avoid employer discrimination by opening small businesses.70 Ethnic enterprises also often serve the special needs of recent immigrants by providing or assisting in the search for housing, food, and services. But ethnic enterprises often expand to the general economy by offering popular goods and services at very competitive prices, as with

22   From High School to College

ethnic restaurants, Hispanic lawn-care services, Korean greengrocers, and Asian nail salons. In addition to the economic resources made available by running a family business, entrepreneurial experiences may reinforce cultural traits of rationality, frugality, and long-term investment that are rewarded in schooling. The entrepreneurial families who run or work in ethnic enterprises are hypothesized to be highly motivated to sponsor the educational attainment of their children.71 The effects of ethnic businesses may also spill over to youths in the broader community through employment opportunities, philanthropy, and providing a template for economic success.72 Quite a different cultural interpretation is sometimes invoked to explain the lower educational outcomes of disadvantaged minorities. Although SES and family-background variables explain a share of the educational deficit among African Americans, American Indians, and Latinos, some of the residual gap is often attributed to cultural factors passed along through socialization and child-rearing patterns. The most prominent interpretation of the lower educational attainments of African Americans is the oppositional-culture theory of John Ogbu and colleagues.73 The theory of oppositional culture holds that expectations of discrimination and the lower returns on investment in education experienced by blacks (and other disadvantaged minorities) have led to a rejection of schooling as a means of upward mobility. The most vivid expression of this theory is the disparagement of studying and doing homework by minorities because these activities are seen as “acting white.” Although oppositional-culture theory is supported by some ethnographic accounts, almost every empirical study, quantitative and qualitative, has refuted its major predictions.74 Black students are generally more optimistic about their future and express more proschooling attitudes than white students. There remains, however, considerable discussion about the gap between schooling attitudes and behaviors among disadvantaged minorities. Roslyn Mickelson offers the interpretation that many African American students hold abstract positive attitudes about the importance of education but are less likely to follow up with concrete attitudes and behaviors, including doing assignments and preparing for exams.75 In an insightful survey of the literature on oppositional culture, Douglas Downey claims that the positive attitudes of black students are predictive of success at the individual level, but that environmental influences of poor schools and neighborhoods overwhelm their positive attitudes.76 The leading sociological theory of racial and ethnic inequality in educational attainment is segmented assimilation, which blends both the social-class and cultural hypotheses with a particular focus on secondgeneration immigrants. Originally sketched in a brief article by Alejandro Portes and Min Zhou,77 segmented assimilation theory has been revised and expanded by Portes and Rubén Rumbaut in successive editions

The Role of Education   23

of Immigrant America and in Legacies: The Story of the Immigrant Second Generation.78 Segmented assimilation has become the major theoretical orientation in the literature, but it has also been challenged by many researchers. The major premise of segmented assimilation theory is that the standard sociological theory of assimilation—continuous upward mobility across immigration generations—does not account for the varied experiences of diverse racial and ethnic groups in America in the late twentieth and early twenty-first centuries. By calling attention to the diversity (segments) of trends and patterns, segmented assimilation provides a clear contrast to the “one size fits all” assumption of classical assimilation theory. Social-science researchers debate the claim that classical assimilation theory assumed a linear path of upward mobility for all groups. With the exception of Robert Park’s 1950 prediction of eventual assimilation in his proposed race-relations cycle (contact, competition, accommodation, and assimilation), most research following the classic assimilation paradigm did not expect to find a linear “straight line” path of upward mobility for all groups.79 Even Park, sometimes considered the originator of assimilation theory, noted that segregation and unequal power relations (forms of accommodation) prevented African Americans from entering the mainstream of American society. Milton Gordon’s Assimilation in American Life (1964) is considered a classic statement of standard assimilation theory, but his primary contribution was to distinguish the many dimensions of the concept; he rejected the notion of a monolithic process of assimilation. His characterization of American society, circa 1960, pointed to “ethclasses,” which meant that the major racial, ethnic, and religious groups were not structurally assimilated. By structure, Gordon meant primary group affiliations, such as families, neighborhoods, clubs, and cliques.80 Although most often contrasted with assimilation theory, segmented assimilation is also a challenge to the theory that white racism is an unchangeable feature of American society, inevitably blocking opportunities for racial minorities. There are many branches of racial theory, contingent on time, place, and specific minority groups, but a common theme is that prejudice and discrimination persist because the pervasive ideology of white supremacy provides economic, social, and psychological advantages to the white majority.81 In rejecting the “one size fits all” model, segmented assimilation theory predicts alternative outcomes depending on the characteristics of different groups, social and economic institutions, and historical and social context. Moreover, the key dependent variable in most tests of segmented assimilation theory has been educational outcomes, a much narrower focus than the exclusionary economic and political arenas examined by more general racial theory. In their segmented-assimilation model, Portes and Rumbaut present three pathways, or ideal types, of intergenerational adaptation by

24   From High School to College

immigrant groups.82 The first type, consonant acculturation, is similar to classic assimilation. This model argues that complete integration occurs within three generations in immigrant communities that possess high levels of human capital, have strong family structures, and receive a positive reception from American institutions. Members of these groups enter American society as college graduates and professionals and settle in middle-class suburban areas. They may experience some discrimination, but strong family structure and economic resources allow for relatively smooth patterns of acculturation and upward mobility. The second pathway, dissonant acculturation, eventuates in downward assimilation into menial jobs and possible reactive ethnicity (militant ethnic identification created by discrimination). This outcome is a common finding in many empirical studies, but it is a central element of segmented assimilation theory—predicting that not all groups are equally likely to experience upward mobility in American society.83 Groups that are particularly prone to dissonant acculturation are those with low levels of human capital, little knowledge of English, and weak coethnic communities that have an absence of professionals and entrepreneurs. Dissonant acculturation is also a likely outcome for groups that receive a hostile or negative reception from American institutions. The dissonant acculturation experienced by the second generation leads to failure in American schools, troubled intergenerational relations (little respect for workingclass parents who do not know English), and a propensity to join gangs engaged in deviant behavior. One frequently mentioned finding among youth experiencing dissonant acculturation is a lack of fluency in their mother tongue as well as English. Lack of native-tongue fluency limits family communication and cohesion, and poor English leads to school failure and poor employment prospects. The most interesting pathway in the segmented-assimilation model is the third—selective acculturation. This pathway partially resembles that of dissonant acculturation in its composition—namely, working-class immigrants with low levels of human capital. The primary difference is that selective acculturation is possible with strong families and cohesive ethnic communities that provide protection from the discrimination and the lack of receptivity from American institutions. Strong coethnic institutions, such as ethnic-enclave businesses, religious institutions, and voluntary organizations, provide positive images and opportunities that can compensate for the lack of economic resources. Children in these communities are encouraged to selectively acculturate—that is, to pursue upward mobility through schooling and mainstream institutions but not to adopt American adolescent culture. Those who selectively acculturate are encouraged to adhere to traditional values of family, culture, and language. If successful, selectively acculturated youth become bilingual, and this serves to reinforce family solidarity and educational advancement.

The Role of Education   25

Segmented assimilation theory predicts integration and upward mobility, along with bilingualism, for those who selectively acculturate. Segmented assimilation theory does not present a simple model for empirical research. Cases, which combine several characteristics (values on variables), are the primary units in the theory. The theory is easier to interpret with representative national-origin groups that easily fit the theoretical cases of consonant, dissonant, and selective acculturation. This approach is illustrated in the second chapter of Portes and Rumbaut’s Immigrant America, which demonstrates how national-origin groups are often closely identified by human capital and legal status.84 Professional and entrepreneurial communities generally have high human capital and enter the United States as legal immigrants. They encounter a neutral or positive reception, as do refugee groups that receive support from the government and community groups. Examples of such groups are Cubans, Vietnamese, Russians, Iranians, Chinese, and some South American and South Asian professionals. These ethnic communities are considered to be cases of consonant or selective acculturation. At the other end are workingclass communities with an unauthorized (undocumented) legal status in the United States, such as Mexicans, Haitians, and some Central American groups—for example Salvadorans and Guatemalans. The children from these groups are considered the most at risk of dissonant acculturation. This short summary does not do justice to the nuances and qualifications of Portes and Rumbaut’s text, which considers multiple legal statuses and the timing of arrival within national-origin groups. The conventional approach to testing the predictions of segmented assimilation theory is to analyze the educational success and other early life-course outcomes of the second-generation national-origin groups.85 National-origin groups are identified as consistent with the predictions of consonant, dissonant, and selective acculturation based on criteria such as parental human capital, occupational profiles, family structure, and bilingualism. Empirical analyses are invariably messier and more complicated than pristine conceptual models, but the results are generally in line with the theory. Cubans and East Asians do relatively well in school, in part because of their measured background variables (high parental human capital, for example) and also, some researchers argue, because of unmeasured characteristics, such as a favorable reception in the host society and strong family solidarity. Mexican American and other immigrant communities do not perform as well because of their low levels of family SES, weak coethnic communities, and a negative, often hostile reception (discrimination). A central theme of the criticism of segmented assimilation is the theory’s inability to sufficiently account for temporal changes in the progress of immigrants and the second generation. Joel Perlmann and Roger Waldinger note that the children of immigrants from southern and eastern

26   From High School to College

Europe in the early twentieth century eventually did climb the educational, economic, and social ladders in American society, but that it was a very long and arduous process.86 Segmented assimilation theory posits that the darker skins of many contemporary (post-1965) immigrants and the new economy with few stable, well-paying, working-class careers have raised the risks for dissonant acculturation. But Perlmann and Waldinger observe that many early twentieth-century immigrants were not considered white and that they encountered poorly paid, unstable employment in nonunionized and very dangerous workplaces. The second generation of Italian, Greek, and eastern European immigrants initially experienced high dropout rates and low rates of social mobility. Stanley Lieberson finds that there was gradual progress in residential integration, occupational patterns, and educational attainment of the second-generation immigrants in northern cities from 1910 to 1940.87 The incremental changes over time and across generations are not emphasized in segmented assimilation theory. Richard Alba and Victor Nee conclude that the classical sociologicalassimilation perspective best describes and explains the long-term trajectory of race and ethnic groups in American society, and of the descendants of immigrants in particular.88 They argue that segmented assimilation theory is about variations in the process—temporal and national-origin differences, which are evident at any moment but not fundamental to the long-term historical pattern. Alba and Nee are most convincing in their revisionist account of the progress of the descendants of southern and eastern European immigrants during the first half the twentieth century. Despite the claims by many leading scholars in the 1960s and 1970s that the “melting pot never happened,”89 Alba and Nee conclude that the majority of all white ethnics moved up the socioeconomic ladder, achieved residential integration in suburban areas, and even intermarried across religious lines at surprisingly high rates.90 In his more recent volume, Blurring the Color Line, Alba presents a broader interpretation of how New Deal policies, the unifying experience of World War II, the expansion of higher education in the 1950s and 1960s, and suburbanization allowed American society to become more inclusive of white ethnics, primarily the children and grandchildren of Catholic and Jewish immigrants.91 The one exception to the master trend of assimilation is the historical experience of African Americans.92 If the integration and socio­ economic mobility of white immigrants can be measured in generations, the progress of black Americans, whose presence in the United States predates the arrival of almost all the ancestors of white Americans, is only evident in glacial terms. Indeed, the upward mobility of white ethnics occurred when there was a hardening of racial lines and widening of residential segregation from 1910 to 1960.93

The Role of Education   27

There is considerable agreement, however, between the proponents and critics of segmented assimilation theory on many of the “facts” regarding post-1965 immigrants and their children, as highlighted in a 2011 issue of Social Forces. An overview of segmented assimilation theory, including a response to its critics and new empirical analyses, is presented by William Haller, Alejandro Portes, and Scott Lynch, followed by comments from Richard Alba, Philip Kasinitz, and Mary Waters and then a reply by the original authors.94 This exchange is stimulated, at least in part, by the book Inheriting the City: The Children of Immigrants Come of Age by Kasinitz et al., which is considered to be an empirical, if not a theoretical, challenge to segmented assimilation theory.95 Kasinitz and his coauthors find, to their surprise, that second-generation immigrants in New York City had higher educational attainment and better career prospects than domestic minority youth—African Americans and Puerto Ricans.96 The exchange in Social Forces highlights broad areas of agreement on many empirical issues, including: • Almost all second-generation immigrants are fluent in English and have acculturated to American society. • The children of immigrants with favorable social origins (high levels of human capital) have above-average educational and occupational attainments that are comparable to those of native-born white Americans. • Family structure, including coresidence in early adulthood with parents, has a significant influence on socioeconomic mobility of the second generation. • Significant minorities of second-generation youth lag behind their peers. In other words, proponents and critics of segmented assimilation theory agree on the positive outcomes of the second-generation groups identified by consonant and selective acculturation in the theory. Critics may question the need for a new theoretical perspective, but there is little disagreement on these empirical findings. The major empirical disagreement is focused on the scope and patterns of the dissonant acculturation group with low human capital, weaker family and community cohesion, and a negative reception by American institutions. In several studies based on the data from their Children of Immigrants Longitudinal Study (CILS), Portes and colleagues found that Central American, Haitian, West Indian, and especially Mexican second-generation youth are much more likely to experience downward assimilation in terms of high school dropout rates, poverty, unemployment, early fertility, and

28   From High School to College

encounters with the criminal justice system.97 Controlling for family background, researchers find that some of these disadvantages are due to selectivity (human capital and family structure), but the remaining negative direct effects indicate the cause as negative reception of these groups, including but not limited to discrimination. Some of the differences in findings and interpretations between proponents and critics of segmented assimilation theory may be due to differences in populations (for example, Mexicans are not included in Kasinitz et al.), study design (cross-sectional versus longitudinal), data sources, and dependent variables. But differences are also apparent in empirical results. Kasinitz et al. and Charles Hirschman find that West Indians are not below average in educational outcomes; however, the same authors find the situation of second-generation Dominicans (Kasinitz et al.) and young immigrants from the Hispanic Caribbean (Hirschman) are consistent with segmented assimilation theory.98 Michael White and Jennifer Glick analyze inequality by immigrant generation (adolescent-arrival, 1.5, second, third) over various dimensions, including educational tests and attainment with national longitudinal surveys conducted by the NCES.99 Without controls for family background, first- and second-generation immigrants have lower educational attainments than third-and-higher-generation (native born of native parentage) Americans. The lower attainments of first- and second-generation immigrants are largely due to “social class”—lower socio­economic origins. However, once socioeconomic background is included as a covariate, immigrant generation is much less likely to be a major predictor of educational attainment. In fact, the second and 1.5 generations perform better in math tests and in college completion.100 These findings provide support for the hypothesis of second-generation advantage rather than for segmented assimilation theory. However, as noted, we see less of a contest between these perspectives than do some of the protagonists. An overall second-generation advantage, however, does not negate the findings that some groups are downwardly mobile, in accordance with segmented assimilation theory. For example, in their multivariate models, White and Glick find a significant negative net effect of Mexican ethnicity on educational outcomes, a finding that could be interpreted as evidence for the dissonant acculturation prediction of segmented assimilation theory.101

Key Assumptions of Models and Measurement Our goal in this work is to describe and explain educational disparities by gender, race-ethnicity, and immigrant generation. Our analytical approach is to consider these three variables as exogenous, or “primary,”

The Role of Education   29

and then to test whether, and how much, educational inequality in these dimensions is mediated by other student characteristics that are posited to be endogenous. The selection of these three characteristics of family of origin as exogenous may appear to be arbitrary—almost all attributes of family background are “prior” to the life of the student, and their temporal order is unknown and probably unknowable. However, the logic of our analysis depends on specifying precise questions with explicit assumptions. The first assumption is that gender, race-ethnicity, and immigrant generation do not directly influence educational outcomes. Significant educational disparities by these variables may be evident, however, because of their association with other background variables that do directly affect educational outcomes, such as SES and other structural attributes of the family of origin. Gender, race-ethnicity, and immigrant generation may also be associated with education outcomes because of their indirect influence via proximate influences on student behaviors, attitudes, and roles. The key factors that might directly affect educational outcomes are socialization from family and friends, primary group associations, student employment and participation in school activities, and the qualities of schools attended. In this introductory chapter, we present an overview of the logic of this approach, with the full explanation and details presented in the subsequent chapters. As already noted, inequality in family SES is the major reason for the lack of educational opportunity in the United States. In the abstract, all families want the best for their children, but they differ in their abilities and resources. Children from advantaged families are more likely to be coached and trained to do well in school, have access to better neighborhoods and schools, and have social networks that reinforce educational attainment. The higher test scores, school grades, and educational achievements of children from advantaged families are not evidence of superior innate abilities but of the greater social and economic investments made by their families. All theories of educational inequality, regardless of their peculiar emphases, stress the centrality of the social class, or SES, of families of origin. Testing the social-class hypothesis in educational outcomes by raceethnicity and immigrant generation is the subject of chapter 5. Since the sex of children is randomly distributed among families, SES has a different meaning for explaining gender differences in educational outcomes. Although the logic of the social-class hypothesis is fairly straightforward, the identification and measurement of socioeconomic characteristics are not. The term social class implies a conceptual framework with a few discrete categories that are hierarchically ranked with clear and substantial differences between groups but internally homogenous. In contrast, the term socioeconomic status suggests a graduated, but continuous, hierarchical

30   From High School to College

scale. Our use of these two terms interchangeably in this work captures the inequality of educational attainment that is caused by the differential statuses and resources among families of origin. Our preliminary efforts to create a parsimonious SES index were stymied by a variety of technical issues, including missing data across a number of dimensions. Our preliminary research revealed that every summary index was inferior to the inclusion of the combined explanatory power (and mediating role) of all measured dimensions of SES and family structure. Since our objective was not to resolve conflicting theoretical claims about different SES concepts and measures, we adopted the pragmatic strategy to include all family-background variables that had a strong impact on educational outcomes and on educational inequality by raceethnicity and immigrant generation. They are: father’s educational attainment, mother’s educational attainment, father’s employment, mother’s employment, father’s occupational status (for respondents with a father or father figure in the labor force), mother’s occupational status (for respondents with a mother or mother figure in the labor force), home ownership, and family status (intact or not). In addition to SES and family background, two other intervening variables are considered in chapter 5: encouragement and GPA. Prior research has identified the influence of significant others (“significant-other influences,” or SOI) and academic performance as the primary mechanisms though which social class works.102 As noted earlier, the ascriptive variables of gender, race-ethnicity, and immigrant generation have been hypothesized to influence educational outcomes through cultural factors that are independent of family SES. To address this broad question, in chapter 6 we formulate a tentative model of specific cultural traits and test whether and how educational disparities by gender, race-ethnicity, and immigrant generation are mediated by cultural mechanisms beyond those influenced by SES and family background. The variables of encouragement and student performance (GPA), introduced as mediators of SES in chapter 5, are reconceptualized in chapter 6 within a broader rubric of cultural context, cultural orientations, and cultural expressions. The goal is not to explore all aspects of the relationship between culture and schooling but to test hypotheses that cultural influences and patterns mediate some fraction of the observed relationships between our three ascriptive variables and educational outcomes.

A Brief Recap In the empirical analysis of the UW-BHS study, we survey the sources, mechanisms, and influences on disparities in college graduation by gender, race-ethnicity, and immigrant generation. We seek to understand the reasons for the below-average rates of college graduation of disadvantaged

The Role of Education   31

minorities as well as the above-average educational attainment of other groups, including females and Asian Americans.103 Immigration status is closely intertwined with race and ethnicity in American society. The children of immigrants, generally considered to be at high risk because of low SES, have done surprisingly well in the American educational system.104 The recent evidence of higher rates of college graduation of women relative to men has become a central issue for research on educational stratification.105 Our analysis is guided by ideas and hypotheses put forth earlier in this chapter. All theories of ascriptive inequality posit that the role of class or family SES is an important determinant of schooling, especially of higher education. The potential of family socioeconomic origins as a major variable in racial-ethnic and immigrant-generation inequality in college graduation is a key hypothesis. We also examine alternative explanations, which state that family and community socialization may serve as mediating variables of SES and may also have direct independent influences. For example, Grace Kao and Marta Tienda posit in their immigrant-optimism hypothesis that immigrants pass along their high expectations and work ethic to their children.106 James Coleman claims that some immigrant groups create high levels of social capital because neighbors and relatives monitor the activities of youths in order to reinforce the values of parents.107 Portes and Rumbaut offer segmented assimilation as a theoretical framework that incorporates a broad array of factors (family SES, community resources, reception by the host community, and family communication) to explain why some second- and 1.5-generation immigrant groups do better than others.108 We posit that social background affects college background and influences educational outcomes through five stages of the educational ladder, beginning with aspiring to graduate from college and culminating in completing college within seven years after high school graduation. First we examine what contributes to one’s aspiration to graduate from college. Aspirations to attend and graduate from college are often considered the major determinant of educational attainment. We then ask what factors allow vague aspirations to be solidified into concrete expectations to graduate from college. Some research has suggested that underachieving groups have high hopes for college but lack realistic expectations that help them follow through on their aspirations. Next, we examine how students’ college expectations affect their preparations for college while in high school. College preparation is measured by taking college preparatory courses, taking college entrance exams (SAT or ACT), and applying for college. The next step on the ladder is college enrollment, particularly in a four-year college, and we analyze why many ambitious and college-prepared students do not make the leap to college enrollment. Finally, we look at the factors affecting the completion of

32   From High School to College

college by college entrants, given that only half of those who begin college eventually earn a BA or BS degree. The overwhelming majority of college graduates follow this sequence, so understanding which transitions are more consequential and which may vary across groups is a useful analytical framework. Of course, some students earn a college degree through an alternative pathway (second chances after deviating from the linear pathway), so we measure this process as well. Our analysis is largely based on data from the UW-BHS project, which surveyed almost ten thousand high school seniors in nine public high schools and three private high schools in the Pacific Northwest from 2000 to 2005. In addition to the baseline senior survey (a “paper and pencil” questionnaire) we also conducted a one-year follow-up telephone survey, with a 90-percent response rate, and matched respondents with college-enrollment and graduation records from the National Student Clearinghouse. The longitudinal research design allows us to compare orientations and social characteristics measured in high school with records of college graduation up to seven years after high school. In chapter 2, we set the stage with a review of educational trends based on national census and survey data, with a focus on race and ethnic differentials in college graduation rates. We also decompose trends and inequality in college graduation rates by race and ethnicity into components representing high school completion, the transition from high school to college, and college completion. In chapter 3, we review the origins of the UW-BHS project and evaluate its coverage and quality. Although the UW-BHS data are not a representative sample of high school seniors, many of the patterns reported here are similar to those from national studies. We also consider potential threats to the meaning and measurement of all major variables in the analysis. In chapter 4, we present the College Pathways Model, which represents key stages on the road to college graduation. We provide a rationale for the model and present an empirical analysis of student attrition at each stage of the process. We also conduct a demographic decomposition of how failure at each stage contributes to disparities in college graduation by gender, race-ethnicity, and immigration-generation status, including alternative pathways to college completion. The objective of chapter 5 is a detailed analysis of how SES and other aspects of family-background variables affect each stage of the College Pathways Model from college aspirations to college completion. In chapter 6, we take a closer look at the role of cultural factors as an explanatory variable for college graduation rates. In addition to being full-time students, the majority of high school seniors are workers (paid employees) and almost half participate in student extracurricular activities (including sports). These nonstudent roles are analyzed in chapter 7, where we examine the stratification of student employment and activities and how

The Role of Education   33

the participation in nonstudent roles affects college aspirations, expectations, preparation, college enrollment, and graduation. Chapter 8 addresses the effects of school settings on college enrollment and graduation and analyzes the impact of the Washington State Achiever (WSA) program in five of the twelve high schools in the UW-BHS project. The WSA program, sponsored by foundation grants, aimed to expand college enrollment and graduation among low-income students with opportunities for college scholarships and other initiatives. It was an experiment introduced in sixteen low-income high schools in Washington State, several of which overlapped with the UW-BHS sample of high schools. This overlap created an opportunity for a quasi-experimental evaluation of the WSA program. With additional foundation support, the UW-BHS data collection was increased from five to nine public high schools, including five WSA schools and four nonprogram schools, plus three private high schools. Chapter 9 summarizes our major findings and their implications for research and social policy on ascriptive inequality and educational opportunity in American society.

Chapter 2 Recent Trends in College Graduation: The National Portrait with nikolas pharris-ciurej

I

2012, the Census Bureau issued a press release announcing that, for the first time, more than 30 percent of Americans aged twenty-five and older held at least a bachelor’s degree—up from slightly less than 25 percent in 1998.1 This story was widely reprinted in the news media and framed in a broader narrative about the continuous expansion of higher education in the United States. The story of growth in college enrollments and graduates—interpreted to mean that America is a land of opportunity—is a dominant theme expressed by American political leaders and social commentators. Since state and federal statistical systems are continuously churning out numbers of enrolled students, graduates, and educational expenditures, virtually endless data points show the growth of American higher education, with the implicit assumption that growth equals expanding opportunities. For example, the 20.6 million students enrolled in degree-granting postsecondary institutions in the fall of 2013 is up from 8.6 million in 1970, 13.8 million in 1990, and 15.3 million in 2000.2 These raw numbers, however, do not mean that 21 million American youth fit the traditional image of full-time students at four-year colleges and universities. The reality is that more than one-third of college students attend two-year community colleges and the majority of community-college students attend part time. Even among students enrolled in four-year institutions, 30 percent attend part time.3 About one in ten students at four-year colleges are enrolled in for-profit institutions. These alternative enrollment patterns have expanded opportunities to higher education for n early

34

Trends in College Graduation   35

many persons who do not have the resources, qualifications, and time for full-time attendance at traditional four-year colleges, but these alternative pathways are much less likely to lead to a bachelor’s degree than full-time enrollment in a four-year college. Here, our focus is on measuring the output of American higher education by looking at the trend in college graduation rates of successive generations (birth cohorts). The trends in high school and college graduation rates are shown in figure 2.1 for persons aged twenty-five to twentynine years from 1940 to 2013 based on the Current Population Survey (CPS)—a large Census Bureau survey of all American households that is widely used to measure social trends in American society.4 The standard census question on educational attainment is, “What is the highest grade of school completed or the highest degree received?” But this question does not capture the eventual educational attainment of many adolescents and young adults because they are still in school. At the other end of the age spectrum, the educational attainments of older persons reflect the experiences of those who were students many decades earlier. The selection of persons aged twenty-five to twenty-nine is intended to capture the “sweet spot” of the age distribution—old enough so that most will have completed their schooling, but not so old that their educational

100%

High school graduate

90 80 70 60 50 40

College graduate

30 20 10

19

19

45 19 50 19 55 19 60 19 65 19 70 19 75 19 80 19 85 19 90 19 95 20 00 20 05 20 10

0

40

Percent of U.S. population graduated

Figure 2.1    Percent of High School Graduates and College Graduates in the U.S. Population, Aged Twenty-Five to Twenty-Nine, Based on the CPS Series from 1940 to 2013

Source: Author’s compilation based on data from U.S. Census Bureau 2014. Note: CPS = Current Population Survey.

36   From High School to College

experiences are far in the past. A simple summary measure of all adults, such as the education of persons aged twenty-five years or more (as was used in the 2012 Census Bureau press release),5 is a weighted average of the educational qualifications of persons who last attended school in the distant past, recent school leavers, and everyone in between. The average educational attainment of all adults is weighted by the age distribution of the population, which means that larger age groups—such as the baby-boom generation—may distort a summary index based on persons above a certain age. To avoid these confounding factors, the Census Bureau (and most researchers) generally publishes temporal comparisons based on the educational attainment of persons aged twenty-five to twenty-nine (which is more closely linked to recent trends in school completion of young adults). The top line in figure 2.1 shows the percentage of young adults (aged twenty-five to twenty-nine) from 1940 to 2013 who have earned a high school diploma, either through graduation from high school or by successfully completing a high school equivalency program, such as the GED program. The bottom line shows the proportion graduating from college with a baccalaureate degree. Even though the education question was changed in 1993 (from years of schooling completed to degrees earned), this time series is a reliable assessment of the long-term temporal trend in educational completion.6 The time period of 1940 to 2013 along the horizontal axis represents the year of data collection. The members of the population at each data point were born twenty-five to twenty-nine years earlier. The schooling of each cohort was probably completed five to ten years before the years listed. Later in this chapter, we index successive cohorts by the approximate year(s) when they were aged twenty—as a proxy for the temporal conditions at the time when each cohort was in the college-going years. During the first decade of the twenty-first century, the high school graduation rate reached 90 percent and the college graduation rate exceeded 30 percent. The historical trends cannot be described as ones of continuous expansion of educational attainment. There are two distinct eras— a period of rapid growth from the 1940s to the 1970s and one of much slower growth from the 1980s to the present. High school graduation rates rose from about 40 percent in the 1940s—reflecting the graduation rate in the 1930s—to over 80 percent in the 1970s—reflecting the graduation rate in the 1960s. After a long lull, the rate was approaching 90 percent by the early 2000s. Given the “ceiling”—the high school graduation rate cannot exceed 100 percent—a slowing rate of increase in rates is to be expected. College graduation was a rare event in American society until the middle of the twentieth century. Fewer than one in ten young adults completed college in the 1940s and early 1950s. From the late 1940s to the early 1970s, public higher education grew by leaps and bounds and the college graduation rate rose to about 25 percent. Then stagnation

Trends in College Graduation   37

set in—perhaps even a small downward movement of a few percentage points. After about twenty years of very little progress, there has been a modest uptick since the late 1990s and early 2000s, when the graduation rate increased slightly to just over 30 percent. The trends in figure 2.1 are largely the same as those reported by other education researchers that have analyzed CPS and comparable census data.7 The stagnation in college graduation rates during the last quarter of the twentieth century poses a fundamental challenge to the standard inter­ pretation of continuous expansion of opportunity and higher education in American society. It serves as an important backdrop and motivation for this study and is also central to the broader literature on education and opportunity in American society.8 To set the stage for our analysis, this chapter presents a detailed description of national trends in college graduation over the second half of the twentieth century by major racial and ethnic groups. Then we decompose the overall college graduation rate into three key educational transitions: high school completion, high school graduation to college enrollment, and college enrollment to college completion. This analysis allows a clearer picture of how the slowdown in the number of students transitioning to college and in the collegecompletion rates affected different groups in American society. Our analysis follows in the well-established tradition of estimating the historical trend in educational attainment from cross-sectional census or survey data with two key assumptions: successive age groups accurately represent successive birth cohorts, and education is generally completed by age twenty-five. For example, in her 1968 article Beverly Duncan examined intercohort trends in educational attainment based on the 1960 census tabulations by age group.9 Using similar data and logic in his 1980 work, Stanley Lieberson measured trends in the schooling of secondgeneration white immigrants and African Americans in northern cities.10 Robert Mare presented a definitive overview of trends in educational attainment and inequality by gender, race-ethnicity, and nativity based on a cross-sectional analysis of 1990 population-census data.11 However, there is a nontrivial risk of bias in estimating trends from cross-sectional data, especially at the oldest and youngest ages. The stock of older persons at any given time is depleted by mortality. Such mortality is invariably selective, as less-educated people tend to live shorter lives, which in turn may cause an upward bias of the reported level of education of older cohorts. At the other end of the age distribution, the educational attainments of younger persons may be biased downward as many individuals do not complete college until their late twenties and early thirties. These potential biases can be estimated using data on completed education for the same cohorts from multiple time points. In table 2.1, we compare reported college graduation rates for the same cohorts at different ages from the 1990 and 2000 population censuses and the 2010 ACS. In

1940–1944 1945–1949 1950–1954 1955–1959 1960–1964 1965–1969 1970–1974 1975–1979 1980–1984 1985–1989 1990–1994 1995–1999 2000–2004

1920–1924 1925–1929 1930–1934 1935–1939 1940–1944 1945–1949 1950–1954 1955–1959 1960–1964 1965–1969 1970–1974 1975–1979 1980–1984

65–69 60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29

1990 Census

65–69 60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29

2000 Census

Source: Author’s compilation from Ruggles et al. 2015.

Year Age 20

Birth Cohort

Age Reported in

65–69 60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29

2010 ACS 12.5% 14.9 17.2 19.4 23.6 27.7 26.4 23.3 22.1

1990 Census

18.0% 20.3 24.7 29.1 28.4 25.9 25.8 27.9 27.2

2000 Census

25.7% 30.1 29.5 27.4 27.9 30.7 31.9 32.1 30.5

2010 ACS

Percent College Graduate Reported in

0.8 0.9 1.1 1.4 2.0 2.5 3.8

1990–2000

1.0 1.1 1.0 1.5 2.0 2.7 4.6

2000–2010

Percentage Point Change

Apparent Intracohort

Table 2.1    Percent of College Graduates in the U.S. Population by Birth Cohort Based on Age Groups in 1990, 2000, and 2010: Based on the 1990 and 2000 Population Censuses and the 2010 American Community Survey

Trends in College Graduation   39

addition to measuring the potential bias by age at reporting, the multiple data sources allow for the measurement of the intercohort trends over a much longer period than is possible from only one data source. The first column in the table labels each birth cohort, from the 1920– 1924 cohort through the 1980–1984 cohort. Given that schooling tends to be completed in late adolescence or early adulthood, birth cohorts are identified by the calendar years in which the cohort reached age twenty (column 2). Labeling birth cohorts by the period (calendar years) when they might have been enrolled in college provides a more intuitive index of the temporal conditions that may have shaped college-going. The next three columns show the reported age for each cohort in the census years of 1990, 2000, and 2010. For example, members of the 1940–1944 birth cohort turned twenty years old during the years 1960 to 1964. Assuming selective mortality is small and there is no age bias in reporting education, the percent completing college for this cohort should be the same in the 1990 census (when they were aged forty-five to forty-nine), the 2000 census (when they were aged fifty-five to fifty-nine), and in the 2010 ACS (when they were aged sixty-five to sixty-nine). The cohort that turned twenty in the early 1960s (born 1940 to 1944), reported 23.6 percent completed college in 1990, 24.7 percent in 2000, and 25.7 percent in 2010. These differences are labeled “apparent intracohort change” in the last two columns of table 2.1. Since little education occurs at older ages, the expected value of intracohort change should be close to zero. This is “almost” true: most of the differences are a little over one percentage point, which is well within the range of sampling and measurement error. This is not the case for younger age groups, with estimates of intracohort change of two to four percentage points. This indicates that a significant number of individuals are completing college in their late twenties and early thirties. This finding—an increase in the average age of college graduation—has been confirmed in additional analyses based on cohort analyses of college graduation from the CPS data (not shown in table 2.1). Some of the reported downturn in college graduation rates in the 1970s may have been due to postponement in the timing of college graduation. The recent uptick in college graduation rates in the 1990s and early 2000s is more evident when college graduation is measured for persons in their thirties than in their twenties. These patterns are evident in figure 2.2, which is a graphic representation of the figures from table 2.1. Each line shows the intercohort change (birth cohorts are shown along the horizontal axis) from one data source. Not all cohorts are reported in all three data sources—there is only partial overlap. For example, the earliest cohorts—those in the college-going years in the 1940s—are only available from the 1990 census. And the most recent cohorts representing the late 1990s and early 2000s are only available for persons who were aged twenty-five to twenty-nine and thirty

40   From High School to College Figure 2.2    Percent of College Graduates in the U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty), Based on the 1990 and 2000 Censuses and the 2010 American Community Survey

Percent college graduates

35%

Based on 2010 ACS

30 25

Based on 2000 census Based on 1990 census

20 15 10

4 00

9 99

–2

20

00

4

–1

99

95 19

–1

19

90

98

9

4 19

85

–1

9

98 –1

97 19

80

4 97

–1

19

75

9

–1

96 19

70

4

–1

96 19

65

9

–1

95 19

60

4

–1

95 19

55

9

–1

19

50

94 –1

45 19

19

40

–1

94

4

5

Year that birth cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015.

to thirty-four in the 2010 ACS. The 1990 and 2000 census data provide estimates of college graduation rates for birth cohorts during the era of growth from the 1940s to the 1960s, and the 2010 ACS provides estimates of college graduation rates for the birth cohorts that witnessed the recent increase in college graduation in the 1990s. For the periods when multiple observations of the same cohort are available, the overall patterns are similar. The slight differences, generally in the range of one to two percentage points, suggest that there is little bias in estimating college graduation for most of the period represented here. Perhaps the expansion of higher education in the early post–World War II era might be understated because of selective mortality of the least educated, but we are confident that the overall trend is robust. The potential bias in estimating college graduation rates for recent cohorts is a more serious problem. For example, persons born in the late 1950s and early 1960s reported a college graduation rate of only 22 to 23 percent in the 1990 census, but 27 to 28 percent in the 2010 ACS. Looking backward, it seems that the reported decline of college graduation rates in the 1970s and 1980s based on data from 1990 was shorter and shallower when viewed from hindsight with data from the 2010 ACS.

Trends in College Graduation   41

Looking forward, the reported uptick in college graduation rates in the 1990s and early 2000s may be much more positive when the youngest cohorts have second chances to graduate from college as they age. In this chapter, we splice the three-time series from figure 2.2 (and table 2.1) to create a time span of sixty years for the birth cohorts of 1920–1924 to 1980–1984 (or persons reaching college-going age from the early 1940s to the early 2000s). Specifically, we take the data for the two oldest cohorts (born in the 1920s) from the 1990 census, the next two cohorts (born in the 1930s) from the 2000 census, and all other birth cohorts from the 2010 ACS. As our preceding discussion suggests, this exercise is not without hazards. For the earliest cohorts, there might be a bias in overestimating college graduation because survivors of the oldest cohorts (in the 1990 census) were better educated that those that died. This bias is probably small and would only have the effect of dampening the dramatic rise of college graduation from the 1940s to the 1960s. More serious is the potential bias for the most recent period—those reaching their twenties in the late 1990s and early 2000s. Accordingly, we are cautious in our interpretation of intercohort change for this recent period.

Changes in Race and Ethnic Composition We now turn from a description of the overall trend in college graduation to variations by race and ethnicity—one of the primary themes of our study. First, it is necessary to review changes in the measurement of race and ethnicity in recent censuses. In the next chapter, we report a new classification of race and ethnicity based on very detailed data from the UW-BHS data. In this chapter, we explain how we created a consistent set of mutually exclusive classifications of race and ethnicity from the 1990 and 2000 censuses and the 2010 ACS. The 1990 census was the last to measure race by asking respondents to make a choice among a set of mutually exclusive categories. Beginning with the 2000 census, respondents could select “two or more races.” The multiple-race population is only a small fraction of the total population (less than 3 percent), but it forms a much large share of some groups— American Indians and Pacific Islanders in particular. We relied on the Integrated Public Use Microdata Series (IPUMS-USA) RACESING variable to assign multiple-race respondents into their most likely single-race group.12 Although Hispanics may be of any race (Hispanic origin is measured with a separate census question), almost half of Hispanics do not report a race or check “some other race” and write in a Hispanic identity.13 The majority of Hispanics consider their ethnicity as their primary identity.14 Accordingly, we created a combined race-ethnicity classification with all Hispanics, regardless of their reported race, in two separate categories: Mexican and Other Hispanics.

42   From High School to College

Since our focus is on educational outcomes from the American educational system, we restricted the analytical sample to the native-born population and immigrants who arrived as children and were most likely educated in American schools. Specifically, we include all persons born in the United States (or who are citizens at birth) and immigrants who arrived before age twelve—typically known as the 1.5 generation. Most adult immigrants (indexed as foreign born who arrived after age twelve) received all or part of their education in their countries of origin and may not have been students in the American educational system. The changing composition of the American population is shown in table 2.2 for seven

Table 2.2    Racial, Ethnic, and Nativity Composition of the U.S. Population by Birth Cohorts from 1920–1924 to 1980–1984 Non-Hispanic White (%) Birth Cohort 1920–1924 1925–1929 1930–1934 1935–1939 1940–1944 1945–1949 1950–1954 1955–1959 1960–1964 1965–1969 1970–1974 1975–1979 1980–1984

African American (%)

Total Population

Native Born

Foreign Born

Native Born

Foreign Born

100% 100 100 100 100 100 100 100 100 100 100 100 100

81 79 77 75 75 74 72 69 65 61 57 56 58

5 4 4 4 3 3 3 3 3 3 3 3 2

8 9 9 9 8 9 10 11 11 12 12 12 12

0 0 1 1 1 1 1 1 1 1 1 1 1

Source: Author’s compilation from Ruggles et al. 2015. Notes: 1920–1924 and 1925–1929 cohorts are based on persons aged 65–69 and 60–64 from the 1990 census. 1930–1934 and 1935–1939 cohorts are based on persons aged 65–69 and 60–64 from the 2000 census. 1940–1944 to 1980–1984 cohorts are based on persons aged 25–29 to 60–64 from the 2010 American Community Survey. “Native born” includes the 1.5 generation (immigrants who arrived before age 12). The percentages for the white and African American populations are rounded to the nearest whole percent, and other groups are rounded to the nearest tenth of a percent. AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

Trends in College Graduation   43

major groups: whites (non-Hispanic), African Americans (non-Hispanic), American Indians (non-Hispanic), Asians (non-Hispanic), Pacific Islanders (non-Hispanic), Mexicans, and Other Hispanics. Five of these groups (all except American Indians and Pacific Islanders) are subdivided into the native-born and foreign-born groups—the 1.5 generation is combined with the native-born. As explained in the source notes, the classification of birth cohorts is based on three major sources of data: the 1990 and 2000 censuses and the 2010 ACS. The changing composition of the American population is shown across birth cohorts identified by period of birth from 1920–1924 to 1980–1984 (in subsequent tables and charts, birth cohorts

Hispanic Asian American (%) Native Born

Foreign Born

0.5 0.5 0.5 0.4 0.4 0.5 0.5 0.6 0.9 1.3 1.9 2.4 3.0

1.1 1.4 2.4 2.8 3.5 3.8 3.8 3.8 4.0 4.4 4.6 3.9 2.9

Mexican American (%)

Other Hispanic (%)

AIAN (%)

NHOPI (%)

Total

Total

Native Born

Foreign Born

Native Born

Foreign Born

0.4 0.5 0.5 0.6 0.5 0.6 0.6 0.7 0.7 0.7 0.7 0.7 0.7

0.1 0.1 0.1 0.1 0.2 0.2 0.2 0.2 0.2 0.3 0.3 0.3 0.3

1.4 1.8 1.6 1.6 2.1 2.3 2.7 3.0 3.6 4.6 5.7 7.0 8.3

0.7 0.9 1.3 1.6 2.1 2.2 2.7 3.4 4.3 5.7 7.0 6.5 5.0

0.5 0.5 0.9 1.0 0.9 1.2 1.3 1.8 2.2 2.5 3.0 3.5 4.2

1.5 1.9 2.3 2.6 2.7 2.5 2.6 2.8 3.3 3.7 3.7 3.7 2.9

44   From High School to College

are indexed by the year the members of the cohort turned age twenty; for example, the birth cohort 1920–1924 is labeled as 1940–1944). The American population is in the midst of a significant shift in race and ethnic composition, largely driven by major waves of immigration from Latin America and Asia since the 1970s.15 The most dramatic change in this series is the decline of the white non-Hispanic population from 85 percent of the birth cohort of 1920–1924 to 60 percent of the birth cohort of 1980–1984. Immigrants were a large share of the white population in the first two decades of the nineteenth century but not for the period shown in table 2.2. For the youngest cohort—those born in the early 1980s, less than 2 percent of non-Hispanic whites were foreign born. The reduced share of non-Hispanic whites has been accompanied by the growth of other racial and ethnic groups: a five-percentage-point increase (from 8 to 13 percent) in African Americans, a four-point increase (from 2 to 6 percent) in Asian Americans, an eleven-point increase (from 2 to 13 percent) in Mexican Americans, and a five-point increase (from 2 to 7 percent) in Other Hispanics. The percentage of American Indians and Pacific Islanders is also growing; however, their combined share is only about 1 percent of the U.S. population (many American Indians and Pacific Islanders report that they are multiracial, which lowers their counts in a mutually exclusive classification system). Immigration has been the major driver of change in ethnic makeup of the American population, especially the increasing share of Asian Americans, Mexican Americans, and Other Hispanic Americans. But for the youngest cohort in table 2.2 (those age twenty in the early 2000s), the native-born (which includes the 1.5-generation population in this table) shares of these populations are actually larger than the foreign-born share. This is particularly pronounced for the Mexican American population.

Trends in College Graduation by Race-Ethnicity Figure 2.3 traces by race-ethnicity the percent of successive generations who have graduated from college based on census data from 1990, 2000, and 2010. (The actual data points for figure 2.3 and subsequent graphs are presented in appendix table 2.A1, available online.)16 The racial and ethnic groups are limited to the native-born population (including the 1.5 generation of each group). The birth cohorts are labeled by the approximate years in which they turned twenty years old—the collegegoing years. The graduation rates for native-born Asian Americans rose to above 50 percent in the late 1960s and then approached 60 per­ cent in the late 1980s and early 1990s. The vertical axis tops off at 40 percent, which truncates the Asian American trend, but allows for a closer examination of all the other groups. The extraordinary college

Trends in College Graduation   45 Figure 2.3    Percent of College Graduates in the Native-Born U.S. Population by Race-Ethnicity and Birth Cohort (Year Cohort Turned Age Twenty) 40%

Asian American White

30 Other Hispanic

25 20

NHOPI

15 AIAN

10

African American

5

4 00

9

–2

99 20

00

4 99

–1

19

95

9

–1

98

19

90

4

–1

98 19

85

9

–1

19

80

4

97 –1

97 19

75

9 96

–1

19

70

4

–1

19

65

9

96 –1

95

60 19

95

–1

19

19

55

9

–1

94

4

–1

94

45

–1 40

19

19

4

Mexican American

0

50

Percent college graduates

35

Year cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015. Notes: AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

graduation rate of native-born Asian Americans since 1960 is literally “off the chart.” The trends in college graduation rates in figure 2.3 can be summarized as follows: • College graduation rates rose rapidly from the 1940s to the late 1960s. • College completion stagnated from the early 1970s to the early 1980s. • College completion showed a modest uptick from the late 1980s to the early 2000s. Before World War II, a college education was largely limited to individuals from elite backgrounds.17 After the war, college graduation increased from a little over 10 percent to almost 30 percent for the cohort reaching the college-going years in the late 1960s. Richard Alba labels the three decades following World War II as an era of “non-zero-sum” mobility, when the children and grandchildren of southern and eastern European

46   From High School to College

immigrants were able to attend college, obtain good jobs, and find good housing in the suburbs on par with old-stock white Americans.18 The slight decline in the proportion of college graduates for cohorts in the 1970s and early 1980s was initially interpreted as a reversion to a more “normal” level of college-going. The assumption was that college enrollments were artificially boosted in the 1960s by young men seeking to avoid military conscription. Although the end of the draft might have been a short-term factor, the longer-term historical perspective shows that the 1970s represented the end of the expansion of higher education in the United States.19 The first wave of baby boomers, born in the late 1940s, were at the tail end of the era of educational expansion in the late 1960s, but the large baby-boom cohorts of the 1950s entered their college-going years at a time when higher education was hit by cutbacks in state budgets.20 This resulted in a decline in the college graduation rate among young adults in the 1970s and 1980s. Figure 2.3 shows a modest and selective uptick in college graduation rates among young white and Asian adults in the 1990s. As noted earlier, the most recent data points (for the 2000–2004 and 1995–1999 cohorts) are probably underestimates of eventual completion of a bachelor’s degree because more people in their late twenties and early thirties have postponed completing college (they had not yet completed a degree by 2010 but will eventually do so). In the oldest cohort represented here (those born in the early 1920s), Asian Americans graduated from college at about the same rate as whites. However, this gap widened dramatically over the next three decades as graduation rates of Asian Americans increased significantly. During the period of expansion, the graduation rate increased to one-third for whites and to over half for Asian Americans. Historically disadvantaged minorities—African Americans, American Indians, Mexican Americans, Other Hispanics, and Pacific Islanders—also experienced growth in college completion but at a slower pace, and they plateaued at a lower level. For example, in the late 1960s (the cohort born in the late 1940s), the college graduation rate of most disadvantaged minorities was in the midteens—only half that of whites. Other Hispanics reached a slightly higher graduation rate, a bit above 20 percent. The overall stagnation of college graduation rates in the 1970s is evident for all groups, though there are significant variations in duration and levels. Whites experienced a modest decrease that persisted for about fifteen years. For Asians, there was a brief lull in the early 1970s before growth resumed (see appendix table 2.A1). Disadvantaged minorities, who experienced much slower growth in the 1950s and 1960s, saw an extended plateau period with little change through the 1980s. Pacific Islanders (native-born), who had made substantial progress in the early 1960s, experienced a marked decrease in the 1970s before growth resumed in the 1980s. Other Hispanics (the small native-born group consisting

Trends in College Graduation   47

of Puerto Ricans, Cubans, and other Latin Americans) made continued progress in the late 1960s and only experienced a modest lull. The uptick in college completion in the late 1980s and the 1990s was experienced unevenly across groups: native-born Asians gained about ten percentage points (rising from 50 percent to 60 percent), and whites about five to six points. The gains for other groups were much more modest, in the range of one to two points. The general pattern here is one of widening race and ethnic disparities in college graduation rates. During periods of expanding educational opportunities, advantaged groups experienced faster gains in college graduation. Among the oldest cohort, the college graduation rate of nativeborn African Americans was about eight percentage points below that of native-born whites; among the youngest cohorts (those in their early twenties from 1995 to 2004), the gap expanded to eighteen percentage points. This trend is even more pronounced when comparing younger American Indians and native-born Mexican Americans with native-born whites. The gaps in college graduation rates between whites and most minorities are generally over twenty percentage points for recent cohorts (36 to 37 percent compared to 13 to 15 percent). These findings are consistent with other research that shows persistent (and sometimes widening) gaps in college graduation by family socioeconomic status.21

Modeling College Graduation with Educational-Transition Ratios Educational attainment is not a single event but the outcome of an agegraded temporal process of school enrollment from early childhood through adolescence and even into adulthood. The traditional census question on educational attainment asked for the “highest level of schooling completed,” often summarized as the “total years of schooling.” These phrases are not synonymous, as grade repetition adds to years of schooling but not to educational advancement. The standard census question on educational attainment and the check-box categories were revised in 1990 to emphasize the completion of degrees or levels of schooling.22 The analytical shift from “years of schooling completed” to “educational credentials” better reflects the role of education as a qualification for employment and other spheres of life. In this study, we focus on college graduation (obtaining a bachelor’s degree) as a key marker of educational and subsequent socioeconomic stratification in American society. College graduation is the product of cumulative transitions—or “educational-transition ratios,” the probabilities of moving from one grade or level to the next grade or level—from first grade through college graduation.23 Educational-transition ratios are conditional probabilities, with the denominator including only persons who have attained the

48   From High School to College

immediate preceding step—those who are exposed to the risk of transitioning to the next step. They are analogous to many other demographic measures, including age-specific probabilities of mortality (qx values) in a life table or parity-progression ratios in fertility analysis. Since not all educational transitions are similar, or have the same causes, analyses of educational-transition ratios are useful indicators for analyzing temporal trend differentials in educational progress by gender, race-ethnicity, and other subgroups.24 For example, the effect of family socioeconomic status is not the same for high school completion, college entry, and college completion.25 Understanding the process of educational stratification is a fundamental objective of this study. College graduation (CG) is the proportion of persons in a birth cohort who graduate from college: CG = graduates/population. CG is the product of three educational-transition ratios: • the proportion of all persons who complete high school: HS = high school graduates/population • the transition rate of high school graduates who enter college: CE = college entrants/high school graduates • the transition rate of college entrants who complete college: CC = college completers/college entrants This gives us the formal identity: CG = HS × CE × CC The logic of this simple model allows for quantitative estimates of the relative changes in high school completion, the transition from high school to college, and college completion on trends and differentials in college graduation. We begin with a simple description of the historical trends in these three critical educational transitions. Figure 2.4 shows the three critical educational transition ratios—HS, CE, and CC—for the thirteen birth cohorts that turned age twenty between 1940–1944 and 2000–2004, based on merged census data series for the total U.S. population. The summary college graduation rate is indexed by the legend on the vertical axis along the right-hand side. The three educational-transition ratios are shown with the index on the vertical axis along the left-hand side. The college graduation rate in figure 2.4 shows the trends of rapid expansion (from 1940–1944 to 1965–1969), stagnation (from 1970–1974 to 1980–1984), and modest uptick (from 1985–1989 to 1995–1999). The high school graduation rate has a straightforward trajectory—rising from a little over 60 percent in the early 1940s to 87 to 88 percent in the late 1960s, with little change since then. The uptick to about 90 percent in the CPS

Trends in College Graduation   49 Figure 2.4    Percent of College Graduates and Educational-Transition Ratios for the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty)

25

High school graduation rate (based on left-hand scale)

80

35%

15 Ratio for transition from high school to college enrollment (based on left-hand scale)

60

5

4

9

00

00

–2

4

99

–1

95

–5

20

9

99

–1

90

19

4

98

–1

85

19

9

98

–1

80

19

4

97 19

75

–1

97

–1

70

19

4

96

65

–1

9

96

–1

60

19

4

95

–1

55

19

9

95 19

19

50

–1

94

–1

45

19

19

40

–1

94

4

40

9

Ratio for transition from college enrollment to completion (based on left-hand scale)

Percent that graduated college

College graduation rate (based on right-hand scale)

19

Educational transition ratio

100%

Year cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015.

series (see figure 2.1) is not evident in the census series that only extends to the early 2000s. Some analysts suggest that rates of regular, on-time high school graduation may have declined in recent years due to the rise in alternative credentials to high school graduation, such as the GED.26 Until recently, census and CPS data did not distinguish between GED completion and high school graduation. Recent studies suggest that 8 to 13 percent of high school completers hold a GED or other high school equivalency credential.27 The second educational-transition rate—for the move from high school to college—is perhaps the most consequential step in the early life course. Using data from the 1960 census, Beverly Duncan reported a decline in the transition from high school to college (measured as the conditional probability from twelve to thirteen or more years of schooling) during the first half of the twentieth century as high school completion became more common.28 Robert Mare showed that this decline was only evident for cohorts born in the early twentieth century.29 Mare found that the transition from high school to college increased for cohorts born in the 1940s. With our merged census data, we find that the transition from high school to college rose steadily from less than 50 percent for those aged

50   From High School to College

twenty in the 1940s to about 67 percent in the late 1960s. Then there was a period of stagnation, actually a decline of a couple of percentage points in the 1970s and early 1980s. There were modest increases in the late 1980s and 1990s, with the transition rate from high school graduate to college entrant reaching 72 percent in the 1990s and early 2000s. Interpretations of changes in the transition from high school to college are complicated by the growth of community colleges and other forms of postsecondary schooling in recent decades. Tabulations from the NCES show that the rise in college enrollments in the late 1980s and 1990s has coincided with increases in the enrollments of students in two-year institutions.30 Although part of the increase in college enrollment (from 25 to 41 percent of persons aged eighteen to twenty-four) from 1973 to 2012 was partially due to more students attending two-year colleges (twoyear-college attendance rose from 7 to 13 percent of persons aged eighteen to twenty-four), most of the increase was “real”—that is, it reflected increased enrollments in four-year colleges. The proportion of eighteen- to twenty-four-year-olds enrolled in four-year colleges rose from 17 percent in 1973 to 28 percent in 2012. The ratio for the third educational transition—from college enrollment to college completion—has increased slightly across the cohorts represented in figure 2.4. The percentage of college entrants who completed college was in the mid- to high-40-percent range in the 1940s and 1950s. The figure rose a tiny bit to 50 percent in the late 1960s and then declined slightly during the era of stagnation in the 1970s and early 1980s. The college completion rate bounced back to 50 to 51 percent in the 1990s. A college dropout rate of 50 percent might be interpreted as a sign that colleges and universities are remarkably ineffective. However, this interpretation assumes that all college entrants plan to obtain a four-year college degree. About 40 to 50 percent of college entrants are enrolled in two-year colleges.31 Many of these students only seek to take a few vocational courses or to earn an associate’s degree. However, there is still a high dropout rate among students enrolled in four-year colleges.

Trends in Educational-Transition Ratios by Race and Ethnicity In the next three figures, we present trends in high school graduation rates (figure 2.5), the transition from high school to college enrollment (figure 2.6), and the transition from college enrollment to college completion (figure 2.7) by major racial and ethnic groups from the early 1940s (birth cohort for early 1920s) to the period from 2000 to 2004 (birth cohort of early 1980s). Because our focus is on American schools, these trends are restricted to the native-born population (including immigrants who arrived before age twelve). Thus, the race-ethnicity comparisons are not

Trends in College Graduation   51 Figure 2.5    Percent of High School Graduates in the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

Asian American

75

Other Hispanic African American

White NHOPI

50

AIAN

4 00

20

00

–2

9

4 19

95

–1

99

99

9

–1

98

90

19

98

85 19

–1

–1

4

9 97 19

80

–1

19

75

97

4

9 96 19

70

–1

4

–1

96

65 19

–1

95

60 19

95

55

19

50 19

–1

4

9

4

94

–1

94

45

–1 40

19

19

9

Mexican American

25

–1

Percent high school graduates

100%

Year cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015. Notes: AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

affected by the inclusion of immigrants who completed their education in their country of origin. In general, the trajectory of a rapid increase in high school graduation rates for the first half of the period (roughly from 1940–1944 to 1965– 1969) followed by a plateau is evident for each racial-ethnic group. The wide gap in rates between whites and disadvantaged minorities (African Americans, American Indians, and Mexican Americans) of thirty to forty percentage points in the 1940s and 1950s narrowed to ten to twenty percentage points by the late 1960s. There were only modest changes in the gaps after this point. High school graduation rates for whites and Asians (and Pacific Islanders) plateaued at around 90 percent to 95 percent, while the rates for African Americans, American Indians, Mexicans, and Other Hispanics were about ten to fifteen points lower. The racial and ethnic gaps in these comparisons are likely underestimates because of the selective undercounts in census data (of minorities and of the least educated).32 Another major limitation is the inclusion of

52   From High School to College

persons with high school equivalency certificates as high school graduates. About 8 to 14 percent of those who reported completing high school in the 1990s and early 2000s were likely to be dropouts who received an alternative equivalency certificate.33 The “true” graduation rates of racial and ethnic minorities were probably considerably lower than the rates reported in figure 2.5. Research based on administrative data (school records) suggests that the on-time graduation rate of African Americans and American Indians was closer to 50 percent.34 Although the GED provides certification of high school completion, the earnings of GED recipients were similar to those of high school dropouts.35 Figure 2.6 shows three distinct trends in the transition from high school graduation to college entry for Asian Americans, whites, and disadvantaged minorities. The most distinctive trajectory is for nativeborn Asian Americans. In the 1940s, when native-born Asian Americans constituted an extremely low portion of the U.S. population, about 40 to 45 percent of Asian American high school graduates went to college—

Figure 2.6    Educational-Transition Ratios for the Transition from High School Graduation to College Enrollment in the Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

Educational transition ratio: high school graduation to college enrollment

100% Asian American 80

60

Other Hispanic

White

AIAN Mexican American

40

African American

NHOPI

19

19

40 –1 9 45 44 –1 19 94 50 9 – 19 195 55 4 – 19 195 60 9 – 19 196 65 4 – 19 196 70 9 – 19 197 75 4 – 19 197 80 9 – 19 198 85 4 – 19 198 90 9 – 19 199 95 4 – 20 199 00 9 –2 00 4

20

Year cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015. Notes: AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

Trends in College Graduation   53

a bit lower than the rate for white students at that time. By the early 1960s, about 80 to 85 percent of Asian American high school graduates enrolled in college—a figure that was about fifteen percentage points higher than the rate of white graduates. The Asian American collegeenrollment rate has fluctuated a bit over time but has more or less stabilized at just under 90 percent. It is only a slight exaggeration to say that virtually all Asian American high school students attend college.36 This is all the more remarkable given the growth in the size and composition of the Asian American population. In addition to long-resident Chinese American and Japanese American populations in California, Hawaii, and New York, there are growing numbers of Filipinos, Vietnamese, Koreans, Asian Indians, and many smaller national-origin groups in many regions of the United States. Because whites are the largest segment of the total population, their trend in college attendance largely parallels that of the total population. During the era of expansion from the 1940s to the late 1960s, college enrollment among white high school graduates rose from 50 percent to just shy of 70 percent—a remarkable gain that appears modest only when compared to the even faster rise among Asian Americans. Over the next decade, when peak baby-boomer cohorts were reaching college age in the 1970s, white college enrollments dipped a few percentage points. By the late 1980s and mid-1990s, however, the white college-enrollment rate (among high school graduates) rose to about 75 percent. In the 1940s, the transition rates from high school graduate to college student among native-born disadvantaged minorities (African Americans, American Indians, Mexicans, and Other Hispanics) were only two to five percentage points below white and Asian American levels. This “slight” disadvantage of minorities in the educational-transition ratio from high school to college must be understood in the context of the times, when minority high school graduates were a relatively select population. In the 1940s, only 30 to 40 percent of minorities graduated from high school, which was well below the comparable white level. During the era of the great expansion in college enrollment—the 1950s and 1960s—the educational transition from high school to college for minorities did rise, but not nearly as fast as it did for white students. The gap between whites and minorities in the transition to college enrollment widened to about 5 to 10 percentage points from the 1950s to the 1960s, in contrast to the narrowing of the gap in high school graduation rates during the 1960s. By the early 2000s, about 60 to 65 percent of minority high school graduates went on to college, which was about ten percentage points below the level for whites and about twenty to twenty-five points below the level of Asian Americans. Figure 2.7 shows the trend in college completion among college entrants for successive cohorts of native-born American youth from the

54   From High School to College Figure 2.7    Educational-Transition Ratios for the Transition from College Enrollment to College Completion in Native-Born U.S. Population by Birth Cohort (Year Cohort Turned Age Twenty) and Race-Ethnicity

Asian American NHOPI

60

White Other Hispanic 40

4 00

9 99

–2

20

19

00

4

–1

9

99 –1

19

90

98

4

AIAN

–1

19

85

9

98 –1

97

80 19

97

–1

19

19

75

9 96

–1

4

–1

19

65

9

96 –1

19

60

4

95 –1

95 19

55

9 94

–1

19

50

4 94

–1 45

–1

19

40 19

4

African American

95

Mexican American 20

70

Educational transition ratio: college enrollment to completion

80%

Year cohort turned age twenty Source: Author’s compilation from Ruggles et al. 2015. Notes: AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

1940s to the early 2000s. In contrast to the earlier eras of expansion and growth, there was only a slight hint of improvement in the transition from college enrollment to college completion. The only exception was among Asian Americans. The Asian American college-completion rate (number of graduates divided by number of entrants) rose to percentages in the mid-50s in the 1960s, then to percentages in the mid-60s in the 1980s, and then to about 70 percent in the late 1990s. The dip in the early 2000s is almost certainly misleading because college completion is censored (more will eventually graduate) for the cohort aged twenty-five to twenty-nine in 2010. Among whites, the college-completion ratio rose a few percentage points to about 51 to 52 percent in the late 1960s and then declined a few points in the 1970s and early 1980s, followed by a slight increase to 52 to 53 percent in the 1990s. The race and ethnicity differentials in college completion in figure 2.7 are largely comparable to six-year graduation rates based on institutional data.37

Trends in College Graduation   55

If stability in college-completion rates describes the trend among whites, college completion among disadvantaged minorities can be characterized as “losing ground.” The completion rate for African Americans was roughly 35 to 39 percent during most the 1950s and 1960s but then dropped to percentages in the low to mid-30s after the mid-1980s. The college-completion rate for American Indians has dropped to below 30 percent. College completion for native-born Mexican Americans rose a few points in the 1960s to percentages in the low 30s and has fluctuated around 30 percent for the last few decades. The college completion rate for native-born Other Hispanics (which includes Cubans) has been about ten percentage points higher and has been around 41 to 42 percent in recent decades. There have been much wider racial-ethnic gaps in rates of college completion compared to other educational-transition ratios. More importantly, the racial-ethnic gaps in college completion have shown no sign of narrowing in recent decades.

Components of Intercohort Changes in College Graduation Rates by Racial and Ethnic Groups The trends and disparities in the three educational ratios are so detailed that even a dedicated reader may lose sight of the overall picture. To summarize the varied patterns, we use the method of demographic decomposition to show the absolute contribution of each educationaltransition ratio to temporal changes in college graduation rates for each race-ethnicity group. The results, presented in table 2.3, show the overall change in college graduation rates (CG) that can be attributed to changes in three components: changes in high school graduation rates (HS), changes in the transition from high school to college entry (CE), and changes in the transition from college entry to college completion (CC). In other words, this table shows the changes in college graduation in terms of the trends exhibited in figures 2.5, 2.6, and 2.7. The change in college graduation rates in the total population is shown for three time periods and for seven race-ethnicity groups by nativity. The time periods, indexed by the years that the birth cohorts reached age twenty, are identified as the eras of expansion (1940–1944 to 1965–1969), stagnation (1965–1969 to 1980–1984), and slow growth (1980–1984 to 2000–2004). As noted earlier, the most recent data point (for persons in the college-going years in the early 2000s) is right censored; the eventual rate will probably be five points higher than the numbers reported here.38 Keeping with the standard practice, the 1.5 generation (immigrants arriving before age twelve) are included as part of the native-born group, since it is likely that all or part of their schooling was in the United States.

56   From High School to College Table 2.3    Decomposition of Intercohort Change in College Graduation Rates Attributable to Educational-Transition Ratios for High School Graduation, Transition to College, and College Completion in the Native-Born U.S. Population by Race-Ethnicity Birth Cohorts Indexed by Year Turned Age Twenty 1940–1944 to 1965–1969

Total population Native-born population White African American AIAN Asian NHOPI Mexican Other Hispanic

Percentage Pts. Total Change in CG

D HS

D CE

D CC

Percentage Pts. Total Change in CG

17.5

6.9

7.1

3.6

-2.3

19.4 11.6 11.6 37.3 13.2 11.3 15.2

7.1 9.1 7.8 7.5 6.0 6.9 7.2

8.2 3.0 3.0 19.5 4.4 2.6 5.0

4.2 -0.5 0.8 10.3 2.8 1.8 3.0

-2.4 -0.5 -4.4 2.8 0.1 -0.4 0.2

Decomposition Attributable

Source: Author’s compilation from Ruggles et al. 2015. Notes: CG = the college graduation rate, is the proportion of the total population that graduated from college. HS = the high school completion rate, is the proportion of all persons who completed high school. CE = the college entry rate, is the proportion of high school graduates who enrolled in college. CC = the college completion rate, is the proportion of college entrants who completed college. AIAN = American Indian and Alaskan Native. NHOPI = Native Hawaiian and Pacific Islander.

The proportion of the total population that graduated from college increased across birth cohorts by 17.5 percentage points from the 1940s to the late 1960s (more than doubling from 12 to 30 percent). The major contributions to the total change were rising high school graduation rates and increasing proportions of high school students going on to college (each accounted for seven percentage points). Increases in college completion contributed a smaller component—about four points of the total increase in college graduation. The metaphor “a rising tide lifts all boats” aptly describes the impact of expanding educational opportunities during the 1950s and 1960s. All racial-ethnic groups registered

Trends in College Graduation   57

1965–1969 to 1980–1984

1980–1984 to 2000–2004

Change HS

Change CE

Change CC

Percentage Pts. Total Change in CG

0.1

-0.5

-1.9

2.6

-0.1

2.3

0.3

0.1 1.2 0.2 0.6 1.3 0.7 1.0

-0.7 0.2 -2.4 1.9 1.5 0.4 1.4

-1.8 -1.9 -2.2 0.2 -2.7 -1.5 -2.2

5.9 1.1 -3.3 0.8 -3.2 -0.5 -0.9

0.2 0.0 -0.1 0.0 0.7 0.5 0.0

3.7 1.7 0.5 -0.1 -0.7 0.3 0.4

2.0 -0.5 -3.7 0.9 -3.2 -1.3 -1.3

Decomposition Attributable

Decomposition Attributable Change HS

Change CE

Change CC

double-digit percentage-point increases in college graduation rates. But the gains were skewed. White students increased their rates by more than nineteen percentage points, and Asian students registered a staggering increase of thirty-seven percentage points. Other minority groups gained about eleven percentage points (Other Hispanics and Pacific Islanders had slightly larger gains). The decomposition results provide a clear picture of the reasons for increases in college graduation rates across the different racial-ethnic groups. Rising rates for high school graduation were important for all groups—generally accounting for about seven to nine percentage

58   From High School to College

points in the rise in college graduation rates during the period of expansion. Higher rates of college entry and rising levels of college completion were important, especially for white and Asian students. These two components—access to college and college completion—were responsible for eight and four percentage points, respectively, of the nineteenpoint gain for whites, and nineteen and ten points, respectively, of the thirty-seven-point gain in college completion for Asian Americans. Minorities had much smaller gains in college access and completion. In an ominous sign, African Americans had a slight decline in their rates of college completion during the era of expansion. During the time of widespread growth in higher education in the 1950s and 1960s, most of the gains in college graduation rates by minorities can be attributed to the rising numbers of high school graduates who were eligible to go on to college. Minorities experienced much less progress in transition rates from high school to college enrollment and from college enrollment to completion. More rapid increases in these transition rates for whites contributed to a widening gap in college graduation rates between whites and disadvantaged minorities. The one minority group that had a completely different experience during the educational expansion was Asian Americans. The increase in their college graduation rates from the early 1940s to the late 1960s was nothing short of spectacular. These gains were primarily attributable to increases in the transitions from high school to college (almost twenty percentage points) in college completion (about ten points). These gains occurred before the major waves of Asian immigration that took off in the 1970s. The 1970s and 1980s saw stagnation in college graduation rates. The only exception to this pattern was among Asian Americans, who registered a small increase of three percentage points in college graduation, primarily because of a higher rate of college entry. But most other groups, including whites, made no progress or actually lost ground during the 1970s and early 1980s. The largest negative factor during this period was the decline in college completion among college entrants. Recall that college-completion rates were only about 50 percent for most groups, but the dropout rate rose and contributed to lower graduation rates for cohorts coming of age in the 1970s and 1980s than those in prior generations. With all but universal stagnation, there was no progress in closing the wide racial and ethnic gaps during this era. The moderate uptick in college graduation during the 1990s and early 2000s was highly skewed. White students gained about six percentage points in their graduation rate, primarily because of higher transition rates of college enrollment and completion. The lack of progress of Asian Americans (and some other groups) may simply be due to postponement—they might well complete college in a few years, but this is not yet evident in the data from 2010. Asian American stu-

Trends in College Graduation   59

dents gained seven percentage points in their graduation rate from the cohort that reached age twenty in 1980–1984 to the cohort that reached age twenty in 1995–1999 but then had a six-point decline for the most recent cohort. This bias (due to right-censoring in the data) is evident for all groups. There was only a one-point gain in the graduation rate of African Americans and a continued negative (though small) trend in college graduation for other minorities. College enrollment was generally stable, but there were declines in college completion for all minorities except Asians. The popular perception that the American educational system is one of “college education for all” does not accurately describe the trends and patterns reported here. The era of growth for the immediate post–World War II era—the 1950s and 1960s—did represent a decisive break from the past. Before the war, college was limited to a relatively small segment of the population—less than 10 percent graduated from college. During the period of expansion, high school graduation rates climbed steadily, and each cohort was more likely to enroll and graduate from college than the preceding one. The high-water mark was the 1945–1949 birth cohort, the first wave of the baby boomers who were in the college-going years in the mid- and late 1960s. One-third of white students in this generation graduated from college, as did more than one-half of Asian Americans. Slower progress was registered for other minorities—only 14 to 18 percent of minorities graduated from college in the late 1960s. For the next fifteen years, college graduation rates stagnated, and rates for some groups actually declined. In addition to slower rates of entry into college, rates of college completion (among those who began college) declined during this period for all groups, except Asians. The increasing share of students beginning higher education in community colleges may account for some of the decline in rates. The uptick in college enrollments and graduation in the 1990s was much more modest than that in the 1950s and 1960s. Moreover, increases in graduation rates were concentrated among whites and Asians. For the 1975–1979 birth cohort that was entering college in the mid- and late 1990s, about 37 percent of native-born whites and 60 percent of native-born Asian Americans earned a college degree. The comparable figures are 19 percent of African Americans, 13 percent of American Indians, and 15 percent of native-born Mexican Americans (see figure 2.3). These patterns do not suggest that the wide racial and ethnic gaps are likely to narrow in the near future. These national trends set the stage for the next chapters, in which we present the context and results of a study of the pathways to college graduation among high school seniors in the Pacific Northwest—the UW-BHS study. The intensive case study attempts to explain how and why raceethnicity, gender, and immigrant generation influence pathways from high school to college graduation.

Chapter 3 The University of WashingtonBeyond High School Project: Data and Description with nikolas pharris-ciurej

T

UW-BHS project originated in the late 1990s, partially by intention and partially by happenstance. Several faculty members at the University of Washington were engaged in discussions about potential research on racial and ethnic stratification in the Pacific Northwest. The region seemed different from the rest of the country— perhaps providing a window on future directions in American society. The Pacific Northwest, along with much of the West Coast, has a much different racial and ethnic composition from the rest of the United States. Although whites are still a majority (at least for the moment), the region is home to a wide diversity of peoples of every hue and nationality, including African Americans, American Indians, Hispanics, Pacific Islanders, and Asians from almost every nation across the Pacific.1 Even whites have migration stories, and local residence is rarely longer than a generation. This diversity, combined with a boom-and-bust economy, created a social atmosphere in the 1990s that was more focused on opportunities and change than memories of past conflicts. The election of blacks and Asian Americans to high political positions by a predominately white electorate at that time was certainly a pacesetter for the rest of American society.2 These inchoate ideas began to take shape in the spring of 1999, when the Andrew W. Mellon Foundation asked if I might be interested in writing a research proposal to evaluate the impact of Initiative 200, a referendum passed by the voters of Washington State in November 1998 banning affirmative action. The research question was whether the elimination of affirmative action would have an adverse effect on the college enrollment he

60

University of Washington-Beyond High School   61

of minorities in the state. In the application to the Mellon Foundation, we proposed a two-part project. The first part was a secondary analysis of administrative records to compare college enrollments of Washington State high school seniors before and after Initiative 200. This analysis was published a few years later.3 The second part was a longitudinal study of high school seniors tracking their transition from high school to college— a study partially inspired by the collegial discussions on race and ethnic stratification in the Pacific Northwest. The Mellon Foundation provided support for the initial wave of data collection, and the UW-BHS project launched in 2000. A longitudinal study of the transition of students from high school to college is a well-established genre of sociological and education research. Perhaps the leading exemplar, and a particular inspiration for UW-BHS, was the Wisconsin Longitudinal Study.4 Other influential studies were the Children of Immigrants Longitudinal Study,5 the National Study of Adolescent Youth,6 and the Baltimore Beginning School Study Youth Panel.7 These studies illustrate the value of a longitudinal research design that tracks adolescents to early adulthood and beyond. The published research from these projects constitutes much of the corpus of basic knowledge on educational careers, intergenerational mobility, and racial and ethnic disparities in American society, even though the data were often based on local or regional populations. The other model for the UW-BHS project was the series of longitudinal surveys conducted by the National Center for Education Statistics (NCES) in the U.S. Department of Education. Many of the items in the UW-BHS questionnaire for high school seniors were taken from NCES surveys, especially the National Educational Longitudinal Survey.8 The UW-BHS survey questionnaire also borrowed standard items from the U.S. Census, the American Community Survey (ACS), and other national surveys.9 School-based samples hold many advantages over conventional household surveys for research on high school students. Only a small minority of households contain high school–age students, so a large screening survey would be necessary to find and interview a sufficient sample of students. Moreover, household surveys are subject to high nonresponse rates, a problem that would be particularly pronounced among adolescents. In contrast, with the cooperation of school authorities, it is relatively economical to survey a substantial number of students in school settings. Most school surveys, including the UW-BHS, experience low levels of student nonresponse. There are, however, sampling problems that are endemic to schoolbased research of students. School administrators are understandably reluctant to allow outsiders, even researchers, to interview students. The primary responsibility of schools is education, and any time allocated to

62   From High School to College

other activities competes with instructional time. As will be discussed, most national surveys of students, including those conducted by the federal government, encounter a significant number of schools that refuse to participate. The UW-BHS project also faced skepticism from central school district officials, school principals, and individual teachers. It was only with dogged persistence, accommodation to local concerns, and measured patience that we were eventually able to persuade every school in our purposive sample to participate. The reluctance from school officials was only evident during the first year of the project. In successive waves of data collection, the UW-BHS team was welcomed enthusiastically by school administrators and teachers. We even became collaborators on school-based research projects. We also obtained informed consent from participants, including the parents or guardians of students below age eighteen. In the end, less than 2 percent of contacted students (or their parents) declined to complete the UW-BHS senior survey. A more serious limitation of the UW-BHS project was the generalizability of a survey of high school students from a nonrandom sample of schools in the Pacific Northwest. The Pacific Northwest is not representative of the United States, and the twelve UW-BHS high schools were not drawn from a probability sample of all high schools in the region. However, the long history of regional-based studies of students in educa­ tional research has not been a problem for cumulative research on edu­ cational stratification. The findings from regional studies generally mirror those based on national data, and we believe that the research reported in this volume continues in this tradition. Two quite different aspects of generalizability are sometimes conflated. Schools differ in their composition (as do regions), and population averages (and other summary measures) are affected by the socioeconomic and ethnic makeup of enrolled students in the area. For example, regional differences in the proportion of students going to college will differ because poverty, immigrant concentrations, and many other factors vary by region. Researchers should be cautious about generalizing descriptive statistics, such as the proportion of students who graduate from college, from nonrepresentative samples. However, the primary focus of the present study is on processes and relationships, such as the effects of family background characteristics on college enrollment. Net of composition, these relationships are much less likely to vary between geographical regions. Several generations of social-stratification research show that statistical relationships between variables are remarkably robust across a broad range of data sources, including nonrandom samples.10 Despite the nonrandomness (and nonrepresentativeness) of the UW-BHS data, many of the descriptive statistics from this study, such as the proportion of high school seniors who attend and graduate from college, are very similar to those from national data. However, our focus is on

University of Washington-Beyond High School   63

explaining social processes, such as whether SES explains educational disparities by race and ethnicity. For these sorts of questions, there is more confidence that locally and regionally based samples can make an incremental contribution to the corpus of social science knowledge. We return to the question of generalizability in the concluding chapter. In the spring of 2000, the UW-BHS project began data collection in a large public school district (referred to henceforth as District 1) that had a history of positive collaboration with the University of Washington. In April and May of 2000, we conducted a baseline survey of all seniors in the five comprehensive high schools in the district. We also conducted a follow-up telephone survey in 2001 of all UW-BHS respondents to the 2000 senior survey to measure college enrollment and a few additional items. Based on the success of the project and continued support from the Mellon Foundation, the UW-BHS data collection was replicated in the spring of 2002 in the same five high schools. During our fieldwork in the spring of 2002, we learned of a major educational intervention under way in three of the five high schools in our sample. In 2001 the Bill and Melinda Gates Foundation, in collaboration with the College Success Foundation (then known as the Washington Education Foundation), launched the Washington State Achiever (WSA) program in sixteen low-income high schools in Washington State. The objective of the WSA program was to assist promising low-income high school students in continuing their education by enrolling in a four-year college.11 The centerpiece of the WSA program was the awarding of five hundred college scholarships annually to promising low-income eleventh-grade students across the sixteen high schools from 2001 to 2010. Three of the sixteen WSA high schools were in our initial sample of five UW-BHS high schools (in 2000 and 2002). The WSA program also included a plan for school restructuring and curricular reform, increased student support, and mentoring to create a “climate of college-going” in the sixteen WSA high schools. School restructuring involved the creation of small schools (dubbed academies or learning communities) within the larger physical school, with the intention of increasing accountability, familiarity, and communication between students and teachers. Additionally, a revised curriculum was designed to increase levels of college preparedness among the student population, while the increased student support and mentoring was intended to help students in the college preparation and selection process. It took several years to plan and implement all phases of the WSA program, but the first round of scholarships was awarded in 2001 and continued for ten years.12 Although we were initially surprised by the presence of the WSA program in three of the five UW-BHS high schools, it did not take long to realize that it presented an extraordinary opportunity to expand the scope of the UW-BHS project. We had already collected data on college

64   From High School to College

plans and college enrollment of high school students in three WSA high schools that could serve as an experimental group, and in two non-WSA high schools that could be considered as a control group. We also had one round of data collection (in 2000) prior to the program intervention and a second round (in 2002) following the program intervention. The overlap between the UW-BHS project and the WSA program was completely fortuitous, but it created the opportunity for a quasi-experimental evaluation of WSA and a rare opportunity for a prospective study of the long-term impact of scholarships on the lives of low-income students. With additional support from the Gates Foundation, we expanded the UW-BHS project to add more schools and three additional cohorts of high school seniors in 2003, 2004, and 2005. The geographic reach of the UW-BHS project was expanded to include two additional school districts. The additional districts were chosen because of proximity and because each contained one WSA high school and one non-WSA high school. We also expanded the UW-BHS data collection to include three nearby private high schools, which provided an additional point of comparison in measuring the transition from high school to college. The private schools were small in comparison to the public high schools. The largest was a highly regarded parochial school, another was a private Christian school, and the last was an elite secular preparatory school. The expanded UW-BHS project surveyed high school seniors in twelve high schools in 2003, 2004, and 2005: three private and nine public. The latter were located in three school districts: District 1, which included the five high schools from the initial 2000 and 2002 baseline surveys, and Districts 2 and 3, with two schools each.

Survey Administration The baseline UW-BHS survey was administered in the spring, generally in April or May of 2000, 2002, 2003, 2004, and 2005. The specific dates and times were initially selected by the school officials to minimize the loss of instructional time. In many schools, the survey was conducted during the week that statewide testing occurred, when regular classes were suspended. Senior students did not take the statewide exams but were still required to attend study halls during regular school hours. In other schools, the dates for the survey were chosen for idiosyncratic reasons. The timing and location for the administration of the survey also varied from school to school. The typical time was the first hour of the school day. Seniors were directed to a common location, usually the school auditorium or cafeteria. In a few high schools, the survey was administered in individual classrooms—usually a senior English class or another class that all seniors were required to take. Our project team developed a standard script

University of Washington-Beyond High School   65

for the administration of the survey, beginning with a brief introduction of the project and the questionnaire. Students were then asked to fill in paperand-pencil surveys as project staff walked around the room to answer questions. In general, almost all students were able to easily complete the survey within the allotted hour, usually in less than forty minutes. The refusal rate was less than 2 percent; cooperation was encouraged with the incentive of a free movie pass to students who completed a survey. We developed a series of follow-up procedures to contact students who were absent on the survey day as well as for students who did not attend one of the five major comprehensive high schools. There were a handful of small-enrollment “alternative-site schools,” primarily for students who had experienced academic or disciplinary problems in regular high schools. The survey was mailed to all seniors enrolled in alternativesite schools and to the students who were absent from one of the regular high schools on the day of the survey. The same procedures, in terms of instructions, the request for informed consent, and the offering of a movie pass as an incentive, were used for the mailed surveys. Using procedures for mailed surveys recommended by Don Dillman, we followed up the initial mailing with postcard reminders, a second mailing of the survey, and finally an express-mail delivery.13 These concerted efforts yielded about a 10-percent increase in the number of completed surveys. The senior survey collected extensive data on college ambitions and plans. Although college plans are a major predictor of college enrollment, the research literature suggests that behavior does not always follow from intentions.14 So, our research strategy included a one-year follow-up survey, scheduled for January to June of the year following the baseline data collection in the high schools. The follow-up survey contained fewer than ten questions (focused on college enrollment) and generally took less than five minutes to complete. The survey was largely conducted over the phone, but a small number of participants responded via e-mail, a web-based version of the survey, and regular mail. The follow-up surveys were conducted by trained and highly motivated University of Washington undergraduate students. The inter­ viewers initially telephoned the home number reported in the baseline senior survey. If the student was no longer at this residence (very few students had cell phones in the early 2000s), the interviewers asked for current contact information (phone, address, e-mail). In most cases, the student was located after two or three calls or e-mails, but there was no limit on the number of calls made to find and interview a UW-BHS respondent. If it was unlikely that the respondent could be reached, the interviewer conducted a proxy interview with a knowledgeable family member or friend. Since the follow-up questionnaire only contained a few objective questions about college enrollment and current employment, proxy responses were considered acceptable if the respondent

66   From High School to College

could not be contacted after multiple attempts. About 10 percent of the follow-up surveys were proxy interviews.

Survey Coverage The primary source of data for this volume is the UW-BHS “core sample” of 9,658 high school seniors from the twelve metropolitan Pacific Northwest high schools that were initially interviewed from 2000 to 2005.15 The core sample is limited to confirmed high school seniors, excluding a handful of students who were foreign exchange students, had severe developmental disabilities, or filled in the survey with random responses. An overview of the UW-BHS data collection is presented in table 3.1. The top panel shows the numbers of seniors who responded by year and school district, and the middle panel reports the numbers of UW-BHS respondents who were successfully reinterviewed in the oneyear follow-up survey. The lower panel reports the coverage rate of the follow-up survey—the percentage of high school seniors in the baseline survey who were located and interviewed one year later. The first two rounds of the senior survey in 2001 and 2002 were limited to five high schools in District 1. The classes of 2003, 2004, and 2005 contained high school seniors from District 1 and also students from the four public high schools in Districts 2 and 3 and from three private schools. Roughly 1,100 students completed the baseline senior survey each year from District 1. About one thousand public high school seniors and three hundred private high school seniors were added to the UW-BHS data base each year from 2003 to 2005 from four high schools in Districts 2 and 3 and three private schools, respectively. The matched data file of high school seniors who were interviewed one year later includes 8,888 students, about 92 percent of the baseline senior sample. The last panel in table 3.1 shows a gradual improvement in coverage rate over the five years of the project. For example, 88 percent of the class of 2000 in District 1 was successfully reinterviewed; this percentage grew to 94 percent for the class of 2005. Virtually 100 percent of the private school students were contacted and reinterviewed. Private school students, primarily from higher-income families, were much easier to track and locate. The respondents lost to follow-up consisted primarily of students who had few ties with family and friends.

Coverage of High School Seniors in the UW-BHS Baseline Senior Survey With a reinterview rate of over 90 percent, the follow-up sample is very representative of the UW-BHS baseline sample of interviewed high school seniors. A more complex and complicated question is whether the baseline

1,226 1,005 285 2,516

2003 1,077 985 307 2,369

2004 1,151 942 334 2,427

2005

1,107 911 284 2,302

2003 969 927 304 2,200

2004

1,088 888 330 2,306

2005

89.2% — — 89.2

2002 90.3% 90.6 99.6 91.5

2003

90.0% 94.1 99.0 92.9

2004

94.5% 94.3 98.8 95.0

2005

Coverage Rate of the One-Year Follow-Up Survey

1,061 0 0 1,061

2002

UW-BHS Seniors Reinterviewed in the Follow-Up Survey

1,189 0 0 1,189

2002

90.4% 93.0 99.1 92.0

All Years

5,244 2,726 918 8,888

All Years

5,800 2,932 926 9,658

All Years

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: Core respondents are defined as confirmed high school seniors, excluding foreign-exchange students, students with developmental disabilities, and students who answered with random responses. UW-BHS = University of Washington-Beyond High School.

2000

1,019 0 0 1,019

District 1 (five schools) Districts 2 and 3 (four schools) Private (three schools) All schools

88.1% — — 88.1

2000

One-Year-Follow-Up

Coverage Rate of Follow-Up

1,157 0 0 1,157

District 1 (five schools) Districts 2 and 3 (four schools) Private (three schools) All schools

District 1 (five schools) Districts 2 and 3 (four schools) Private (three schools) All schools

2000

Senior Survey

Number of Respondents in Baseline UW-BHS Senior Survey

Table 3.1    Core Respondents in the UW-BHS Baseline Survey and in the One-Year Follow-Up Survey: 2000 to 2005

68   From High School to College

sample of UW-BHS high school seniors is representative of all high school seniors listed in the rosters (administrative records) of students enrolled in the targeted high schools. The UW-BHS project aimed to interview all eligible students in the baseline senior survey. To insure a high response rate, we invested considerable time in meetings with school administrators, teachers, and students to explain the objectives of our research and to answer questions. We solicited newspaper stories to raise awareness about the project and its objectives. Yet, when we compared the rosters of enrolled high school seniors with the students that completed the senior survey, there were many inexplicable anomalies. A significant number of supposedly currently enrolled seniors did not complete the survey—a figure much higher than normal absenteeism. Additionally, a number of students showed up to take the survey, claiming to be seniors, who were not on the schools’ lists of enrolled seniors. Evaluation of the completeness of coverage of the senior survey is clouded by the vague definition of high school senior status, the logistics of locating students who are nominally registered as high school students but do not attend on a regular basis, and the fluid nature of student enrollments. The textbook definition of a high school senior is a student who completed the eleventh grade, is currently enrolled in the twelfth grade, and is likely to graduate from high school at the end of the year. In practice, however, there are considerable variations from this standard definition. Some students consider themselves to be seniors, are taking senior classes, and are listed as seniors in the school yearbook, but are classified in school records as juniors because they have not earned sufficient credits. Along with “fourth-year juniors,” there are a number of “fifth-year seniors,” who did not graduate on time and continued parttime enrollment in high school to earn sufficient credits to graduate. In addition to the problems of identifying the potential universe of seniors, errors of coverage arose because about 10 percent of students were not enrolled in one of the five comprehensive high schools in the school district. These students enrolled at “alternative sites.” They included home-schooled students as well as students enrolled in a range of special programs because of academic, behavioral, or disciplinary problems. Although registered as high school seniors, many of these students had only a nominal affiliation with the regular school system. The largest alternative-site program consisted of students who had dropped out of “regular” school and were enrolled in high school equivalency courses at community colleges. Despite repeated efforts to contact alternative-site students, the response rates to mailed questionnaires were very low. Even among students enrolled in the five comprehensive high schools, there were “non-mainstream” students who were not present in regular senior classes. For example, about 7 percent of students were in special

University of Washington-Beyond High School   69

education classes for part or all of the school day. Depending on the school and the particular program, special education students were much less likely to have been included in the call to “all seniors report to the auditorium to take a survey.” Another 6 percent of all high school seniors were enrolled in a Running Start program, which meant that they were earning college credit as high school students by taking classes at nearby community colleges. Some Running Start students took one or two community college classes, but others took all their classes in community college and had little contact with their nominal high school. Another fundamental problem was the fluidity of student enrollment in high school. The counts of enrolled students reported in administrative records by the schools and school districts—the benchmark used to estimate survey coverage—is not really an accurate measure of students who are regularly physically present in the schools. The count of enrolled high school seniors changes on a near daily basis as students transfer in and out, drop out, or re-enroll after dropping out. Dropping out is generally determined only after weeks or months of chronic absenteeism. Record keeping of student transfers is generally incomplete. Because “senior status” is dependent on the number of course credits completed, every semester some students fall behind their expected grade, while some other students “catch up” by making up course credits. In a separate study, we analyzed five years of school-district enrollment records and found that nearly a quarter of students enrolled in a given year did not return the following year, and that about 15 percent of the school population was composed of students who were not enrolled the previous year.16 The composition of the student body was constantly in flux, making it difficult to accurately assess who was enrolled in a given grade at a specific point in time. To illustrate the problems of survey coverage, in table 3.2 we compare counts of high school seniors from administrative records with participation in the UW-BHS senior survey conducted in the spring of 2000 in District 1. The issues addressed here were common to our survey operations in all the public schools in all years. The first row in the first panel of table 3.2 shows the total number of senior students (1,435) who were reported to have taken a course in the spring of 2000 in school administrative records. For the reasons already discussed, we suspect that many of those listed in school records were “phantom” students that did not regularly attend school. The second panel consists of 490 students who were listed as seniors in other administrative records at some time during the 1999–2000 academic year but were not listed in the final course rosters of the spring term. This category includes a few students who were not in any administrative record but were listed as seniors in their school yearbooks or claimed to be seniors in a completed UW-BHS survey. The sum of these two figures,

Not enrolled in spring  semester    Listed in the yearbook    Not in yearbook

Enrolled in spring  semester    Listed in the yearbook    Not in yearbook

School Records

1,056

925 131

101 41 60

942 146

108

44 64

(2)

Coreb

1,088

(1)

Alla

 3  4

 7

17 15

32

(3)

Non-Corec

UW-BHS Senior-Survey Respondents

162 220

382

203 144

347

(4)

Not Interviewedd

206 284

490

1,145 290

1,435

(5)

Totale

Universe of High School Seniors by Any Reckoning

(6) = (1)/(5)

21 23

22

82 50

20 21

21

81 45

74%

(7) = (2)/(5)

All UW-BHS 76%

Core UW-BHS

Coverage Rate

f

 11  15

 25

  4   5

  9

 80  11

 91%

 75%  59  15

(9)

Of UW-BHS Core (column 2) (8)

Of Total (column 5)

Compositiong

Table 3.2    Coverage of the UW-BHS Survey of High School Seniors by Enrollment Status and Yearbook Listing: Class of 2000 in District 1

1,157

966 191

1,196

986 210

20 19

39 365 364

729 1,351 574

1,925 73 37

62 72 33

60  70  30

100  83  17

100

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample is limited to high school seniors in District 1 in 2000. Seniors enrolled in the spring semester are defined as students who completed one or more courses in the spring semester of 2000 based on school administrative records. Seniors listed in the yearbook are defined as seniors whose names appeared in the high school yearbook. UW-BHS = University of Washington-Beyond High School. aAll interviewed high school seniors (core and non-core sample respondents). bCore sample includes the high school seniors (however defined) who completed a usable survey. cNon-core respondents include foreign-exchange students, juniors, students with severe developmental disabilities, and students who answered the survey with random responses. dPotential eligible respondents (based on administrative records) who were not interviewed. e Senior students reported in any school administrative record. f The percent of total students who completed the baseline senior survey. g Column percentages of high school seniors: column (8) in any school administrative record, column (9) in the UW-BHS core sample.

Total seniors (by any  definition)    Listed in the yearbook    Not in yearbook

72   From High School to College

1,925, represents the number of unique names (actually student IDs) that appeared as high school seniors in various records. Each of these major categories is subdivided into yearbook and non-yearbook seniors (those not listed in the school yearbook). Our count of yearbook seniors also includes all students listed as seniors in the yearbook regardless of whether a picture was published. Yearbook status is subject to similar problems of measurement (transfers and dropouts), but it provides an independent estimate of high school seniors for our evaluation of the completeness of coverage of the UW-BHS senior survey. The first three columns in table 3.2 show participation in the UW-BHS senior survey for the class of 2000. Although 1,196 completed surveys were collected, we excluded thirty-nine questionnaires, leaving 1,157 cases in the “core sample” for our analysis. The excluded cases consisted of exchange students, juniors who mistakenly took the survey, developmentally disabled students, and a few students who appeared to have answered the questions randomly. The columns labeled “coverage rate” show the percent of completed questionnaires for alternative definitions of the population of eligible seniors. For the sample of enrolled spring seniors, 74 percent completed the UW-BHS questionnaire. For the subset of yearbook spring seniors, the coverage rate was 81 percent. Although not perfect, this is a fairly respectable rate. Among “marginal” students, the coverage rates were much less—only 45 percent among non-yearbook spring seniors and only 21 percent of students who were not reported to have completed a course in the spring term. For the broadest definition of senior status (enrolled at any time during the year or listed in the yearbook or completed a UW-BHS survey), there were a total of 1,925 “ever seniors,” of whom 60 percent completed the survey. Which of these figures is “right”? The coverage rate of the UW-BHS survey is best for enrolled spring seniors, especially for those who were also listed in school yearbooks. We suspect that the majority of the students with a marginal connection to high school, for example the 290 non-yearbook seniors and the 480 students who did not complete a spring 2000 course, may not have actually been present in the school when the UW-BHS survey was conducted in April or May of 2000. Some of these students may have graduated at the end of fall semester, transferred to another school district, or dropped out of school. A reasonable course of action, and one that we considered, would be to limit our analysis to the sample of UW-BHS students who were regular students—the 1,056 core sample respondents who were reported by school records to have completed a course in the spring semester—and to discard the completed surveys from the 101 students who were not in the final administrative database for the spring term. Although these students were in the category that we consider

University of Washington-Beyond High School   73

“marginal,” they were present in the school when the survey was conducted (or mailed back a survey sent to their home). The ultimate question is, what is the universe of students whose behaviors we seek to explain? The conventional portrait of high school enrollment, with clear boundaries between ninth-, tenth-, eleventh-, and twelfth-grade classes and continuity over the course of an academic year, is not realistic. Imagine an alternative model using the metaphor of a time-lapsed photograph of senior students based on their attendance during the school year. The same picture of students is taken every day with each student assigned to a unique location. Then at the end of the year, the daily pictures could be superimposed, one upon the other, in a collective representation of student enrollment. Students, who were present every day, or almost every day, would stand out in bold relief and be clearly recognizable. Other students, who were only present for one semester, or for a substantial part the year, would be visible but hazy. The students with irregular attendance or who dropped out would appear only as shadows in the montage of daily photographs. The metaphor of the collective image of student enrollment—with clarity weighted by daily presence—provides a benchmark from which to evaluate the coverage of “enrolled” seniors in the UW-BHS senior survey. The count of 1,925 “ever seniors” is too broad—many of these students transferred, dropped out, or were chronic absentees. Many of these “phantom” students were not part of the life of the school and, therefore, should not be weighted the same as “regular” students who were usually present. On the other hand, the sample of 1,145 yearbook seniors who were enrolled in the spring term (when the UW-BHS survey was conducted) is too narrow. This sample would exclude all “marginal” students who were occasionally present and appeared in some school records. These students, who could only be seen hazily in the montage of daily pictures, were part of the social world of the school. Our “preferred” sample is the 1,157 high school seniors who completed a UW-BHS senior survey (our core sample). This is a sample with implicit weights based on the frequency of presence in the school. Regular students are overrepresented because they are typically present throughout the school year. Marginal students, who appear in some, but not all, school records are not ignored but are underweighted based on the likelihood that they would complete a UW-BHS survey. The implications of this weighting scheme are shown in the last two columns of table 3.2, which compares the composition of the universe of 1,925 “total seniors” (listed as a senior in any record) with the sample of 1,157 UW-BHS core senior respondents (those who completed a survey). The UW-BHS core sample consists of 80 percent regular students (yearbook spring seniors), 11 percent non-yearbook spring seniors,

74   From High School to College Table 3.3    Coverage Rates and Composition of the Total Sample of UW-BHS High School Seniors by Enrollment Status and Yearbook Listing: 2000 to 2005 Enrollment and Yearbook Listing

UW-BHS Core Samplea

Compostionb

Coverage Ratec

Enrolled in spring semester    Listed in the yearbook    Not in yearbook Not enrolled in spring semester    Listed in the yearbook    Not in yearbook Total seniors (by any definition)    Listed in the yearbook    Not in yearbook

9,248 8,326 922 309 143 166 9,658 8,504 1,148

96% 86 10 3 1 2 100 88 12

69% 78 35 17 16 19 61 72 30

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: UW-BHS = University of Washington-Beyond High School. aAll UW-BHS senior-survey respondents, excluding foreign-exchange students, juniors, students with severe developmental disabilities, and students that answered the survey with random responses. bColumn percentages by yearbook listing and enrollment status. cThe ratio of UW-BHS core respondents to the total number of students in school administrative records.

and 9 percent students who did not complete a spring course. This sample does not represent equally everyone who appeared on any school record during the school year. On the other hand, it is more inclusive than if we limited the sample to just “regular” students. We have included data from students in the shadows in proportion to their likelihood of completing the senior survey. This is an arbitrary weighting scheme, but one that is correlated with their frequency of participation in the school. A summary of our analysis of coverage rates for the entire UW-BHS universe of all three school districts for all schools in all years is shown in table 3.3. The school administrative records from District 2, District 3, and the private schools were limited to a list of currently enrolled (for the spring semester) seniors. We also had yearbooks from each school, so we were able to compute coverage rates for students defined by enrollment status and inclusion in the yearbook. The results are very similar to those in table 3.2. Among regular students—those enrolled in the spring semester and listed in the school yearbook—78 percent of seniors completed a UW-BHS senior survey, only slightly less than the 81 percent reported for District 1. Coverage in the UW-BHS was much lower for non-yearbook seniors—only 35 percent and less than one in five students who were not currently enrolled. The composition of the UW-BHS sample is heavily

University of Washington-Beyond High School   75

weighted toward regular students, but about 10 percent consisted of current students who were not in the yearbook and an additional 3 percent of students who were not currently enrolled. To sum up, the core sample of 9,658 respondents includes all high school seniors who were interviewed in the UW-BHS baseline senior survey in nine public high schools in three school districts and in three private high schools in a metropolitan region of the Pacific Northwest. The core sample represents our best effort to contact and interview the universe of all currently enrolled seniors in the targeted high schools. Since the selection of the three public school districts and three private high schools was largely opportunistic, the standard logic of generalizing from a sample to a known universe does not apply. Nonetheless, we follow the rules of interferential statistics as a rough guide for interpretation of substantively significant findings.

The Attrition of Students in High School Before Their Senior Year The objective of the present study is to describe and explain inequality by ascriptive background in the college enrollment and completion process. We focus on higher education because there is much less concern with the completion of high school as a source of educational inequality in the twenty-first century. Upwards of 86 percent of adult Americans complete high school, and the figure is even higher for native-born Americans.17 Significant racial and ethnic disparities in high school completion among native-born minorities remain, but they are much smaller than in prior generations. The interpretation of the almost universal completion of high school is challenged, however, by figure 3.1, which shows the number of students enrolled in eighth through twelfth grades in District 1.18 These enrollment figures are averaged over seven academic years (1997–1998 to 2004–2005) to adjust for annual variations in cohort size. The most striking statistic is that the average number of seniors (1,452) was less than half of the average number of freshmen (3,006). The count of freshman was inflated because of retentions and in-transfers. Also, second-year high school students who did not pass sufficient courses to earn sophomore status were classified as ninth-graders in school records. There was also a modest influx of students who transferred to public high schools in ninth grade because of the sharp increase in private school tuition. However, other alternative indexes, such as the ratio of twelfth-grade enrollment to the average of eighth-, ninth-, and tenth-grade enrollments was still only 57 percent (1,452 divided by 2,570). The implication of a 43-percent dropout rate does not square with national census data, which show a 90-percent graduation rate (see figure 2.1).

76   From High School to College Figure 3.1    Average Annual Number of High School Students Enrolled in School District 1, by Grade Level: 1997–1998 to 2004–2005 3,500 3,006

Number of students

3,000 2,500

2,387

2,316

2,000

1,733 1,452

1,500 1,000 500 0

8th grade

9th grade

10th grade

11th grade

12th grade

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016).

The conventional estimate of a 90-percent graduation rate, based on retrospective reports of educational attainment from census and survey data, is certainly exaggerated. Until recently, national survey data did not distinguish typical high school graduates from alternative certificates of high school completion. Other factors include the overreporting of educational credentials by survey respondents, household surveys being more likely to miss (undercount) persons with lower educational attainment, and the exclusion of the nonhousehold sample in many national surveys. The conclusion of several recent national-level studies is that the “true” rate of on-time high school graduation (from high schools) is closer to 70 percent than the 85 percent from national surveys.19 The problem of high school attrition is linked to the previous discussion on measuring who is an enrolled high school senior. The majority of high school–aged youths who regularly attended school were also listed in high school yearbooks and also likely graduated from high school. Students in this category were also likely to have completed the UW-BHS senior survey if they were enrolled in one of sample high schools. However, there was a significant minority of high school–aged youth, perhaps 25 to 35 percent, who were marginal attenders of high

University of Washington-Beyond High School   77

schools. Some dropped out soon after reaching age sixteen. Others drifted in and out of formal schooling, transferred from school to school, or attended remedial programs in community college or alternative high schools. This “churning” made it difficult for the high schools, especially those serving communities with high proportions of students from dis­ advantaged backgrounds, to keep a firm count of enrolled students. Many, perhaps the majority, of these students who dropped out will, at some later point, return to complete an alternative high school credential.

Comparison of UW-BHS Senior Respondents with Respondents to Other Educational Surveys The limitations of the UW-BHS high school senior sample include: a selective (nonrandom) sample of twelve high schools; a nonresponse rate of 20 percent among “regular” enrolled students; an even higher non­ response rate among students who were marginal attendees; and the limitation to seniors, who may represent only 65 percent to 75 percent of students who began high school. The appropriate benchmark to evaluate the quality of the UW-BHS senior survey (and the one-year follow-up) is not perfection but other comparable surveys of high school students. Our assessment is that the problems of coverage and non­ response experienced by the UW-BHS project are endemic to all surveys of high school students. William Sewell and Robert Hauser reported in 1975 that the baseline senior survey in the Wisconsin Longitudinal Study had a response rate of 94 percent based on the ratio of completed surveys to all high school graduates in the state of Wisconsin in 1957.20 They attribute nonresponses primarily to the failure of 47 (out of 501) high schools to return the questionnaires. However, they also noted that an unknown number of students were absent at the time of the administration of the survey. The actual response rate might have been considerably lower if the denominator of “eligible” was based on enrolled high school seniors rather than high school graduates. In an appendix, Sewell and Hauser used a variety of census and state data sources to estimate the cohort high school graduation rate of Wisconsin in the late 1950s. They concluded that 75 to 80 percent of the Wisconsin students graduated from high school in the late 1950s—a figure that was considerably higher than that for the United States as a whole.21 The Wisconsin Longitudinal Study research on college attendance and graduation was based on the sample of original high school respondents (or the parents of respondents) who were contacted and interviewed seven years later in 1964, with a total response rate of 88 percent of the baseline sample of high school seniors.22 Although nonrespondents to the 1964 Wisconsin Longitudinal Study follow-up survey

78   From High School to College

were negatively selective (lower SES), the authors concluded that any biases from incomplete coverage were relatively small and unlikely to affect research results. The gold standard for high-quality educational data is the longitudinal surveys conducted by the NCES. The Education Longitudinal Study of 2002 (ELS:2002), the most recent NCES survey, included 15,400 tenthgrade students drawn from a sample of 750 high schools. The sample of 750 public and private schools represents a 68-percent (weighted) sample of the 1,220 eligible schools that were initially contacted.23 Each participating high school provided a list of enrolled tenth-grade students in addition to a random sample of twenty-six students selected from the roster of eligible students. If selected students were absent on the survey day, repeated follow-up efforts were made. The sample of 15,400 participating (interviewed) students represents an 87-percent (weighted) response rate of the 17,600 initially selected. Student-questionnaire response rates varied modestly—from 82 percent to 92 percent—by race-ethnicity, region, and urban-rural location.24 In earlier NCES student surveys (the 1980 High School and Beyond Survey and the 1988 National Educational Longitudinal Survey), several categories of students were considered ineligible for participation: foreign-exchange students, students with disabilities, and students with limited command of English. The ELS:2002 sampling design was more inclusive, as it offered special accommodation for students with learning disabilities, and students who had been receiving English instruction for less than three years were deemed eligible if school staff felt they were capable of participating.25 Students who were considered questionnaireineligible were excluded from the response rates. The ELS:2002 study reported that “identified biases due to nonresponse were small and that the data could be used with confidence.”26 The first follow-up of the ELS:2002 baseline sample, conducted two years later, when the students were scheduled to be high school seniors, yielded a weighted response rate of 89 percent. One of the most widely used data sources for educational research on adolescents is the National Longitudinal Study of Adolescent to Adult Health (Add Health), which interviewed students in a large nationally representative sample of 90,118 students enrolled in 132 middle and high schools.27 Most studies using Add Health data have relied on the subset of 20,745 adolescents who were also interviewed at home—the Wave I sample. The Wave I response rate was 79 percent.28 Our conclusion is that most highly regarded surveys of high school students have experienced problems in selecting fully probabilistic samples of all students in the target universe. This is true of national samples as well as regionally based projects. Lack of response from students contacted in schools appears to be less of a problem than the refusal of many

University of Washington-Beyond High School   79

schools to participate in research studies. Students who are not present in school, for whatever reason, are often difficult to contact. Attrition from longitudinal studies, even with the best of procedures, is always a problem. There are no precise metrics for comparing the coverage and completeness of surveys among the many studies of high school students. Despite these problems of obtaining representative surveys, in addition to errors arising from nonresponse (missing data) within surveys, most studies have remarkably similar findings. The UW-BHS sample of students from selected schools in the Pacific Northwest was clearly not representative of all students in the country. The key issue here is not whether national population parameters, such as the proportion of high school seniors who go on to attend college, can be generalized from a geographically defined sample—clearly they cannot. The more relevant question is whether findings about relationships and processes, such as the effect of family status on the likelihood of college attendance, are similar to those reported by other data sources, including national surveys. Given the similarity of findings from this study with those from regional and national data, we think that measures of relationships between variables in the UW-BHS data (the aim of our research) hold considerable relevance for understanding educational stratification among American students. Another difficulty was the underrepresentation of students with marginal attachments to (and limited presence in) high schools in the UW-BHS senior survey. The implication is that educational disparities, as measured in the UW-BHS follow-up survey, are likely to be smaller than figures based on data from all youths of a comparable age. This problem is generic in all research because data are collected from populations that are more easily found, observed, and interviewed. Readers should assume that all measures of inequality reported in this study (and most all research) are underestimates of true population differences—less than would be found in population-representative data.

Measurement of the Dependent Variables of the College Pathways Model Our primary focus is on the transition from high school to college, which we have defined as a series of steps beginning with the formation of college aspirations and ending with the completion of a baccalaureate degree from a four-year college or university. A detailed empirical account of the entire process—the College Pathways Model—will be presented in chapter 4; here we focus on the measurement of the key educational outcomes and the logic underlying the creation of these measures.

80   From High School to College

College Aspirations and Expectations The first step in the process of completing college is the formation of college ambitions—the desire and planning to enter and complete higher education after graduating high school. Desires must be acted on in order to reach a goal, and we posit a cumulative sequence of actions necessary to translate college ambitions into college graduation: • college aspirations • college expectations • college preparation • college enrollment • college completion To assess educational ambitions, we used two similar yet conceptually distinct measures: college aspirations and college expectations. The measures were based on the following questions in the UW-BHS senior survey: “How far would you like to go in school?” and “Realistically speaking, how far do you think you will get in school?” The seven response categories ranged from “less than high school graduation” to “PhD, MD, or other professional degree.” Both questions tapped desires and plans for higher education. The first question was about aspirations and is very abstract; it simply asked the student how far he or she would like to go regardless of circumstances. By adding the word realistically, the expectations question asked students to take into account potential constraints that could hinder their ability to achieve their desired level of educational attainment. Prior research has shown that many students report lower educational expectations than their unconstrained educational aspirations.29 The distribution of responses to these questions, with the original cate­ gories, is shown in appendix table 3.A1 (available online).30 In general, college expectations of American adolescents were high, and aspirations were even higher.31 We have recoded the detailed responses regarding educational aspirations and expectations to be binary variables that capture the important distinction of college graduation. Specifically, responses up to and including “some college” are coded as “0,” while those including a bachelor’s degree or higher are coded as “1.” Research has shown that having a college degree, not just college attendance, is the most significant predictor of occupational and income attainment in American society.32 Three-fourths of all students we surveyed aspired to obtain a college degree, while 68 percent expected to do so. The difference of 7.6 percentage points (75.5 minus 67.9) between aspirations and expectations

University of Washington-Beyond High School   81

indicated the dampening effect of the phrase “realistically speaking.” Females had higher college aspirations and expectations than their male peers, and we found similar gender gaps in nearly all other measures of educational ambitions and achievement. Excluding the roughly 8 percent of students who skipped these questions, there were 8,860 and 8,844 cases with valid responses to the aspirations and expectations questions, respectively. The relatively high nonresponse rates to these questions (compared to other survey questions) might have been due to their placement as the initial items in the questionnaire booklet. The first survey questions followed an instructions page and a blank page. It appears that some students inadvertently turned the page not realizing that they had skipped the first page of the questionnaire.

College Preparation The next step in the college pathway is preparation, or becoming “college ready,” for a four-year college during the senior year of high school. College requirements vary, but they generally include the completion of a minimum number of high school courses in English, mathematics, science, and other subjects (sometimes with a minimum grade). Our objective in creating the college preparedness variable was to find indicators that would be highly correlated with enrollment in college preparatory courses in high school. We identified three items in the UW-BHS questionnaire: (1) whether the student had taken (or planned to take) a college entrance exam, such as the SAT or ACT; (2) whether the student had taken (or planned to take) an advanced placement (AP) exam; and (3) whether a student had applied to a four-year college by April or May of their senior year (the approximate time when the UW-BHS senior survey was administered). Unlike college aspirations and expectations, which are purely subjective orientations, college preparedness is measured by specific actions that the student has taken. The first two items concerning college preparedness were directly addressed in questions in the UW-BHS senior survey. The third item was measured indirectly. Students were asked to write in the name and location (city, state) of the college or colleges that they were most likely to attend. These colleges were coded as two-year or four-year based on the Carnegie Classification of Institutions of Higher Education.33 The Carnegie Classification provides a systematic and detailed classification of nearly every postsecondary institution in the United States along fifteen dimensions. The measurement was based on whether the student listed at least one four-year college and reported that he or she had applied to it. The responses to these items are shown in appendix table 3.A2 (available online).34 While there was a moderate degree of missing data for

82   From High School to College

the first two items (3.9 percent and 11.6 percent, respectively), there was a much higher level of nonresponse to the college-application question (almost 21 percent), primarily because students did not write in the name of a specific college. For the measurement of college preparation, we only included confirmed “yes” answers as indicators of college preparedness. Missing data were coded as “not college prepared.” Only one of three high school seniors reported that they had taken (or planned to take) an AP exam, but three of five had taken the SAT or ACT. AP credits are generally considered an indicator that a student is preparing to attend college. However, not every high school offers AP classes, and some college-bound students may decide not to take them— perhaps because of other commitments, such as extracurricular activities and work. A significantly higher percentage of high school seniors take a college entrance exam. Most two-year colleges do not require submission of SAT or ACT scores, but college entrance exams are generally considered a “must” for those planning to attend traditional four-year colleges. About 57 percent of UW-BHS seniors reported that they had applied to a fouryear college by the spring of their senior year. These three activities tap somewhat different behaviors but are all highly correlated with each other and, presumably, to college preparedness in high school. The second panel in appendix table 3.A2 shows the distribution of UW-BHS seniors by sum of the three behavioral indicators of college preparation. Only one in four high school seniors (23.7%) was college ready on all three indicators, but more than half (54.7%) had completed two of the tasks. For this study, we coded students with a score of two or three of the three indicators of college readiness as college prepared (value = 1) and those with less as not college prepared (value = 0).

College Enrollment and Graduation The UW-BHS project was designed as a longitudinal study with follow-up surveys to measure college enrollment, graduation, and other early lifecourse outcomes. As reported earlier, a one-year follow up survey was conducted in the year following high school graduation for each cohort of UW-BHS seniors from 2000 to 2005. The follow-up surveys were remarkably successful, and we were able to obtain responses from 92 percent (8,885) of the complete UW-BHS universe of 9,658 high school seniors. Our initial plan was to use data from our follow-up to measure college enrollment. Midway through the UW-BHS project, we learned of the National Student Clearinghouse (NSC). The NSC is a nonprofit organization that collects individual-level student enrollment and graduation records on more than 95 percent of all students in higher education from more than 3,300 colleges and universities in the United States.35 The NSC provides

University of Washington-Beyond High School   83

enrollment verification information for institutions of higher education, employers, and lending institutions. It also supports research on higher education by matching students with college enrollment records, following the federal guidelines that protect student confidentiality. All students enrolled in a participating university or college are included in the NSC database unless they specifically request that their enrollment information not be shared. After the conclusion of the UW-BHS fieldwork in 2012, we contracted with the NSC to obtain matched college enrollment information on the universe of 2000 to 2005 UW-BHS high school seniors who were enrolled in a college at any time from 2005 to 2012. After obtaining necessary approvals for the protection of respondent confidentiality, we obtained college enrollment and graduation records for the universe of UW-BHS students. UW-BHS high school seniors who were not matched with any NSC college enrollment record were assumed not to have been enrolled in college. Taking appropriate safeguards to protect confidentiality, we attempted to match NSC records with individual students from the UW-BHS database of student names and birthdates (we did not have social security numbers). The process of record matching was much more complicated than we initially assumed. Some students had duplicate names and birthdates. A more widespread problem was the uncertainty of matching students with similar but not exact names. Some students changed their names, but more often, they simply wrote variants of their names in different records. Some students may have used middle names or nicknames rather than their first names. Names, both surnames and given names, may also have varied across records because of misspellings or errors in coding. For a small number of cases, we had to rely on additional information from the senior survey to make informed choices about possible matches between UW-BHS records and NSC data. The first question was whether NSC records provided more comprehensive (and more accurate) data on college enrollment than our own one-year follow-up surveys. To address this question, table 3.4 compares estimates of college enrollment of UW-BHS seniors based on the followup survey and NSC administrative records. The timing of the comparison is not exact. The college enrollment question in the follow-up survey referred to the period since high school graduation, but given that the follow-up surveys were conducted from January to June of the year after graduation, some respondents may have answered in terms of current college enrollment. The NSC measure was based on college enrollment anytime during the thirteen-month period following high school graduation (from June to July of the following year). Both the UW-BHS and the NSC data of college names were coded as two- or four-year institutions in accordance

84   From High School to College Table 3.4    Comparison of College Enrollment One Year After High School Between UW-BHS Follow-Up Survey and Administrative Records from the National Student Clearinghouse (NSC) Row Totals NSC Administrative Records UW-BHS Follow-Up Survey Not enrolled Enrolled in two-year  college Enrolled in four-year  college Total

Not Matched 88.9% 20.1

Enrolled in Two-Year College 8.7% 77.4

Enrolled in Four-Year College 2.4% 2.5

19.7

1.3

78.9

100.0

26.7

33.3

100.0

40.0

Total 100.0% 100.0

Column Totals NSC Administrative Records UW-BHS Follow-Up Survey Not enrolled Enrolled in two-year  college Enrolled in four-year  college Total

Not Matched 64.8% 15.3

Enrolled in Two-Year College 9.5% 88.4

Enrolled in Four-Year College 2.1% 2.3

19.8

2.0

95.6

40.3

100.0

100.0

100.0

100.0

Total 29.2% 30.5

Number of Observations NSC Administrative Records UW-BHS Follow-Up Survey Not enrolled Enrolled in two-year  college Enrolled in four-year  college Total

Not Matched 2,306 546

Enrolled in Two-Year College 226 2,098

Enrolled in Four-Year College 62 68

Total 2,594 2,712

706

48

2,825

3,579

3,558

2,372

2,955

8,885

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample is limited to the 8,885 students in the one-year follow-up survey. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

University of Washington-Beyond High School   85

with the Carnegie Classification. NRC records were searched for all 9,652 UW-BHS seniors in the core sample, and nonmatches were con­ sidered not college enrolled. The universe in table 3.4 is limited to the 8,885 UW-BHS students who responded to the one-year follow-up survey. There were three possible responses for each data source: not enrolled (nonmatched in the NSC data), enrolled in a two-year college, or enrolled in a four-year college. The top panel of table 3.4 evaluates the NSC data relative to the follow-up survey data as the standard (row percentages), while the middle panel does the reverse with the NSC as the standard and column percentages. The lower panel shows the actual number of observations (UW-BHS respondents) in each cell. Although the data from the follow-up survey and the NSC records month are not perfectly aligned, they are reasonably close. The total number of cases on the diagonal (7,229 = 2,306 + 2,098 + 2,825), or 81 percent of the matched sample, represents exact agreement on college enrollment status one year after high school. There were a handful of cases (116 = 48 + 68) that were in agreement that the student was enrolled, but one source indicated a two-year institution and the other indicated a four-year college. A slightly larger number of students responding to the follow-up survey indicated not being enrolled, but the NSC reported their enrollment in a two-year (226 students) or a four-year college (62 students). A closer look at the month of enrollment in the NSC indicates that most of these students were enrolled in college during the fall term but then dropped out before they were interviewed in the follow-up survey. Although one can argue that a few months or even a few weeks in college should not be classified as college enrollment, we conclude that the NSC data are more valid. Students who do enroll in college, even for a short time, have made the transition from high school to college. A more serious inconsistency is found in the large number of students who were not enrolled in college according to the NSC but who reported their enrollment in a two-year or four-year college in the UW-BHS follow-up survey (Ns equal 546 and 706, respectively). This discrepancy had an impact on the marginal distributions of college enrollment. The follow-up survey indicated that 40 percent (3,579) of UW-BHS respondents were enrolled in a four-year college and another 30 percent in a two-year college. The NSC data, however, counted only 33 percent (2,955) in a four-year college and 26 percent in a two-year college. The middle panel of table 3.4 shows that 20 percent of students who were enrolled in college according to the follow-up survey were not matched with a college record in the NSC records. It is possible that some of these students exaggerated their college attendance—reporting what they think they should be doing or will be doing in the future—but it is unlikely that such errors of exaggeration account for more than a small share of the

86   From High School to College

discrepant cases. We tend to believe that the UW-BHS follow-up survey respondents accurately reported their college enrollment status. Our finding of underestimation of college enrollment in NSC records has been confirmed by other researchers.36 The NSC acknowledges that its listings are incomplete—the 3,300 NSC participating colleges and universities cover only 96 percent of all college students. More importantly, NSC records are incomplete even among reporting colleges. For example, students can opt out of allowing their colleges to share student records with third parties, including the NSC. Another possible contributor to undercoverage of college enrollment by the NSC is name mismatches. As noted earlier, college student records are often different from those completed in high school. Some students add (or drop) nicknames, initials, spellings, or even names—for example, Barry decides to be called Barack. Change in surname following marriage also complicates name matches; for example, a student known as Susie Smith in high school records could be recorded as Susan S. Jones in NSC records. Many Asian adolescents adopt (or drop) Western-given names, following the longstanding American tradition of changing unfamiliar ethnic names to more familiar Americanized names. In additional analyses, we examined the correlates of the undermatching of college students from the UW-BHS follow-up survey with NSC data. We found that Asians and low-SES students were more likely to be misclassified as not college enrolled in the NSC data when they were enrolled in the UW-BHS follow-up survey. The problem of name matching may be partially responsible for the undermatching of Asian students. Other problems of inconsistent information on names and birthdates may be partially responsible for the lower matching of low-SES students. After weighing all the pros and cons, we decided to rely primarily (but not exclusively) on the NSC measure of college enrollment. A key consideration was that the measure of college graduation was only available from NSC records. Another factor in our decision was the problem of missing data from the UW-BHS follow-up survey. Although less than 10 percent of UW-BHS seniors (8,885 of 9,652) were not successfully reinterviewed in the follow-up, nonresponse was almost certainly overrepresentative of students who did not enroll in college. The nonrespondents in the follow-up survey appeared to be students who maintained little contact with family or friends after high school. We suspect that relatively few of these cases “lost to follow-up” were enrolled in college. The NRCbased estimate of four-year-college enrollment—33 percent of UW-BHS respondents (2,955/8,885)—is biased downwards but the estimate based on the UW-BHS follow-up survey of 40 percent of UW-BHS respondents (3,579/8,885) is probably biased upwards. In several analytical chapters in this book, we compare the rate of enrollment in a four-year college

University of Washington-Beyond High School   87

based on NSC data alone with another estimate based on whether the student was enrolled in college according to either the NSC or UW-BHS follow-up survey data. The final step in the sequence is college completion—the likelihood of obtaining a BA or BS degree from a four-year college or university. Appendix table 3.A3 (available online) shows the percent of all UW-BHS seniors who enrolled in and graduated from college who were matched with NSC records.37 There are slight differences in the reported levels of college enrollment in table 3.4, limited to students who completed the one-year follow-up survey, and appendix table 3.A3, which is based on the complete (core) sample of 9,652 UW-BHS seniors. For example, the matched NRC records for the complete sample of 9,652 UW-BHS high school seniors shows that 31.4 percent were enrolled in a four-year college (appendix table 3.A3) compared to 33.3 percent (table 3.4) for the sample of 8,885 UW-BHS high school seniors who were reinterviewed in the follow-up survey. This small difference confirms our hunch that most of the students lost to follow-up did not go to college. The “traditional” expectation is that students who enroll in college will graduate with a baccalaureate degree four years later. However, the assumptions of continuous enrollment and college completion have always been exaggerated. Many students attend college part-time, often while working, which delays graduation for one or more years. Many students drop out of college, sometimes involuntarily based on their performance. But far more students leave college due to a lack of interest, a change in career plans, demands of employment, marriage or migration decisions, or other reasons.38 Students also transfer from one institution to another, which often adds time to completion. Four years after high school graduation, only 16 percent of UW-BHS seniors had graduated from a four-year college, compared to the 31 percent who enrolled in a four-year college (appendix table 3.A3). With the passage of time, the fraction of college graduates (of UW-BHS seniors) rose to 24.6 percent after five years, 27.8 percent after six years, and 29.3 percent after seven years.

Independent Variables: Ascriptive Characteristics Our primary aim is to explain differences in educational outcomes for three ascriptive characteristics: gender, immigrant generation, and raceethnicity. The measurement of gender and immigrant generation is relatively straightforward. However, the discussion of the concept and measurement of race and ethnicity could easily fill another volume. Here, we provide an overview of conceptual and measurement issues, and our strategy to assign a single race-ethnicity identity to each UW-BHS

88   From High School to College

respondent. The UW-BHS senior survey included separate questions on race, Hispanic origin, and ancestry, following the standard questions of the Census Bureau. The major problem, shared by every data-collection effort that uses the standard census questions, is how to deal with respondents who report multiple racial identities or write in “noncodable” responses, such as a religion or just “American.” We address this issue by creating a set of mutually exclusive racial-ethnic categories based on additional data from our survey. The key additional measure is a survey question on “primary race and ethnicity,” which is the name of the revised classification.

Primary Race and Ethnicity Racial and ethnic origins are highly correlated with educational outcomes in American society. Some historically disadvantaged groups, including the children of southern and eastern European immigrants, have been able to close the educational gap with the majority population.39 African Americans have made progress in completing high school in recent decades, but they lag behind whites in college graduation rates. This is also true of other disadvantaged minority groups, such as Hispanics, American Indians, and Pacific Islanders.40 Historically, race was measured in surveys by asking respondents to choose only one category. Following the lead of the 2000 census, the UW-BHS survey asked respondents to choose “one or more races” that they considered themselves to be. This seemingly small change in wording in the standard census race question opened the door to multiraciality in official statistics. The new policy, mandated by the 1997 revisions to Statistical Policy Directive 15 by the Office of Management and Budget, was intended to reflect changes in American society, including increasing levels of intermarriage and mixed ancestry.41 Although the change in data collection and racial classification represents an important and progressive step in recognizing that many Americans have multiple identities, it does complicate our analysis. The basic problem is there has not yet emerged a consensus on how to classify multiracial persons for statistical tabulations and social-science data analysis. The current Census Bureau convention is to classify persons by race in two ways: by “race alone” (for single-race identifiers) and by “race in combination” (for persons who report multiple races). This practice creates some confusion because of the overlap of racial groups. For example, the sum of the numbers of “blacks alone” and “blacks in combination” yields the total number who identify as black. However, the sum of all whites (“whites alone” and “whites in combination”), all blacks (“blacks alone” and “blacks in combination”), and so on yields a number that exceeds the total population of respondents because multiracial persons

University of Washington-Beyond High School   89

are double counted or triple counted (if they report three identities). The alternative is simply to consider each racial group as consisting of singleidentifiers, but this leaves the awkward residual of the multiracial population. Efforts to classify the multiracial population by its constituent groups lead to the creation of dozens of additional categories (the sum of all possible combinations). For some populations, especially American Indians, the multiracials (persons who are part American Indian) are as large as the monoracials (“American Indians alone”). Another complicating aspect of measuring race (and ethnicity) is that “Hispanic origin” is a separate classification in Census Bureau practice. Although many people, researchers included, considered Hispanicity to be on a par with racial identities, the inclusion of “Hispanic origin” as a separate classification allows Hispanics to be members of any race. However, about half of all Hispanics consider their ethnic identity to be equivalent to their racial identity, and they skip the race question or check the “Other Race” category and write in a specific Latin American national origin, such as Mexican, Cuban, or Puerto Rican. For this project, our aim was to create a single identity—labeled primary race-ethnicity—for each UW-BHS respondent. This classification combines responses from the race and Hispanic-origin questions with explicit assumptions on classifying persons who gave multiple, inconsistent, or no responses. In general, our methods attempted to allow the respondents to speak for themselves by relying on additional questions in the UW-BHS senior survey. Based on the UW-BHS senior survey, we created several alternative race-ethnicity classifications, which are shown in table 3.5. The list of race-ethnicity groups is inclusive of all the major race categories and Hispanic-origin categories that were listed on the original questionnaires. In addition to the listed groups, the questionnaire also included spaces for respondents to write in responses under American Indian, Other Asian, Other Pacific Islander, and Some Other Race on the race question, and under Other Hispanic on the Hispanic-origin question. Many of the writein responses simply provided additional information that was consistent with the major category. For example, some students wrote in “Thai” on the Other Asian write-in line or “Navajo” on the American Indian write-in line. These cases were left unchanged. However, there were also a number of inconsistencies between specific write-ins and the broader categories. For example, some students wrote in a non-Indian identity on the write-in line for American Indian tribes (for example, Irish) or a non-Asian identity on the write-in line for Asian (for example, Hawaiian). Our coding practice was to “reassign” these responses to be consistent with the specific write-in response. For example, if a person wrote “Italian” on any write-in line, that person was classified as white. If a respondent wrote “Japanese,” she or he was coded as Asian, regardless of the location of the write-in response. Given that we consider Hispanicity to be equivalent to a racial

Non-Hispanic White African American American Indian Asian American   East Asian     Korean     Chinese     Japanese   Filipino   Vietnamese   Cambodian   Other Asian Pacific Islander   Samoan    Other Pacific Islander

6,300 1,387 533 2,046 842 446 219 177 369 322 292 221 294 124 170

(N) (1) 54.3% 11.9 4.6 17.6 7.3 3.8 1.9 1.5 3.2 2.8 2.5 1.9 2.5 1.1 1.5

Identities (2) 65.3% 14.4 5.5 21.2 8.7 4.6 2.3 1.8 3.8 3.3 3.0 2.3 3.0 1.3 1.8

Persons (3)

Percent of

Major and Minor Identities: Alone and in Combination

51.5% 8.4 0.6 11.1 3.6 2.8 0.5 0.3 1.8 2.8 1.9 1.0 1.1 0.7 0.4

Percent (5)

Persons

4,972 806 61 1,069 347 266 52 29 175 267 188 92 106 68 38

(N) (4)

Single Race-Ethnic Individuals and Other Responses

Race-Ethnicity Responses

60.6% 13.7 1.5 16.0 5.6 3.8 0.9 0.9 2.8 2.9 2.8 1.9 2.0 0.9 1.0

Percent (7)

Persons

5,846 1,325 148 1,546 543 370 85 88 269 283 269 182 190 91 99

(N) (6)

Primary Race-Ethnicity (adjusted)

Table 3.5     Alternative Measures of Race and Ethnicity Among UW-BHS High School Seniors

702 406 77 434 194 104 32 58 92 16 81 51 83 23 60

UW-BHS Questions (8)

172 113 10 43 0 0 0 0 0 0 0 43 1 0 1

School Records (9)

Multiracial, Uncodable, and No Response Assigned By

— — 11,607

Other responses Multiple identities No response and uncodable Total 100

9.0 3.8 1.2 4.1

120

10.8 4.5 1.4 4.9 2,059 397 9,652

182 129 19 34 21.3 4.1 100

1.9 1.3 0.2 0.4

28 9,652

569 339 71 159

0.3 100

5.9 3.5 0.7 1.6 104 285 2,456

365 208 52 105 0 28 389

22 0 0 22

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: Column 1 is based on the tally of responses to the race (Q159) and Hispanic origins (Q158) questions in the UW-BHS senior survey. This is a tally of racial and ethnic identities, not persons. Column 2 shows the percentage of identities noted for each race-ethnicity group relative to the total number of listed identities. Column 3 shows the percentage of identities noted for each race-ethnicity group relative to the number of survey respondents. Column 4 is the number of single racial-ethnicity responses to Q158 and Q159. All students with more than one response were coded as “multiple identities,” and students with no response (left questions blank) and uncodable answers were coded “no response and uncodable.” Column 5 shows the percentage distribution of single racial-ethnicity responses based on column 4. Column 6 is the number of students reclassified to “primary race-ethnicity.” “Multiple identities” and “no response and uncodable” responses were reclassified based on responses to additional questions in the UW-BHS senior survey and school district administrative records. Column 7 shows the percentage distribution by “primary race-ethnicity” based on column 6. Column 8 includes respondents who were originally classified as “multiple identities” and “no response and uncodable” but were reclassified to a single race-ethnicity on the basis of additional questions in the UW-BHS senior survey. Column 9 includes respondents who were originally classified as “multiple identities” and “no response and uncodable” but were reclassified to a single race-ethnicity on the basis of the school district administrative data.

1,047 438 135 474

Hispanic Mexican Puerto Rican Other Hispanic

92   From High School to College

identity, a person who checked “white” and also wrote in “Mexican” under Other Race, was considered to be multiracial. Our logic is more fully explained in the following paragraphs. In table 3.5, we trace the steps from the complexity of the original responses to the race-ethnicity questions to our preferred summary classification of a set of mutually exclusive categories for primary race-ethnicity. There are several pan–race-ethnicity groups along with constituent subgroups. For example, Asian American is subdivided into East Asian, Filipino, Vietnamese, Cambodian, and Other Asian. East Asian is further subdivided into Korean, Chinese, and Japanese. The panethnic Pacific Islander (Native Hawaiian or Pacific Islander, or NHOPI) population includes two groups: Samoan and Other Pacific Islander. The Hispanicorigin population is subdivided into Mexican, Puerto Rican, and Other Hispanic categories. Column 1 of table 3.5 includes counts of all persons who “ever identify” with each category—“alone or in combination,” which means that some persons are double (or triple) counted. All single-race identifiers (“race alone” in census parlance) are counted once in column 1, while multiracial students are counted multiple times—once for each race or ethnicity checked (or written in). Respondents who checked both a specific Hispanic group and a racial group are counted as having both a Hispanic identity and a distinct racial identity (for example, Mexican and white), while only the racial identity (or identities) for non-Hispanic students were counted. For example, if a person checked white, Filipino, and Other Hispanic, he or she would be counted as having three identities in column 1. Students who checked black and white would be counted as having two identities in column 1. Respondents who skipped these questions or wrote in something odd, such as “Martian,” are not counted as having an identity (at least one of the standard census identities) in column 1. In their responses to the race and Hispanic-origin questions, the 9,652 UW-BHS respondents reported 11,106 identities, alone or in combination, which are presented as percentages in column 2 (relative to total identities) and in column 3 (relative to all persons). The sum of column 3 is 120 percent, which simply means that there are more identities than persons. In contrast to total (“alone” and “in combination”) counts of everidentifiers in column 1, the counts in column 4 only include “single raceethnicity” (“alone” in census parlance) persons for each group. Persons who checked multiple races (or a race and a Hispanic identity) were assigned to the residual category of “multiple identities” in column 4. The count of monoracial (or monoethnic) Hispanics in column 4 is consistent with the logic of single-identifier race groups, but it requires a bit more explanation. The count of single-identity Hispanics in column 4 includes

University of Washington-Beyond High School   93

persons who checked “yes” to the Hispanic-origin question and either skipped the race question or wrote in a Hispanic response (for example, Mexican or Latino) to the race question. If persons checked “yes” to the Hispanic-origin question and checked one of the listed races (or wrote in a non-Hispanic “some other race”), they were classified as having multiple identities. There is a final residual category under column 4—“no response and uncodable”—for respondents who either skipped both the race and Hispanic-origin questions or wrote in an uncodable response, such as a religion, a region, or generic response (“American”). The comparison of the numbers of students who ever-identify (column 1) and only-identify with a specific group (column 4) illustrates the fluidity of race and ethnicity in American society. For example, almost twothirds (65 percent) of UW-BHS high school seniors claimed to be white or part white, but only one-half (51 percent) reported themselves as only white. Since there are more identities than persons (the sum of column 3 is 120 percent), the counts of single identifiers in column 4 are always smaller than the total identities of the same group in column 1, but some of the discrepancies are huge. For example, more than 5 percent of students (N = 533) reported having some American Indian ancestry, but less than 1 percent (N = 61) reported being “only American Indian.” Over 10 percent of UW-BHS students checked “yes” on the Hispanic-origin question, but only 2 percent reported that they were Hispanic-only in column 4. The remainder of Hispanics were persons who checked “yes” on the Hispanic-origin question but also checked one of the listed groups on the race question, where Hispanic was not a listed category. So, who really are American Indians, Hispanics, or any of the other groups in table 3.5? The problem with the broad definition of raceethnicity in column 1 is the double counting of multiracial students— many students identify with two or more racial-ethnic groups. The problem with the narrow definition of race-ethnicity in column 4 is that over 21 percent of students are classified in the residual “multiracial” category, and another 4 percent are in the “no response and uncodable” category. We propose an alternative classification of primary race-ethnicity. The UW-BHS senior survey contained several additional questions about identities beyond the standard census questions of race and Hispanic origin. The most important was question 161: “Considering all the race and ethnic categories, what is your primary ethnic and/or racial identity?” The objective was to allow respondents to report their preferred single racial or ethnic identity if they had to choose only one. A common practice by institutions and researchers is to assume that multiracials should be classified by their minority status even if they are part white. For example, survey respondents who check “yes” to Hispanic origin are generally considered Hispanic regardless of their responses to the race

94   From High School to College

question (as we did in chapter 2). However, there is much less consensus on how to classify persons who report multiple minority identities. Overall, the question on primary identity was understood by UW-BHS respondents, and the overwhelming majority of multiracials responded with a single identity. However, not all cases of multiple identities and nonresponses were resolved with this question. Some students skipped it, while others wrote in multiple identities (for example, “mixed black and white” or “white and Native American”). To resolve these ambiguous cases, we developed a few ground rules. If the respondent wrote in multiple identities, we selected the first write-in response. If the respondent skipped the primary-race question, we used the responses to the questions on ancestry (census-style questions on ancestry were asked for the respondent, the respondent’s birth mother, respondent’s birth father). The first listed ancestry was given priority if multiple ancestries were listed. Column 8 in table 3.5 shows the absolute numbers of UW-BHS respondents with complex or problematic responses to the race and Hispanicorigin questions (coded as multiple identities, no response, or uncodable in column 4), which were recoded to a specific group by the question on primary race-ethnicity. Almost all persons with multiple identities (1,955 out of 2,059) and little over a quarter of those with missing or uncodable responses (112 out of 397) were classified to a single racial-ethnic group by using the question on primary race-ethnicity or related items on the UW-BHS survey. The remaining problematic responses consisted primarily of persons who skipped the primary race-ethnicity and ancestry questions after having skipped the Hispanic-origin and race questions, thus providing no information whatsoever. Fortunately, we were able to access one additional source of information independent of the UW-BHS senior survey—the administrative records from the high school. In most cases, we received the names of seniors from each participating school along with basic demographic information, including date of birth, gender, and race-ethnicity. The school classification was a set of mutually exclusive categories: white, black, American Indian, Asian, Pacific Islander, and Hispanic. Using these records, we were able to assign a generic race-ethnicity to all but 28 of the 389 students (see column 9) who still had multiple racialethnic identities or a missing race-ethnicity after the adjustment based on responses to the questions about primary race-ethnicity and ancestry on the senior survey (column 8). Note that the school data do not allow for assignment to specific Asian or Hispanic groups but only to the panethnic category. Drawing on both the additional survey questions about primary raceethnicity and school administrative records, column 6 presents our final “preferred” counts of all students by a single race-ethnicity group. This measure is labeled primary race-ethnicity because the overwhelming

University of Washington-Beyond High School   95

majority of multiracial students are assigned to a singular race or ethnic group based on the questions on primary race-ethnicity. To avoid the deletion of the twenty-eight cases with an unknown race-ethnicity, these students are assigned to the white population in subsequent tabulations. While the primary race-ethnicity classification (column 6) does not resolve all problems, it does create a relatively parsimonious set of mutually exclusive categories that largely reflect respondent’s preferred identities. While we are sympathetic to the claim that multiracial or multiethnic persons should not be classified into a single category, we believe that other methods of assignment are even more problematic. For example, leaving 25 percent of respondents in global “multiracial” and “unknown” categories would be very confusing. It is also counterproductive to create dozens of new categories that contain small numbers of specific racialethnic combinations (for example, Chinese-Vietnamese, black–Native American). Our goal is to measure and explain racial and ethnic dis­ parities, and this objective requires a finite, or at least a relatively parsimonious, classification of racial and ethnic groups. In the long run, we think that a new conceptualization of racial-ethnic identities that allows for variations between single-identifiers and partial-identifiers would be an important contribution to research and popular understanding, but it is a task that would require a separate project. For our purposes, we will rely on the primary race-ethnicity classification of UW-BHS respondents, as presented in column 6 of table 3.5. The racial and ethnic composition of the UW-BHS sample is reflective of the ethnographic makeup of the Pacific Northwest region among adolescents of 60 percent white, 15 percent black, 15 percent Asian American, and smaller proportions (2 to 4 percent) of Mexicans, other Hispanics, Pacific Islanders, and American Indians. Each of these populations is incredibly diverse, with subgroups of Asians identified in table 3.5 as East Asians (5 to 6 percent), Cambodians (3 percent), Vietnamese (3 percent), Filipinos (3 percent), and Other Asians (2 percent). One of our main objectives is a close examination of ethnic diversity, especially among the hetero­ geneous Asian American panethnic population.

Gender Appendix table 3.A4 (available online) shows the primary race-ethnicity classification by gender.42 Gender (or sex) is an ascriptive characteristic that is correlated with educational outcomes. The historical gender differential in higher education that favored men was reversed in the 1970s and, at present, young women are more likely to graduate from high school, enroll in college, and graduate from college than their male peers.43 Indeed, women are superior to men in every level of schooling and in both attitudes and behavior.44 The assumption that gender is

96   From High School to College

a binary trait is an oversimplification that masks considerable overlap in identities and sexual orientation. This complexity is avoided here with a survey question that asked “What is your sex?” and had the traditional response categories of female and male. For the handful of survey nonresponses, administrative data from school records were used to assign codes for gender. Females outnumbered males (55 to 45 percent) in the UW-BHS sample of high school seniors. The overrepresentation of females in the UW-BHS sample was indicative of the gender imbalance in the later years of high school due to sex-selective high school dropout.45 This gender imbalance was found in every racial-ethnic group but was most acute among American Indians and Mexican Americans (61 percent and 58 percent female, respectively). There was near gender parity among Asian Americans—52 to 48 percent. The ethnic composition of males and females differed a bit because of variations in the underrepresentation of males among high school seniors. The male sample tended to have larger proportions of whites and Asian Americans due to the higher dropout rates among minority males.

Immigrant Generation The assimilation or incorporation of immigrants and their children in the American social fabric is a classical sociological theme. Immigrants—the first generation—are unique because of their birth and socialization in another society. In general, most immigrants arrive in the United States with lower levels of human capital and occupational skills than the native-born population, though some immigrant streams (from India, Taiwan, Iran, and Nigeria) are highly selective and have high levels of educational attainment.46 As newcomers, immigrants often experience short-term adjustment problems as they develop new linguistic skills and learn the rules and norms of a new society. In spite of these problems, many immigrants adapt quickly and some are remarkably successful. These successes are often attributed to perseverance and extra effort that comes with high ambitions to succeed in a new society. Though immigrants experience a combination of advantages and disadvantages, their children—the second generation—are generally upwardly mobile. The children of immigrants may be disadvantaged by the lower SES of their families, but they experience fewer problems of acculturation. Regardless of the language spoken at home, children who begin their formal schooling in the United States invariably become fluent in English, indeed most prefer to speak English instead of their mother tongue.47 The high ambitions of immigrant families are generally absorbed by the second generation—a phenomenon that sociologists have named “immigrant optimism.”48 The first generation is identified by

University of Washington-Beyond High School   97

birthplace and the second generation by parental birthplace. The thirdand-higher generation is often labeled third generation, but it is really a residual category that includes the grandchildren of immigrants with others whose ancestral origins are much more distant. In recent years, there has been considerable interest in the educational progress of immigrant youth. Most studies have found that the secondgeneration youth excel in the educational system, while the first generation may experience lower levels of success.49 In some of these studies, the definition of second generation is broadened to include the foreign born who arrive as children (the 1.5 generation). More nuanced research has found that the educational success of second-generation youth is conditional on legal status, the presence of strong coethnic community, and their familial context.50 The immigrant generation variable was derived from three questions asked in the UW-BHS senior survey: • Where were you born? • Where was your biological mother born? • Where was your biological father born? The measurement of immigrant generation in the UW-BHS is defined in the conventional style: (1) first generation—the student was born outside of the United States (or American territories); (2) second generation— the student was born in the United States and at least one parent was born outside of the United States (or American territories); (3) third generation—the student and both parents were born in the United States (or American territories). Following the convention established by the Census Bureau that all persons who are American citizens at birth are considered native born, UW-BHS respondents with one or more nativeborn parents are coded native born regardless of the respondents’ birthplace. For example, students born in Japan to native-born Americans are classified as native-born Americans. Respondents with missing data on birthplace or parental birthplace were assigned an immigrant generation based on the partial information that was reported. For example, if birthplace information was only reported for only one parent, then immigrant generation was defined in terms of the single parent. If respondents reported that they were born outside the United States but that they were currently American citizens and spoke English at home, then they were classified as third (or higher) generation. We followed similar logic for coding the small number of respondents who did not report their birthplace but reported that their parents were native born, that they spoke English at home, and that they were currently American citizens, classifying them as third generation or

98   From High School to College

higher. There were fewer respondents with missing data on birthplace, but some of them reported American citizenship, speaking English at home, and one foreign-born parent. We coded these students as second generation. Finally, some respondents (about 6 percent) had multiple missing records and were impossible to code. These cases are excluded from the descriptive tables here but are included in subsequent multivariate analysis with multiple imputation of missing data on immigration generation. Primary race-ethnicity and immigrant generation are closely intertwined (see online appendix table 3.A5).51 For the total UW-BHS population, only about one in ten students (13 percent) was foreign born and another 20 percent were second generation. However, about nine in ten Asian Americans were first or second generation. In contrast, more than eight out of ten white and black students were third-andhigher generation (this figure was even higher for American Indians). Virtually no Vietnamese or Cambodians were third-and-higher generation. Hispanics and Pacific Islander students had a more balanced distribution by immigrant generation. The racial-ethnic composition of the different immigrant-generation populations is shaped by historical patterns of immigration and also the relative size of groups. For example, nearly a quarter of first-generation immigrants in our survey and one-third of the second generation were white. About 60 percent of the foreign born and 40 percent of the second generation were Asian. The second generation was a particularly heterogeneous population with representation from all groups.

Mediating Variables: Social Origins In addition to the three key ascriptive variables of gender, primary raceethnicity, and immigrant generation, our analysis includes a number of other independent variables. To clarify our analytical focus (see chapter 1), these additional independent variables are identified as “mediating variables” because we use them to test hypotheses about the mechanisms that account for ascriptive inequality in educational outcomes. The most important of these mediating variables are socioeconomic status (SES) and family structure, which we collectively call social origins.

Social Origins: SES and Family Structure Advantage begets advantage—those who start out ahead typically finish ahead. This pattern is an axiom of stratification research. Specifically, adolescents who are reared by parents with higher education and more economic resources have a marked advantage in completing high school, attending college, and completing college over their peers from less

University of Washington-Beyond High School   99

advantaged families of origin.52 Family advantage, of course, does not guarantee success—there are numerous instances of students from privileged backgrounds who do not complete college as well as students from disadvantaged backgrounds “beating the odds” to reach the highest levels of educational attainment. Yet, if we could identify just one factor that contributed to educational and career success, it would certainly be an advantaged family background. In spite of the recurrent findings of a positive impact of SES on educational attainment, there remains considerable debate over the conceptualization and measurement of the social and economic characteristics of family background. One key dimension is total family income. Families with higher incomes can more easily afford the costs of higher education and private school tuition. Money can also pay for books, tutoring, and other experiences that promote learning and positive educational outcomes. Income and wealth also allow families with children to live in safe neighborhoods with better schools. However, SES is more than just familial income. The educational background, social standing, and occupational positions of parents are also important. The multiple dimensions of SES raise questions of conceptualization and measurement, especially if different aspects are not consistent across groups. One popular strategy has been to create an overall index of family or household SES—a composite score that reflects a composite of multiple dimensions of income, parental employment and occupations, and parental education. Unless there is theoretical interest in each dimension of SES, one variable is simpler to manage and interpret, especially if there are problems of measurement and missing data. However, there are still many unresolved conceptual and measurement issues with a global measure of family SES. Is the aggregate index simply the sum of its components, and how should the individual components be weighted? Are the observed indicators really an accurate reflection of an underlying global dimension of SES?53 Robert Hauser found that the observed indicators of mother’s and father’s education and occupation often had different and independent effects on educational outcomes.54 Rather than just one, there are probably many dimensions of SES with varying effects on educational and occupational outcomes. Without prejudging these issues, our “empiricist” approach is to operationalize eight key dimensions of family SES as separate empirical indicators, namely: father’s educational attainment, mother’s educational attainment, father’s employment, mother’s employment, father’s occupational status, mother’s occupational status, home ownership, and family structure. Although a more parsimonious measurement of family SES would be preferable, our preliminary research indicated that each background variable made a net independent contribution to explaining the influence of social origins on educational outcomes.

100   From High School to College

Father’s educational attainment and mother’s educational attainment were measured in the baseline survey with the standard census-style question, “What was the highest degree or level of school that your mother or mother figure (and father or father figure) completed?” followed by detailed response categories and a “not applicable” category for those without a mother (or mother figure) or father (or father figure). Preliminary research showed that a simple three-category classification—high school graduate or less, some college, and college graduate or higher—plus an additional category for “no response” and “not applicable” was the most efficient summary classification of parental education (that is, neither a linear term nor more detailed educational categories added significant explanatory power). The parental-education questions, as well as those about employment, occupation, and industry, asked the student to respond about their father (or father figure) and mother (or mother figure). A father or mother figure was defined as the person who is most like a father or mother, respectively, to the respondent. The actual relationships of parental figures could be an absent birth father, a coresident stepparent, a grandmother, or another relative. Despite the broad reach of the question, almost one in ten respondents (9.4 percent) did not know or report the education status of a father figure, and one in twenty (5.8 percent) did not know or report the education status of a mother figure. Parental employment (measured when the respondent was a high school senior) was coded using dummy variables (0 = not employed, 1 = employed). The responses to the questions about parental occupation were first coded by the detailed three-digit census occupational classifications, replicating standard Census Bureau procedures.55 Occupational socioeconomic index (SEI) scores, ranging from 0 to 100, were then assigned to each detailed occupational code.56 However, about one-third of student responses did not yield a specific parental occupation (or SEI score). In some cases, the parent was not working, but many more students did not report (or know) the occupation of their parent (or parental figure). To avoid the potential hazards of excluding cases or imputing values for such a large number of respondents, we coded parental occupational SEI as the interaction of the reported (not missing) SEI score multiplied by a dummy variable measuring whether the parent was employed with reported occupation (0 = missing occupation, 1 = employed with a reported occupation). Therefore, the impact of parental occupational SEI was only estimated for the subsample of respondents with a reported occupation (a pairwise-present assumption), and the impact of an unknown parental occupation was captured with a missing-data dummy variable. Home ownership, an additional indicator of family SES and residential stability, was measured by the question, “Does your family own or rent

University of Washington-Beyond High School   101

the home you live in?” Students living in owner-occupied housing were coded “1,” and renters were coded “0.” Our index of family structure was based on responses to the question, “Are you living with both your mother and father (biological or adoptive)?” Students were classified as having intact families only if they were living with birth (or adoptive) parents when the respondent was a high school student. We experimented with a series of alternative measures of family structure; however, we found only marginal gains in predictive power from increasingly complex family classifications relative to a simple dichotomous variable of intact versus nonintact family. Descriptive tabulations of social origins (SES and family structure) for UW-BHS high school seniors by gender, primary race-ethnicity, and immigrant generation are presented in appendix tables 3.A6 and 3.A7 (available online).57 The overall distribution of father’s education can be summarized as 30 percent, 30 percent, and 30 percent—roughly similar proportions in the three categories of parental education: high school or less, some college, and college graduate or higher. About 10 percent of students did not know the educational attainment of their father or father figure (or did not have a father or father figure). The average educational attainment of respondents’ mothers was slightly lower than that of their fathers. About 20 percent of African American, Cambodian, and other Asian students did not report a father’s education (the figures were somewhat lower for mothers). Parental education varied widely by primary race-ethnicity and immigrant generation. If parental education is the starting line for the race to pursue one’s own education, white students and East Asian students in our sample were more likely than the average student to have parents who graduated from college (about four in ten). American Indians, Pacific Islanders, Mexicans, Cambodians, and Vietnamese were much further behind—about one half had a father with a high school education or less and were even more disadvantaged in terms of maternal education. African American students fell between these groups—they had a much lower proportion of parents with college degrees (only half the level of white students), but they had smaller proportions of parents in the lowest educational level than did American Indians, Asian refugee populations (Cambodians and Vietnamese), and Mexican students. The Hispanic population was very heterogamous, with Mexicanorigin youth having considerably lower parental educational levels than “Other Hispanics.” Our data showed that parental educational levels rise monotonically across immigrant generations. Both first- and second-generation students are the children of immigrants, but the second generation has parental educational resources that are only marginally lower than those of

102   From High School to College

third-and-higher-generation students. The first generation is much more likely to have parents with very low educational attainments. Summary measures for other dimensions of social origins are shown in appendix table 3.A7, but with somewhat different samples for each variable. For example, the universe of students for the parental employment and parental occupational SEI variables was not precisely the same. About 82 percent of the total UW-BHS sample reported that their fathers (or father figures) were working in the preceding month, but only 65 percent reported a father’s occupation. In the descriptive tables, we report summary values (means) for parental occupational SEI for those with a valid (nonmissing) value for the variable. In the multivariate analyses in subsequent chapters, we rely on multiple imputation methods to include as many cases as possible. As with parental education, there were significant differences in the SES and family-background characteristics by ascription. We even found modest gender differences because male UW-BHS respondents were slightly more positively selective than females: they were four percentage points more likely to live in an owner-occupied home and 2 percentage points more likely to have an intact family. These “population” differences were due to the higher dropout rate among high school males (school dropouts were disproportionately from lower socio­ economic origins). When we analyze the gender gap in educational outcomes in subsequent chapters, disparities adjusted by social origins are considered a better measure of the baseline gap. There were much wider gaps in social origins by primary race-ethnicity and immigrant generation. In general, whites and East Asians were much more likely to have higher-status social origins, including higher rates of home ownership than all other groups. Filipinos were just a step behind and were more likely to live in owner-occupied homes than white students. Racial and ethnic inequality in parental employment and parental occupational SEI was more striking for respondents’ fathers than mothers. African Americans, American Indians, Mexicans, Cambodians, and Vietnamese were the most disadvantaged—employment rates among their fathers were ten to twenty percentage points lower than whites and six to twenty points lower on the parental occupational SEI scale. The maternal employment and occupational SEI indexes for African American students were also below those of whites, but not as low as those of Mexicans, Cambodians, and Vietnamese. Only about one-half of African American, Cambodian, Vietnamese, Pacific Islander, and Mexican students lived in houses that were family-owned, compared to over 80 percent of white students. Much has been written on the significance of family structure to children’s well-being and educational attainment. Youth who grow up in intact, two-parent families are more likely to graduate from high school

University of Washington-Beyond High School   103

and attend college than are students who have experienced a parental divorce or separation while growing up.58 A large part of the effect of family structure is the diminished economic resources available in a single-parent household. With only one potential adult earner in the household, single-parent families are generally poorer than two-parent households.59 Moreover, two-parent families can more easily monitor their children’s activities to curtail actions that are detrimental to their future success. It is also easier for parents in intact families to invest time with children and to assist with schoolwork.60 However, not all children in two-parent families are educationally advantaged relative to single-parent families. Children reared in blended families—with stepchildren and half siblings who are the joint children of both parents—have, on average, lower educational achievement than children in a single-parent family.61 Overall, 61 percent of UW-BHS high school seniors reported that they were currently living with both of their biological parents (or with both of their adoptive parents). The remainder—“not intact”—were not necessarily living in a one-parent or no-parent household; some students in this category may have been living with one biological parent and one stepparent. The category is also heterogeneous in terms of closeness and contact with nonresident parent(s) and the role of other parental figures (stepparents, parental-partners, grandparents, and other relatives) in the day-to-day lives of student respondents. After considerable exploration of these patterns, we focus on a simple indicator that distinguishes between intact biological and adoptive families (intact) versus all others (not intact). This simple measure captures almost all of the explanatory power of more complex classifications of family structure and its impact on educational outcomes. In contrast to how they fared with the socioeconomic dimensions, white students did not have the highest fraction with respect to intact family structures. More than 70 percent of Asian students—East Asians, Cambodians, and Vietnamese in particular (80 percent)—reported they currently live with both parents compared to only 64 percent of white students. The values for Mexicans and “Other Hispanic” students were only moderately lower than those for whites. American Indians and African American students were least likely to live in intact families; the proportions were 47 percent and 36 percent, respectively. First-generation immigrants experienced significantly lower levels of paternal SES and home ownership than second- and third-and-highergeneration students. The second generation ranked slightly below the third-and-higher generation on most family SES measures, but the differences were small and not significant. This pattern was reversed for family structure across immigrant generations—the first generation had the highest level of intact families, and the third-and-higher generation had the lowest.

104   From High School to College

Additional Mediating Variables: Encouragement and Student Performance Beyond social origins, two key variables are hypothesized to play important mediating roles in explaining educational disparities in terms of our three key ascription characteristics. These additional mediating variables are encouragement from significant others and student performance.

Encouragement from Significant Others to Attend College Students are likely to attend college not just because they can but also because they are encouraged to do so by family members, friends, and other influential persons such as teachers and other adult mentors. Encouragement from significant others (family, friends, teachers, and others) is generally expressed in socialization, in everyday interaction, and by the power of example (“I did it, so can you”). Parents often tell their children that they must study hard, get good grades, and go to college in order to have a successful life. Students may also feel encouraged to pursue college because of parental behaviors, such as taking an interest in their child’s homework, saving money for college, and helping in their child’s college search. Peers, teachers, and other adults can also play a significant role through their expressions of confidence in a student’s ability and potential. Adolescents often absorb the expectations of significant others and make them their own. The Wisconsin school of social stratification, which blended structural and social-psychological theories, emphasized the role of significantother influence (SOI) as a key mediator of the impact of family background on educational and occupational attainment.62 In addition, SOI may have a direct impact on educational outcomes independent of its role as a mediating variable. The UW-BHS senior questionnaire included the question, “What does [the significant other] think is the most important thing for you to do right after high school?” The significant-other question was asked six times, once for the father (or father figure) and then repeated for the mother (or mother figure), siblings, friends, favorite teacher, and an adult whose advice the student trusts. The possible responses were: (1) go to college; (2) enter a trade school, vocational school, or work apprenticeship program; (3) enter military service; (4) get a job; (5) get married; (6) doesn’t know; and (7) does not apply (for father and mother, only). For the measures of significant-other encouragement, students that checked “go to college” were considered encouraged, while all other responses, including “doesn’t know” and “does not apply,” were coded as not encouraged.

University of Washington-Beyond High School   105

The percentages of students whose significant other thought that the most important thing to do after high school was to attend a college are displayed in appendix table 3.A8 (available online) by type of significant other and by gender, primary race-ethnicity, and immigrant generation.63 Nonresponse was low for the encouragement questions, only about 2 percent. Assuming that SOIs are additive, we computed a summary index of encouragement as the sum of significant others who thought that the respondent should go to college. The overwhelming majority of high school seniors reported that family, friends, and teachers encouraged college attendance. None of the values were below 50 percent, and most were in the range of 70 to 80 percent. The overall encouragement index was 4.6 out of a possible total of 6.0. Three of the significant others—mothers, favorite teachers, and adult mentors (“an adult whose advice you value”)—were reported to have encouragement levels of over 80 percent. Since favorite teachers and mentors were chosen by the respondent (in the survey), there may have been some degree of favorability bias—students chose to report on teachers and mentors they considered supportive. The lower levels of encouragement from fathers and friends—around 73 percent—may reflect a deficit of persons in these categories (for example, in cases of absentee fathers and a lack of friends), but also that fathers and friends were modestly less nurturing than the other significant others. The still-lower average level of encouragement (62 percent) from brothers and sisters may have been due to similar factors. Some students had no siblings, or their siblings were too young or distant to provide encouragement. Female students reported higher levels of college encouragement than males. On the six-point summary scale of encouragement, females reported that 4.8 persons encouraged them to go to college, compared to 4.3 for males. In general, Asian Americans were much more likely to feel encouraged to go to college than all other racial and ethnic groups—about ten percentage points higher for individual significant others and about half a point on the six-point scale. Some of the disadvantaged groups reported less encouragement from fathers (African American, American Indian, Pacific Islander, and Mexican). Teachers appear to hold universally high expectations across all groups. The only minority group receiving consistently below-average levels of encouragement was American Indians, whose summary index was 4.2 compared to 4.5 for all other groups. In contrast to how they fared with the parental SES variable, whites did not stand out as advantaged on encouragement. Immigrants reported receiving higher levels of encouragement than did second and third-and-higher generations—particularly among family members. It may be that families and friends of immigrants (and the second generation) are more ambitious for their children, but

106   From High School to College

immigrants are also more likely to have parents and siblings to provide encouragement.

Academic Performance One of the most important, and proximate, determinants of educational attainment is high school academic performance.64 The standard indicator of academic performance is school grades, most often summarized in the composite measure of grade point average (GPA). Every study that includes GPA finds that it is invariably the single most important predictor of college enrollment and graduation.65 Two fundamental issues need to be addressed in research that includes academic performance— its measurement and interpretation. These are complex issues that have given rise to considerable confusion and controversy, even within the research community. We first review the procedures and assumptions of measuring academic performance and then address the more complicated questions of meaning and interpretation. The key measure of academic performance in the UW-BHS project was self-reported GPA, a fairly standard measure in most educational surveys, though it needs to be used with caution.66 The UW-BHS senior survey included the following question: “In general, what grades do you get?” The possible responses were: (1) mostly As, (2) half As and half Bs, (3) mostly Bs, (4) half Bs and half Cs, (5) mostly Cs, (6) half Cs and half Ds, (7) mostly Ds, and (8) mostly below Ds. Students showed little reluctance to answer the question—more than 98 percent checked one of the response categories. There are, however, reasonable concerns about the validity of self-reported grades. In addition to the relatively crude response categories in the survey question, some students may not have known their overall grades or may have been inclined to misreport them, perhaps inflating their grades above their true level. To address the question of validity of self-reported GPA, we compared survey responses with actual school grades for the subset of all respondents in the five high schools in District 1. School transcripts provided letter grades for each completed course for each student. The letter grades in school records were converted into a summary index of school GPA with the following scale: A = 4.0, A- = 3.7, B+ = 3.3, B = 3.0, B- = 2.7, C+ = 2.3, C = 2.0, C- = 1.7, D+ = 1.3, D = 1.0, and all other letter grades = 0. The responses to the survey question on self-reported grades were converted into a GPA metric using the approximate midpoints of the range. For example, the category “half As and half Bs” was recoded to 3.5; “mostly Bs” was recoded to 3.0; and so on. The top category, “mostly As,” was recoded to 3.9, and the bottom category, “mostly below Ds,” was recoded to 0.5.

University of Washington-Beyond High School   107

The validity of self-reported GPA was evaluated for the subsample of 5,744 UW-BHS respondents from District 1 for whom we had actual school grades from administrative records. We compare these two indicators of academic performance for the matched sample of students in appendix table 3.A9 (available online).67 The first column shows the assumed midpoints of each interval of self-reported grades from the UW-BHS senior survey. The second column shows the “true” (from school transcript records) mean GPA for the matched sample of students. For example, the students who checked “mostly As” had an average GPA of 3.62 (based on school records), which is roughly .30 lower than our assumed value of 3.90 for “mostly As.” The apparent inflation of self-reported grades relative to school records was also evident for the next two grade categories. For example, students who reported “half As and half Bs” should have had an average grade of 3.5, but this value was well above their actual mean GPA of 3.15 based on school grades. Similarly, the assumed midpoint of “mostly Bs” of 3.0 was higher than the actual GPA of 2.83 from school grades. The tendency for students to inflate their high school grades, however, did not hold throughout the grade distribution. The third column of appendix table 3.A9 shows the difference between self-reported grades (assumed midpoint values) and actual average GPA based on school records. In general, students in the upper range of the grade distribution inflated their grades when self-reporting, while students in the lower range appear to have deflated self-reported grades relative to school records. We should be cautious about overinterpreting the comparisons between self-reported and actual school grades; the differences may be artifacts of measurement rather than conscious behavior. Students may have had varied understandings of the reference period when asked, “In general, what grades do you get?” For example, some students may have been thinking about the current academic year, while other students were considering their entire high school career. Although school grades might be considered the gold standard of measurement, school records were only available for one school district and only for courses taken in local schools. Despite the variations in measurement, all measures of student achievement were highly correlated with one another. The correlation between self-reported GPA and school-based grades was .79. All in all, self-reported GPA appears to have been a reliable and valid measure of student achievement in school. The one adjustment made to our measure of self-reported GPA was to substitute the means from column 2 (based on school GPA) in appendix table 3.A9 instead of the assumed midpoint of each interval for the nine discrete response categories. This provided slightly more “empirically based” values to transform the discrete response categories into a continuous variable.

108   From High School to College

Differences between the “true” school-based GPA and the recalibrated self-reported GPA by gender, race-ethnicity, and immigrant generation are shown in appendix table 3.A10 (available online).68 The first data column, labeled “actual,” contains the school-based GPA. The second column, “self-reported GPA,” is based on responses to the sub­jective question from the UW-BHS survey, but we recalibrated the values with the mean school-based value. With this adjustment, the overall (all UW-BHS seniors in District 1) means of the school-based grades and selfreported grades were approximately the same: 2.96 and 2.94. In general, self-reported GPA is a close representation of actual GPA based on school records. We found modest differences between selfreported and actual GPA across race-ethnicity and other groups, but the difference was usually within one-tenth of a GPA point. Females tended to slightly underestimate their grades, while males tended to slightly overestimate them. Asians and whites tended to underreport, while other minorities tended to overreport. Although school-based GPA would be a more objective measure of academic performance, self-reported GPA (recalibrated) appears to be a valid proxy for objective grades. Moreover, self-reported GPA was available for the universe of UW-BHS seniors, while school records were limited to students in District 1. For these reasons, we generally relied on the recalibrated self-reported GPA as the measure of academic performance in this study.

Interpreting the Direct and Indirect Role of Academic Performance We now turn to the more complicated and contentious issue of the interpretation of academic performance, as measured by GPA. Some researchers, and many policy makers, assume that academic performance is a simple product of inherited ability—the capacities for learning and intellectual development. There are three problems with this interpretation: • the assumption that ability is innate and heritable • the assumption that ability can be measured by academic performance • the assumption that ability is unitary and can be measured with a single summary index Each of these assumptions needs to be questioned. Most of the existing research on educational outcomes that uses standard measures of academic performance takes their significance and meaning to be selfevident. Higher test scores and school GPA are considered positive phenomena that are highly correlated with many other dimensions of family background and subsequent socioeconomic outcomes. The problem is how to interpret the mediating role of GPA on between-group

University of Washington-Beyond High School   109

(by gender, SES, race-ethnicity, and immigrant generation) differences in educational outcomes. Continuing the discussion introduced in chapter 1, we now examine how measures of academic performance are frequently misinterpreted by laymen and also in some academic studies. Individuals differ from each other in a wide variety of physical and mental attributes, and there is little doubt that a significant fraction of these differences are biological in origin. Individuals have varying innate capacities for learning as they do for physical and social development. Some of these innate capacities are probably heritable: children are more likely to resemble their parents and siblings than unrelated persons. However, there is also a huge stochastic or random distribution of innate capacities across generations. Despite the claims by ideologues that observed relationships in SES across generations are due to the heritability of abilities, there is almost no reliable scientific evidence to support these claims.69 The biological capacity to learn cannot be measured directly, but only indirectly through measures of performance, such as GPA and test scores. Children who perform well in academic tests may have been born with greater capacity for learning, but their performance has likely also been enhanced by greater exposure to a stimulating environment with motivated parents and caregivers. Students from advantaged homes are more likely to have had early life experiences—such as being read to—that better prepare them to learn and take tests upon entering school. Students who learn quickly, answer questions in class, and obtain higher grades on exams are often praised and rewarded by teachers, parents, and even some peers. This positive reinforcement certainly plays a key role in motivating students to continue their schooling through high school and into college. The social environment, “nurture,” is deeply involved in the translation of inherited ability, “nature,” into academic performance or school grades. The moderately high correlation between the measured IQ of parents and their children (estimates range from .4 to .7) may have some component due to the inheritance of capacity to learn between families, but it also may be explained by differences in the social environments between families. Research has shown that the inheritance of IQ explains only a small share of the intergenerational correlation of income and other socioeconomic outcomes.70 In this study, our goal is to measure and explain the observed disparities in educational outcomes by the ascriptive characteristics of gender, primary race-ethnicity, and immigrant generation. Although individuals differ in their learning abilities (inborn predispositions or capacities for learning), there is no justification, theoretical or empirical, for assuming that ascriptive groups differ in their potential to learn. In other words, ability may vary widely within groups, as within the general population, but there is no evidence of differences, on average, in the distribution of innate abilities between groups.

110   From High School to College

There are, however, differences in measured academic performance between groups. The translation of ability into academic performance is strongly affected by social processes. Social processes begin at birth with interactions between infants and families (or other caregivers). Social interactions increase as children interact with neighbors and relatives and even more as they enter formal schooling. These social environments vary in terms of resources and culture (the content of communication and socialization). The association between ascriptive background and academic performance is not evidence of differences in inherent ability between groups, but rather is evidence of differences in the social worlds between groups (the socioeconomic resources of families, neighborhoods, and schools) and perhaps in socialization (the communication of culture across generations). As noted earlier, the inborn potential to learn cannot be measured, since nature interacts with nurture immediately after birth. Our focus is not on the direct effects of academic performance on educational outcomes but on the role that academic performance plays in mediating the impact of ascription—gender, primary race-ethnicity, and immigrant generation—on educational outcomes. Some of this role is likely due to social origins (including SES and family structure)— advantaged families, those with more resources, time, and motivation, will coach and train their children to do well in school. Net of social origins, certainly some fraction of the observed differences in educational outcomes by gender, primary race-ethnicity, and immigrant generation is mediated by academic performance. We posit that some part of these net ascriptive disparities in GPA may be due to “effort,” which could be due to the role of communities and social networks to socialize and mold their children to do well in school, independently of their socioeconomic resources. The logic of this approach is tested, in a preliminary fashion, in table 3.6 with decomposition of the total effects of ascription on GPA (school grades) into components explained by test scores and a residual, which might be considered as a proxy for effort. We posit that a substantial component of the net impact of GPA on educational outcomes is due to differential effort, which is affected by family socioeconomic background, socialization, and early life experiences. Regardless of their abilities, students who put more effort into their schooling can improve their grades. For example, students who regularly attend classes, take an interest in the subject matter, and study the material outside of class are able to achieve higher grades than students of comparable ability who do not apply themselves. This assumption draws support from recent research that compared the relative impact of SAT scores and high school GPA on college success. William Bowen, Matthew Chingos, and Michael McPherson found that high school GPA is a much better predictor of college graduation

University of Washington-Beyond High School   111

than SAT scores. They concluded that high school grades reflect both academic ability and effort.71 Generally, high-ability students are able to score well on standardized exams, but getting high grades in school classes also requires sustained effort and motivation. Persistence in attendance, paying attention and participating in classes, and doing homework can boost GPA. Since these same traits are needed to complete college, high school GPA provides an independent measure of college success above and beyond SAT performance. Although some people believe that between-group differences in test scores are due to group differences in inborn abilities, this ideological claim is not supported by theory or research. As we have argued, these group differences are much more likely to be explained by early life experiences, such as socialization, that better prepare some children to do well in standardized tests. In table 3.6, we estimate how much of the observed differences (gaps) in self-reported GPA among the three ascriptive groups can be explained or mediated by test scores. The residual gaps—differences in GPA that are not explained by test scores—are our first-order approximation of effort. We define effort as a group-specific property that motivates and prepares children to do well in school, net of test scores. To address these issues, we compared the actual and self-reported GPAs of students with their predicted GPAs based on scores on standardized exams. From District 1 school records, the UW-BHS project had access to students’ Washington Assessment of Student Learning (WASL) scores based on a standardized statewide exam. For the 2002 to 2005 cohorts, we had WASL scores for 4,224 tenth-grade students in District 1 who took the WASL. The following equation was estimated from a linear regression of actual GPA on math and reading WASL scores for the sample of UW-BHS students with matched District 1 records of grades and WASL scores: Actual (school) GPA = α + β1 (WASL math score) + β2 (WASL reading score) + μ Test scores proved to be a good predictor of measured GPA with an adjusted R-squared for the model of .383. Based on the results of this exercise, we estimated a predicted GPA score for each UW-BHS student in the matched sample: Predicted GPA = –2.07 + .00646 (WASL math score) + .00613 (WASL reading score)

2.96

2.79 3.10

3.03 2.67 2.73 3.10 3.15 2.86 3.35 3.02 3.00

By gender  Male  Female

By primary race-ethnicity  White   African American   American Indian   Asian American    East Asian    Cambodian    Vietnamese    Filipino    Other Asian 2.98 2.78 2.84 3.00 3.04 2.85 3.17 2.96 2.95

2.81 3.04

2.94

(2)

(1)

For total population

SelfReported

Actual

GPA

2.98 2.82 2.90 2.95 3.04 2.88 2.96 2.96 2.88

2.93 2.94

2.94

(3)

Predicted

Test Scores

Residual (effort)

(9)

Race-Ethnicity Gap with Whites — — — -0.21 -0.15 -0.05 -0.14 -0.08 -0.07 0.02 0.05 -0.03 0.06 0.06 0.00 -0.13 -0.10 -0.04 0.19 0.21 -0.02 0.00 -0.02 -0.02 0.06 -0.04 -0.10

Total GPA Gap

(8)

Race-Ethnicity Gap with Whites — — — -0.36 -0.15 -0.21 -0.30 -0.08 -0.22 0.07 0.09 -0.03 0.12 0.06 0.06 -0.17 -0.10 -0.07 0.32 0.34 -0.02 0.01 -0.01 -0.02 0.07 -0.03 -0.10

Residual (effort)

(7)

Gender Gap Between Males and Females — — — 0.23 0.01 0.22

Test Scores

Total GPA Gap

(6)

Self-Reported GPA

Gender Gap Between Males and Females — — — 0.31 0.01 0.30

(5)

(4)

Actual (school-based) GPA

Intergroup Disparities in GPA Due to WASL Test Scores and Residual (effort)

(N)

2,529 674 73 628 161 165 159 82 61

1,873 2,348

Table 3.6    Decomposition of Intergroup Gaps in GPA (Actual and Self-Reported) by Washington Assessment of Student Learning (WASL) Test Scores and the Residual (Effort) Among UW-BHS High School Seniors in District 1

3.04 2.99 2.98 2.99

By immigrant generation   First generation   Second generation   Third-and-higher generation  Total 2.97 2.97 2.96 2.96

2.78 2.86 2.85

2.86 2.96 2.96 2.95

2.86 2.86 2.87

-0.12 -0.12 -0.10

-0.20 -0.08 -0.10

Immigrant Generation Gap with Third Generation 0.06 0.16 -0.10 0.01 0.00 0.01 — — —

-0.32 -0.21 -0.20

-0.12 -0.12 -0.10

-0.09 0.00 -0.03

Immigrant Generation Gap with Third Generation 0.01 0.11 -0.10 0.01 0.00 0.01 — — —

-0.21 -0.12 -0.14

451 720 2,681 3,852

79 143 98

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample is limited to 4,224 UW-BHS seniors in District 1 with a reported school GPA and Washington Assessment of Student Learning (WASL) test scores. The coefficients in the prediction equation were obtained from: GPA (self-reported) = a + b1(WASL math score) + b2 (WASL reading score) + µ. Column 1: School-based (actual) GPA is the mean grade of all high school courses from the transcripts of matched UW-BHS high school seniors in District 1 with at least four graded courses. Column 2: Self-reported grades are based on responses to the question, “In general, what grades do you get?” Self-reported grades are recalibrated to the mean values for each response category from column 2 in appendix table 3.A10. Column 3: Predicted GPA is the value predicted from the equation: GPA (self-reported) = a + b1(WASL math score) + b2 (WASL reading score) + µ. Columns 4 and 7: Total gaps are the absolute differences in GPA by gender, race-ethnicity, and immigrant generation relative to a reference population. Columns 5 and 8: The share of the intergroup difference attributable to group differences in test scores, the GPA gap minus the predicted GPA based on WASL test scores. Columns 6 and 9: The share of the intergroup GPA gap attributable to effort in school—the residual intergroup GPA gap not explained by group differences in WASL test scores. Effort = Self-reported GPA - predicted GPA. Predicted GPA = 2.07 + .00646 (WASL math score) + .00613 (WASL reading score). GPA = grade point average. UW-BHS = University of Washington-Beyond High School. WASL = Washington Assessment of Student Learning.

2.71 2.82 2.83

  Pacific Islander  Mexican   Other Hispanic

114   From High School to College

The first two columns of table 3.6 show the actual and self-reported (recalibrated) GPA of UW-BHS students (similarly to appendix table 3.A10). The predicted GPA for each ascriptive group is shown in the third column. The next two panels show how much of the observed intergroup gaps can be accounted for by test scores. For example, consider the male-female gap in actual GPA in column 4. There is a substantial difference, with females outperforming males by .31 of a point in actual GPA (3.10 - 2.79). Virtually none of the gender gap in GPA (only .01) can be explained by differences in WASL test scores. In general, men scored a bit higher in math tests and women scored a bit higher in reading, but the composite model shows that almost all of the difference is unexplained. We conclude that the gender gap in GPA was due to differences in social environments and processes between male and female students but was unrelated to differences in test taking. The next panel repeats the exercise for self-reported GPA (see columns 7, 8, and 9). The gender gap is smaller, but the decomposition shows the same results. Our preliminary interpretation is that females put more effort into their schooling—their academic performance exceeded that which would be predicted by measured test scores. There were modest differences by primary race-ethnicity in actual GPA (column 4) and self-reported GPA (column 7) in the neighborhood of .2 to .3 of a point. East Asians had grades slightly higher than white students (.12 actual and .06 self-reported), but all other groups had lower grades than whites. African Americans, American Indians, and Pacific Islanders had average GPAs about -.3 (actual) and -.2 (self-reported) below white students. A fraction of this gap (about -.1, but with wide variations) is explained by test scores. Racial and ethnic differences in GPA attributable to test scores were most likely due to social conditions in childhood, not ability between groups. Note that a more substantial component of the variation in raceethnicity differences in GPA is independent of the between-group variation in test scores—factors that lower students’ performance relative to their measured potential. These factors are not measured here, but they could be products of opportunities and socialization that varied between groups: school attendance, incentives to study, and so on. It is interesting to note that the Asian GPA advantage was almost entirely endogenous—they exceeded their potential as assessed by test scores. Vietnamese students had lower test scores but received higher grades in school than did white students. There were only slight differences in observed GPA by immigrant generation. First-, second-, and third-and-higher-generation students had the same grades in school. However, the lack of a difference in GPA between the first and third-and-higher generation is the result of offsetting forces. The first-generation students performed lower on standardized tests than students who had been in the United States for multiple

University of Washington-Beyond High School   115

generations. This difference can be explained by limited exposure to American society, incomplete fluency in English, or lack of familiarity in taking standardized tests. This deficit was, however, completely offset by over­performance (endogenous) in GPA. It seems that first-generation students do indeed “try harder” to make up for their deficit. In the analytical chapters that follow, we find that self-reported GPA mediates a fraction of ascriptive differences in educational outcomes. Some of this is accounted for by family SES and other background variables that are exogenous to GPA (because higher-SES families are better able to prepare their children for school). However, there are unmeasured factors associated with gender, primary race-ethnicity, and immigrant generation that may also explain variations in academic performance that affect educational outcomes. We suspect that some groups translate the same resources into differential outcomes by training and motivating their children to try harder in school.

The UW-BHS Data The UW-BHS project evolved with a conscious awareness of the history of prior research and many of the landmark studies in the field of educational stratification. Indeed, many aspects of the study design, the survey questions, and follow-up procedures were based on the examples from the Census Bureau, the Department of Education, and several regional studies. We closely studied the published scholarship, including methodological appendices, and consulted with the leading experts in the field. Nevertheless, our data collection encountered many problems, as discussed in the first part of this chapter. These problems are not ours alone. Despite valiant efforts, problems of school nonparticipation, selective nonresponse within schools, and attrition in follow-up surveys are endemic in the field of educational-stratification research. Recognizing these limitations, we expect that the quality of the UW-BHS data is sufficient to make a significant contribution to research on educational stratification. This expectation draws on an established base of knowledge of American education that draws on varied sources of data, including censuses, national surveys, and many regional and local studies. The robustness of findings from many empirical studies adds to the confidence that our results might be of interest beyond the geographical region of our data collection.

Chapter 4 The College Pathways Model

C

is an event—a moment in time—that is celebrated by graduates and their friends and family. But college graduation is also a process, or the culmination of a process, that begins in childhood and extends through adolescence and often into the early years of adulthood. In chapter 1, we used the analogy of a college-bound train to describe the sequential process leading to college graduation. The surest way to graduate from college is to get on the college-bound train and avoid falling off. It is possible to recover from early missteps, but the odds of college graduation are much lower for those who drift off track. In this chapter, as in chapter 2, we describe the process leading to college graduation as a series of steps or sequential educational transitions. In chapter 2, the process consisted only of three steps—high school graduation, college entry, and college completion. This model, though deceptively simple, provided a clear understanding of how and why college graduation rates expanded and stagnated during different periods and generated new insights into the reasons for differential progress by racial and ethnic groups (table 2.3). During periods of educational expansion, whites and Asian Americans were able to increase their rates of college graduation because of higher levels of achievement at each stage of the process. For native-born, disadvantaged minorities (African Americans, American Indians, Pacific Islanders, and Mexicans), higher rates of high school completion were the major component behind their increasing numbers of college graduates. They also experienced some gains from higher rates of transition from high school to college, though less than whites and Asian students. But disadvantaged minorities lost ground, both relative to their advantaged peers and in absolute terms, in the transition from college enrollment to completion. In this chapter, we formulate and describe a more complex and nuanced model of the pathways to college graduation based on data from the UW-BHS project. The College Pathways Model identifies the five key steps or transitions that lead to college enrollment and completion. College ollege graduation

116

The College Pathways Model   117

graduation is defined as the receipt of a bachelor’s degree from a fouryear college or university. The objectives of this chapter are to describe the model and present educational-transition rates for each step by gender, race-ethnicity, and immigrant generation. We then use demographic decomposition to explain disparities in college graduation produced by inequality at each of the five steps of the College Pathways Model. The results show that the transitions to enrollment in a four-year college and to completing college are more consequential than those at the earlier stages of the College Pathways Model. By definition, models are simplified abstractions of reality. Focusing on a few points of very complex educational histories allows researchers to summarize broad patterns and to discover connections and associations that are obscured when looking at the totality of every possible data point. But models are only as good as their assumptions. A simplified version of the College Pathways Model assumes that students move sequentially through the five steps—in a linear progression—leading to college graduation. The simple three-step model presented in chapter 2 also assumed linear progression—high school graduation, college enrollment, and college completion. However, many students who dropped out of high school eventually did resume their schooling, and some even graduated from college. These nonlinear progressions to college graduation were not measured in the analysis in chapter 2 because persons were simply classified by their highest level of schooling. We did not have direct measurement for students who resumed their schooling after dropping out at an earlier stage. The assumption of sequential progress through the College Pathways Model is more complicated and problematic. In this chapter, we begin with a comparison of a (complete) model of all possible pathways to college graduation with a simplified model of linear progression along the college-bound track. The simple model works reasonably well. Of the 30 percent of UW-BHS seniors who graduate from college within seven years after high school graduation, three-quarters of them (or 23 percent of all UW-BHS seniors) followed a linear path through the five sequential transitions of the College Pathways Model. This means that an assumption of linear progression through the College Pathways Model is a useful summary of the complexities leading to college degrees. However, a quarter of college graduates (7 percent of all UW-BHS seniors) followed an alternative pathway. Alternative pathways to college graduation are too important, both empirically and theoretically, to ignore. After presenting the results of progress through linear-progression steps, we analyze which students obtain a college degree though an alternative pathway. To our surprise, we find that students from advantaged backgrounds are not only more likely to graduate from college via the linear pathway but also through alternative pathways.

118   From High School to College

Pathways to College Graduation The immediate precursors of college entry occur in high school, mostly during the student’s senior year. For students who plan to attend a fouryear college, there are a variety of steps that lead to submitting a college application. Because college admissions can be competitive, students are usually advised to apply to several colleges, including their top choices and “back-up” schools. In addition to the formal application for admission, prospective college students must also arrange for SAT or ACT scores, high school transcripts, letters of recommendation, and a report on family finances to be submitted as part of the application process. Applications are generally due in the late fall or early winter of students’ senior year of high school for admission the following fall. Aspiring college entrants begin their senior year with a series of impending tasks, including identifying the list of colleges to which they will apply, asking teachers and counselors to write letters of recommendation, scheduling college entrance exams, and beginning to fill out application forms. In reality, the process of planning to attend college is not always so organized. Some students begin the process but do not submit complete applications. Other students submit partial or late applications. The vast range and nature of postsecondary institutions further diversify the application process. Traditional four-year colleges, especially those with competitive admissions, have formal processes that follow a standard calendar. However, there are variations from school to school and for different types of applicants. Most community colleges have flexible admissions policies allowing students to enroll and register on the day that classes begin. The messiness of reality—and the uncertainty of student attitudes and behavior in planning for college—is reflected in the responses to the UW-BHS senior survey. When asked during the spring of their senior year about college plans, 21 percent of UW-BHS students said they were unsure if they would continue on to college the next year. Even more students gave mixed signals on their immediate plans for college. Some students reported that they planned to attend college but did not name a specific college in a follow-up question. Based on these results, it seems that many high school students have a vague idea that they want to (or should) go to college but postpone making a commitment as they weigh other options. Although college plans are supposed to crystallize during the senior year of high school, college aspirations and planning probably begin many years earlier. As soon as students enter high school, they (and their teachers and guidance counselors) begin to sort themselves into different trajectories as they enroll in honors and college preparatory classes or in a general or vocational curriculum. Course planning during high school is influenced by college admission requirements.

The College Pathways Model   119

Many four-year colleges have minimum expectations for high school credit hours in English, math, sciences, foreign languages, and other fields. For example, many colleges will not consider students for admission unless they have completed three years of math, including two years of algebra and one year of geometry (or more). In practice, many community colleges (and some four-year colleges) offer remedial classes, but the general rule is that college-bound students should begin taking courses that meet college admission requirements in the ninth grade. These courses also prepare students to take college entrance exams, such as the SAT or ACT, during their junior or senior years. College aspirations may have even deeper roots. Students in middle school, and even in elementary school, often hear teachers, family, friends, neighbors, and other adults offer encouragement and advice to plan on college if they want to get ahead in life. Indeed, 84 percent of UW-BHS respondents reported that their family always expected them to go to college. Given all the potential variations and complications, it may seem impossible to establish a temporal pathway of the critical steps from high school to college completion. Our objective is not to search for every factor that might explain how and why students are able to find their way from high school to college but to identify a few key turning points that account for most of the variation between ascriptive groups in the transitions to college enrollment and completion. There is a long tradition of research that attempts to explain the determinants of educational outcomes, including college graduation. Many studies are published every year identifying the correlates and antecedents of educational success and failure. The findings of these studies are rarely surprising—children from poor families, poor neighborhoods, and poor schools are much less likely to succeed than their peers from advantaged backgrounds. These patterns are intertwined with race and ethnicity, teacher characteristics and skills, tracking within schools, peer influences, and many other factors. The problem for educational researchers and policy makers is not too little research but the lack of a conceptual framework that can bring order and meaning to the complicated movements of students through schools. Having studied NCES longitudinal surveys, Clifford Adelman proposed in 1999 the conceptual model of the Toolbox to organize the sequence of pathways from high school to a bachelor’s degree with the following steps:1 Step 1:  Demographic background and high school history Step 2:  Postsecondary entrance (timing and type of institution) Step 3:  First postsecondary year history (curriculum and performance) Step 4:  Factors of financing postsecondary education in the early years

120   From High School to College

Step 5:  Postsecondary attendance patterns Step 6:  Extended postsecondary history (curriculum and performance) Step 7:  Final model, with complete academic history The Toolbox has been widely used by a variety of researchers as well as policy makers to identify critical roadblocks to college completion and to estimate the impact of student backgrounds on educational outcomes. However, the complexity of many of the models and limitation of the analytical sample to students who entered postsecondary education means that the framework cannot be applied to a model of sequential conditional probabilities of advancing from one step to the next. Our approach is to blend several genres of prior research—the simplicity of the Wisconsin model of educational aspirations and attainment,2 the idea of stages or pathways from the Toolbox model, and the logic of educational-transition ratios (conditional probabilities) from demography. The basic College Pathways Model is illustrated in figure 4.1. This aim of the model is to simplify complexity with a narrow focus on the key steps leading to a college degree. The value of this model, or any model, is not that it explains the trajectory of each and every student, but rather that it identifies critical turning points that account for a large share of the inequalities in college graduation rates by gender, race and ethnicity, and immigrant generation. In chapter 5, we test a broad range of hypotheses organized by the key transitions of the College Pathways Model. One of the key assumptions underlying this model is that high school students internalize educational goals based on their social environment, including the expectations of people around them. The first step in the model is the formation of aspirations to graduate from a four-year college. Aspirations may range from generalized beliefs to firmly held plans. The next step is turning college aspirations into realistic expectations. If a student believes that a college degree is both desirable (aspirations) and possible (expectations), the next logical step is to prepare for college by taking the requisite coursework in high school. Preparation is more than coursework, however, as it also involves taking college entrance exams and applying for college while in high school. There are all sorts of postsecondary educational institutions, but only one type has the primary goal of a baccalaureate degree—a four-year college. Enrollment in a four-year college keeps a student on the path. The final step is the completion of college coursework and other requirements of a four-year degree. Figure 4.2 presents the basic descriptive “facts” of progression through the College Pathways Model based on the UW-BHS data. Each box in the top row represents a step in the process and includes the percentage of all UW-BHS high school seniors who arrived at this stage, regardless of

The College Pathways Model   121 Figure 4.1     The College Pathways Model College aspirations

College expectations

College preparation

College enrollment

College completion

Source: Author’s compilation.

their prior trajectory. For example, 76 percent of UW-BHS high school seniors aspired to a college degree, 69 percent expected to obtain one, 49 percent were college prepared, 32 percent enrolled in a four-year college right after high school, and 30 percent completed college, earning a BA or BS degree within seven years of leaving high school. These percentages differ slightly from some of the descriptive tables in chapter 3 because the sample of all UW-BHS seniors in figure 4.2 is limited to those with nonmissing responses to the survey questions on college aspirations, expectations, and preparation (N = 8,719). The numbers (N) in each box represent the number of UW-BHS high school seniors who reached that stage. The arrows identify the possible pathways, and the numbers embedded in each arrow are the outflow (conditional) probabilities from each state to the next. Another way to think about figure 4.2 is to consider the boxes as states of the model and the arrows as representing the probabilities of movement from one state to the next. The top horizontal pathway represents continuity on the college-bound path—linear progression through the College Pathways Model. The downward sloping arrows represent attrition—veering off the college-bound track. The upward sloping arrows represent alternative pathways—the probabilities of returning to the college-bound path after having left it. The two arrows from each stage add up to 100 percent, representing all possible outcomes. The first step in the process leading to college graduation is aspiring to complete a college degree, which was reported by 76 percent of UW-BHS high school seniors. The complement—about one in four high school seniors—represents high school seniors who did not aspire to complete a college degree, though they may have been planning some sort of postsecondary schooling. Almost nine out of ten students (89 percent) who aspired to graduate from college also expected to do so. The 11 percent of students who aspired to graduate from college but did not “realistically” expect to do so represent dashed aspirations. Perhaps these students anticipated barriers that would prevent them from realizing their aspirations. The percentage of all students (of the total sample) who expected to receive a college degree—69 percent—is the sum of the products of pathways

.24

No college aspirations N= 2,084 .96 .96

.04

.11

.89

No college expectations N=2,741

College expectations N= 5,978 [69% of all seniors]

.32

.90 .90

.10

.68

Not college prepared N= 4,405

College prepared N= 4,314 [49% of all seniors]

.97 .97

.03

.37

.63

Did not enroll N= 5,888

Four-year enrollment in college N= 2,831 [32% of all seniors]

.27

.90 .90

.10

.73

Did not complete N= 6,068

College completion N= 2,651 [30% of all seniors]

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The UW-BHS sample includes all respondents with nonmissing data on college aspirations, college expectations, and college preparations. Educational-transition ratios (embedded in arrows between boxes) are defined as the conditional probability of continuing from one stage of the College Pathways Model to the next. Having college aspirations and expectations is defined as aspiring and expecting to complete a four-year college bachelor’s degree. Being college prepared is defined as completing two of the following: took SAT or ACT, took or plan to take AP exam, or applied to a four-year college.

All UW-BHS seniors N= 8,719

.76

College aspirations N= 6,635 [76% of all seniors]

Figure 4.2     The College Pathways Model of Educational-Transition Ratios with All Possible Paths

The College Pathways Model   123

(educational-transition rates) leading to the state. For example, .69 is the sum of .676 (which is .76 × .89) and .0096 (which is .24 × .04). The percentage of all UW-BHS seniors who graduated from college, 30 percent, is the sum of the products across all possible pathways leading to college graduation. The majority of college graduates followed the linear-progression path directly across the top from college aspirations to expectations to preparation to enrollment to completion. However, many students strayed from the college-bound path at each transition. The cumulative toll of these downward paths along with very low rates of recovery are the reasons that 70 percent of students did not get a college degree within seven years after high school graduation. Behind the diagram of arrows and pathways lies the drama between the ambitions and achievements of high school seniors. Almost every high school student in the UW-BHS project planned on some sort of postsecondary schooling in the future. The 24 percent who did not aspire to graduate from college may have had career plans that only required a high school credential or vocational training. The 11 percent of students who aspired to get a college degree but acknowledged that they did not expect to do so were an especially interesting case. Limited financial resources, poor academic performance, or lack of confidence are possible reasons for their attrition from college aspirations to college expectations. It is much harder to offer a rationale for the 4 percent of students without college aspirations who nevertheless expected to earn a college degree (see the first upward arrow in figure 4.2). Perhaps, these students simply did not read the questions carefully and checked an illogical response. Or perhaps some students really did not want to go to college but expected that they would have to go because of family pressures. The translation of subjective plans into actions is a crucial point on the path to a college degree. We constructed an indicator of college preparedness to include only students who appear to have taken purposeful actions. A college-prepared student was defined as a high school senior who had completed at least two of three tasks: (1) taken AP tests, (2) taken a college entrance exam (SAT or ACT), or (3) applied to a four-year college. According to this index, only 68 percent of seniors who expected to graduate from college were prepared. One out of three students fell off the path to a college degree at this juncture on the pathway. There are many popular aphorisms—“all talk and no action” or “all hat and no cattle”—that capture the lack of behavioral follow-through that is all too common among high school students (and many adults as well). Along the lower pathway—the non-college-bound track—90 percent of students who did not expect to graduate from college did not bother preparing for college. This seems logical, but what about the upward path of 10 percent of students who had no college expectations but were college prepared? These students may have been preparing

124   From High School to College

for a demanding postsecondary career that did not necessarily require a college degree (such as nursing or an engineering trade). Another possibility is that these students were taking the steps to prepare for college even though they did not expect to earn a college degree. The cumulative toll of the first three stages of the model (aspirations, expectations, and preparation) left slightly less than half (49 percent) of all high school seniors prepared for college. However, being college prepared in high school is not a guarantee of success. In our sample, over one-third (37 percent) of highly motivated, ambitious, and college-ready students did not enroll in a four-year college right after high school. Lack of resources would seem to be the most likely reason that these able and prepared students did not immediately enroll in a degree-granting college. Many of these motivated students may have postponed college entry or enrolled in a less expensive community college. Other students may have had family or employment obligations that hindered college enrollment. The final step along the college pathway is college completion by those who enrolled in a four-year college right after high school. Almost three out of four students who enrolled in a four-year college obtained a college degree within seven years. College completion, like all the preceding steps in the pathways model, requires considerable motivation and persistence to stay on track, especially in the face of possible failure and the attractions of short-term economic gains from full-time employment. The major takeaway point from figure 4.2 is that significant numbers of students drop off the college-bound path at each stage of the process. The cumulative impact in our study was that only 30 percent of high school seniors earned a college degree, compared to the 76 percent who expressed aspirations to do so. As noted in chapter 3, the NSC data underrepresent college enrollment (relative to the UW-BHS follow-up survey data), but the basic conclusion would not change with a slightly higher rate of transition from being college prepared to college enrollment. In general, movements along the upward pathways in figure 4.2 (back to the college-bound route) were relatively rare. For example, only one in ten students who did not enroll in a four-year college right after high school graduated from a four-year college (compared to a 73 percent college-completion rate of students who enrolled in a four-year college). As noted, economic conditions often force some highly motivated and prepared students to enroll in a two-year college. The tuition costs of a two-year college are considerably lower, and many community college students defray expenses by living at home and working a part-time job. After completing an associate’s degree, community college students can transfer to a four-year institution and receive academic credit for most of their academic coursework. This sounds easy, but in reality, transferring from a two-year college to a four-year college requires considerable motivation and planning.

The College Pathways Model   125

Of course, there are “late bloomers,” who manage to enroll and grad­uate from college in spite of the odds. Conceding this point and also acknowledging that not every measurement is exact (some students undoubtedly checked the wrong box on our questionnaire), the overwhelming evidence from figure 4.2 points to the conclusion that the odds are stacked against college graduation for those who drop off the college-bound train.

The Linear-Progression College Pathways Model Figure 4.3 presents a simplified version of the College Pathways Model with only a straight-line trajectory to college graduation along the top row. There are no upward arrows or alternative pathways in this version. The conditional probabilities leading to college graduation in figure 4.3 only include students who have consistently remained on the collegebound track. How good is the simplified, linear-progression version of the College Pathways Model? Based on the criterion of completeness, the complete College Pathways Model is superior to the linear-progression variant because it covers all possible trajectories. The linear-progression model predicts that only 1,996 UW-BHS students graduate from college, which is only 75 percent of the “true” count of 2,651 students who actually did graduate from a four-year college (based on NSC data). However, the linear-progression model has one major advantage—parsimony—which allows us to focus on only five key transitions leading to a college degree. Parsimony is essential to understanding the key stages leading to college graduation. However, parsimony comes with a cost if we ignore the 25 percent of college graduates who made the journey through an alternative pathway. After taking a closer look at each of these five key transitions of the linear-progression version of the College Pathways Model, we will return to the question of alternative pathways.

College Aspirations and Expectations Before presenting more detailed empirical data on each stage of the College Pathways Model, it is important to remind readers of the distinction between the percentages of all UW-BHS seniors who reached each stage of the model (the percentages in figure 4.2) and educationaltransition ratios—the conditional probabilities of advancement from one stage to the next, which are shown embedded in arrows in figures 4.2 and 4.3. Both figures are shown in table 4.1. The percentages, which are shown in the left-hand panel, tell us who (of the initial population) made it to this stage. The educational-transition ratios, which are shown in the middle panel of table 4.1, are the conditional probabilities that (Text continues on p. 130.)

.24

No college aspirations N = 2,084

.76

1.0

College aspirations N = 6,635

.11

1.0

College expectations N = 5,891

No college expectations N = 2,828

.89

.32

College prepared N = 4,033

Not college prepared N = 4,686

.68

1.0

.34

.66

Did not enroll N = 6,053

Four-year enrollment in college N = 2,666

1.0

.27

College completion (in seven years) N = 1,996

Did not complete N = 6,723

.73

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The UW-BHS sample includes all respondents with nonmissing data on college aspirations, college expectations, and college preparations. Educational-transition ratios (embedded in arrows between boxes) are defined as the conditional probability of continuing from one stage of the College Pathways Model to the next. Having college aspirations and expectations is defined as aspiring and expecting to complete a four-year college bachelor’s degree. Being college prepared is defined as completing two of the following: took SAT or ACT, took or plan to take AP exam, or applied to a four-year college.

All seniors N = 8,719

Figure 4.3     The Linear-Progression College Pathways Model of Educational-Transition Ratios

76 75 66 82 91 67 81 80 77 69 66 71

Primary race-ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic 73 81 77 77

74 78

Gender Male Female

Immigrant generation First generation Second generation Third-and-higher generation Total reporting immigrant  generation

76%

All students

Aspired to Grad College

64 73 70 70

70 65 55 74 86 53 76 70 68 52 53 64

66 71

69%

Expected to Grad College

42 57 51 51

51 43 35 56 75 37 50 52 46 40 33 41

46 53

49%

College Prepared

21 36 35 34

36 25 18 31 41 18 27 32 28 18 20 26

29 35

32%

Enrolled in College

Percent of All High School Seniors (N)

1,022 1,585 5,441 8,048

5,308 1,181 136 1,414 499 243 266 243 163 173 300 207

3,890 4,829

8,719

(Table continues on p. 128.)

23 32 33 32

35 19 13 31 36 18 37 28 29  9 18 19

27 33

30%

Graduated in 7 Years

Table 4.1    Percent of All High School Seniors at Each Stage of the College Pathways Model and Educational-Transition Ratios by Gender, Race-Ethnicity, and Immigrant Generation: UW-BHS High School Seniors

76 75 66 82 91 67 81 80 77 69 66 71

Primary race-ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic 73 81 77 77

74 78

Immigrant generation First generation Second generation Third-and-higher generation Total reporting immigrant  generation

76%

Gender Male Female

Aspired

All students

Table 4.1    (Continued)

86 89 90 89

91 84 79 88 93 77 92 86 86 73 79 88

88 90

89%

Expected/ Aspired

62 74 70 70

70 60 56 73 85 61 63 71 66 66 56 62

65 71

68%

Prepared/ Expected

Educational-Transition Ratios

50 63 70 66

71 57 58 56 55 52 56 61 63 51 60 65

64 67

66%

Enrolled/ Prepared

77 72 76 75

78 62 61 73 71 65 87 67 76 41 68 63

72 77

75%

Graduated/ Enrolled

742 1,282 4,188 6,212

4,040 886 90 1,153 454 164 216 194 125 119 199 148

2,879 3,756

6,635

Aspired

641 1,138 3,767 5,546

3,678 747 71 1,020 422 126 198 166 108 87 158 130

2,526 3,365

5,891

Expected

396 839 2,622 3,857

2,568 450 40 749 359 77 124 118 71 57 89 80

1,651 2,382

4,033

Prepared

199 527 1,828 2,554

1,832 255 23 422 196 40 69 72 45 29 53 52

1,064 1,602

2,666

Enrolled

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The universe of students in the first panel is based on all UW-BHS high school seniors with nonmissing data on college aspirations, expectations, and preparedness. The educational-transition ratios in the middle panel are percentages based on the number of students reaching the prior stage of the College Pathways Model. The numbers of UW-BHS students enrolled in a four-year college and graduated from college are based on National Student Clearinghouse (NSC) administrative records.

1,022 1,585 5,441 8,048

5,308 1,181 136 1,414 499 243 266 243 163 173 300 207

Primary race-ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic

Immigrant generation First generation Second generation Third-and-higher generation Total reporting immigrant  generation

3,890 4,829

Gender Male Female

Total 8,719

All students

Number of UW-BHS Respondents at Each Transition

130   From High School to College

a student will continue from one stage to the next. In other words, the educational-transition ratios measure the “risk” (conditional probability) of educational advancement based on those who have reached the preceding state of the model. The numerators are the same in both measures, but the denominators differ. The percentages are based on the full sample of 8,719 UW-BHS seniors who responded to the survey questions on college aspirations, expectations, and preparation, while the educational-transition ratios are based on the numbers of students “at risk” of each transition (who reached the prior stage). The numbers reaching each stage (the denominators for the educational-transition ratios) are presented in the right-hand panel of table 4.1. For the first step up the ladder, college aspirations, the percentages of all students and the educational-transition ratios are the same. More than three-fourths of all UW-BHS high school seniors (76.1 percent) aspired to obtain a BA or BS degree or more. There are a variety of ways to conceptualize and measure educational ambitions—a generic concept meant to encompass the ways that children and adolescents express their educational intentions and goals.3 In this study, we emphasize the distinction between aspirations and expectations, even though past research has sometimes conflated the two concepts.4 An aspiration is akin to a wish— something that is desired in the abstract, without any consideration of the likelihood of attaining it, while expectations are constrained by a sense of realism.5 Brian Jacob and Tamara Wilder Linkow observed that expectations refer to what individuals think will happen while aspirations refer to what they hope will happen.6 In the UW-BHS senior questionnaire, educational aspirations were measured by the question “How far would you like to go in school?” Possible responses were: (1) less than high school graduation; (2) high school graduation only; (3) less than two years of college, vocational, or business school; (4) two or more years of college, including a two-year degree; (5) four- or five-year degree; (6) master’s degree or equivalent; (7) PhD, MD, or other professional degree. This question was immediately followed by a similar question designed to measure educational expectations: “Realistically speaking, how far do you think you will get in school?” The response categories were the same. Educational expectations are similar to aspirations but with an awareness of constraints, such as the costs of schooling, family resources, academic interests, and abilities. Although there may still be some degree of “wishing for the best” in educational expectations, we assume that expectations are a more realistic assessment of the future than are aspirations. In our analysis, educational aspirations and expectations have been recoded as dichotomies, with responses indicating college graduation or higher (the top three codes) equal to 1 and those indicating less than college graduation equal to 0. The recoding follows the standard practice to

The College Pathways Model   131

focus on the major dividing point of college graduation. We use the terms educational aspirations and college aspirations (and educational expectations and college expectations) interchangeably. At present, only about one-third of young American adults have graduated from college.7 This figure stands in stark contrast to the high level of students who aspire and expect to graduate from college: 76 percent and 69 percent, respectively (table 4.1). The UW-BHS estimates are in line with national surveys of high school seniors.8 The educational ambitions of American youth have skyrocketed in recent decades, especially from the 1970s to the 1990s.9 This disconnect between college aspirations and graduation rates was captured in the title of an important book, The Ambitious Generation: America’s Teenagers, Motivated but Directionless.10 A generation or two ago, Americans were worried about the leveled aspirations (and anti-intellectualism) of American youth, particularly in working-class communities.11 However, this no longer appears to be the case. A significant share of the rising educational and occupational ambitions is due to changes in the status and roles of women in American society.12 The traditional role of women as primarily homemakers, dependent on their husband’s income, declined dramatically after the 1960s. Although gender differences in occupational roles persist, women are much more likely to participate in the paid labor force than ever before. The data in table 4.1 show that the young women in our study had slightly higher educational aspirations than their male peers. Both groups held high aspirations, but females were three to four percentage points more likely to aspire to graduate from college. National data show that the narrowing of the traditional gender gap (and emerging female advantage) in educational aspirations and attainment is largely due to a change in the ambitions and behavior of women from lower SES families.13 One hypothesis is that the erosion of commitment to lifelong marriage has increased the awareness among women of the need for higher educational credentials to compete in the labor market and be self-supporting. We found moderate differentials in educational aspirations by race and ethnicity and by immigrant generation. White and black youth shared similar aspirations—about 75 percent hoped to graduate from college. Because the UW-BHS survey was conducted in the spring of the students’ senior year, most high school dropouts were not included in these data. The higher high-school dropout rates among men and minority youth means that measured college aspirations are skewed (slightly) upwards for these groups (relative to those who entered high school). With the exception of Cambodians, Asian students had higher educational aspirations than white students did. This finding is consistent with research based on national data.14 In sharp contrast to the above-average levels of aspirations among the Asian American students, American Indians, Pacific Islanders, and

132   From High School to College

Hispanic students had lower aspirations for college graduation. They were still high—about two-thirds of students in these groups aspired to graduate from college—but these figures were about eight to ten percentage points below those of the other racial and ethnic groups. There was a modest gap (about five percentage points) between the low college aspirations among Mexican-origin youth and the slightly higher levels among other Hispanics. First-generation youth who arrived in the United States as children or teenagers had slightly lower college aspirations than third-and-highergeneration youth, whereas second-generation youth had the highest aspirations for college graduation—a few percentage points above the third-and-higher generation. Although first- and second-generation students share a common characteristic—both are children of immigrant parents—their differences probably arise from the impact of duration of residence. The first generation experiences the handicaps of being new to the United Sates and the American school system, while the second generation has only experienced American society.

The Translation of College Aspirations into Expectations Although the concepts of college aspirations and college expectations are very similar, there was a slight nuance in the survey question that asked respondents to consider potential barriers—economic, social, and cultural— that might make it impossible (or unlikely) to realize their ambitions to complete college. As might be expected, our data showed a drop-off in expectations relative to aspirations. A little over 10 percent of students who responded that they wished to complete college said that they did not expect to do so. Only 69 percent of all 8,719 UW-BHS seniors expected to graduate from college, but 89 percent of the 6,635 seniors who aspired to graduate from college expected to do so. What factors might account for the falloff in expectations relative to aspirations? In his 2005 assessment of the logic and assumptions of the Wisconsin model of educational attainment, Stephen Morgan observed that aspirations tap goals while expectations represent beliefs.15 Aspirations or goals are largely shaped by external social influences, including socialization from family and the broader society. Aspirations are not necessarily “rational”—they embody hopes and dreams. Expectations, however, require an awareness of what is likely to happen based on prior experiences, including prior educational performance. Expectations are also informed by knowledge of real-world constraints—family finances, personal limitations, proximity of colleges, the likelihood of admission, and other factors. Morgan argued that the net effect of educational expectations on educational attainment (not simply mediating-family and

The College Pathways Model   133

significant-other influences) may reflect nonsocialization factors, such as the structural barriers stressed by Alan Kerckhoff.16 This theoretical distinction between aspirations and expectations may actually be wider than our empirical measures designed to capture them. The questions were asked sequentially in the questionnaire with the hope that the students would be prodded to reassess the college aspirations with the prompt, “realistically speaking.” However, many students may not have weighed these words carefully and simply answered the questions in the same way.17 Overall, American youths, as represented by UW-BHS respondents, are very optimistic about their educational trajectories. Over 76 percent of our students aspired to complete college, and almost 90 percent of these students (66 percent of all UW-BHS seniors) expected to earn a bachelor’s degree or higher. Because going to college is so highly valued in American society, high school seniors may be expressing socially desirable norms as much as individual preferences. Survey questions that ask about other highly valued behaviors, such as voting in the next election, visiting and keeping in touch with family and friends, and losing weight are also likely to elicit very high levels of reported aspirations. But since most good intentions are not realized, these abstract claims may not be highly predictive of behavior. A generation or two ago, high school students did not need a college degree to find a good middle-class job that would support a family. There were blue-collar jobs in factories, as well as many careers in retail and wholesale trades and the service sector, for which a high school diploma was a sufficient credential for an entry-level position. In recent years, the wages and career prospects have declined for traditional blue-collar workers in many sectors because of deindustrialization, deunionization, and increasing economic competition with low-wage countries. Many other careers have been “professionalized” with expectations of college degrees for entry-level positions in criminal justice, health care, and business. These changes have been the most significant among women, as many of the traditional “feminine occupations,” such as nursing and office assistants (formerly secretaries), have been educationally upgraded, with a college degree now considered the usual prerequisite. Although the high levels of reported college aspirations and expectations by high school seniors may be platitudes (“yes, I know that ‘I would like to earn a college degree’ is how I am supposed to answer this question”) and unrealistic, they also represent an awareness that a college degree is one of the few protections against low wages and marginal career prospects. Despite the high levels of college aspirations and expectations among all UW-BHS high school seniors, there were important differences among groups.

134   From High School to College

The gender differences were small but consistent for both aspirations and expectations. Young women were three percentage points more likely to aspire to complete college than young men, and women with college aspirations were two percentage points more likely to expect that they would reach their goal. As we shall see, there were even wider gender differences in behavior than in attitudes. The female “edge” was pervasive, cumulative, and statistically significant. The fact that women put more stock in the importance of a college degree than men may reveal a slightly more realistic orientation to career planning. This pattern was consistent with the finding that most women intend to work full time, regardless of their marital status.18 There were no differences between African American and white high school seniors when it came to buying into the American Dream—about 75 percent of each group aspired to a college degree. However, black youth were somewhat less likely to expect that they would obtain one. The difference was relatively small, about five to seven percentage points. This pattern was consistent with the interpretation that black youth, when planning their careers, “factor in” the likelihood of barriers, such as poverty and poor academic performance, that are independent of preferences and socialization. Whether the lowered expectations are based on an accurate assessment of specific barriers, or just a generalized fear, it is an important distinction that is lost in the conflation of educational aspirations and expectations. Asian American high school seniors are, in general, more likely than any other group to aspire to and expect to graduate from college. Kim Goyette and Yu Xie conclude that high educational ambitions are the characteristic most widely shared by Asian national groups that differ widely in culture, history, migration experience, and location in American society.19 Specifically, East Asians (Koreans, Chinese, and Japanese) and Vietnamese students are very confident that they will realize their high aspirations for college graduation. Filipino and Cambodian students, on the other hand, are less likely to expect that they will achieve their educational aspirations. We found nonblack and non-Asian minorities, including American Indians, Pacific Islanders, and Mexican Americans, to be least likely to aspire and expect to graduate from college. Their absolute levels were quite high—two-thirds aspired to complete college, and three-fourths of those with college aspirations expected to reach their goal. Yet, these figures were ten to fifteen percentage points below those for other races and ethnic groups. There were also slight differences between expectations and aspirations by immigrant generation. First- and second-generation immigrants, by three and one percentage points, respectively, were less likely to expect to realize their aspirations than third-and-higher-generation native-born Americans.

The College Pathways Model   135

The Translation of College Expectations into Behavior: College Preparedness One of the key issues in the educational literature is the gap between attitudes and behavior. It is relatively easy for a student to think about going to college, but preparing for college is quite another matter. This problem is sometimes identified as the attitude-achievement paradox.20 Some groups hold very positive abstract attitudes about wanting to go to college but not the specific attitudes that lead to preparing for college. Being fully prepared for college requires a concentrated effort to complete the high school courses required for college admission. College preparatory courses are usually more rigorous than regular courses, with more required homework and term-paper assignments. Some students may simply be expressing their “real attitudes” by avoiding difficult courses, but the differences between choices and “steering” may be conflated, as teachers, guidance counselors, and peers are likely to direct many students to general education courses. To create a survey-based measure of four-year college preparedness, we considered three behavioral measures from the UW-BHS senior survey: whether the student had applied to a four-year college, whether the student had taken (or planned to take) a college placement exam (SAT or ACT), and whether the student had taken or planned to take an AP exam. Column 3 in table 4.1 shows that 68 percent of students who expected to earn a college degree were college prepared according to our index. As before, this number is not the percent of all high school seniors who were college prepared; that would be a much smaller number (approximately 49 percent). It is not too surprising that students without college ambitions were not preparing for college; the critical question is why one-third of students who expected to graduate from college were not preparing for college in high school. The disparities in the college-preparedness transition ratio by gender, race-ethnicity, and immigrant generation bear a close resemblance to the patterns evident in the transition from aspirations to expectations. Overall, female students were about five percentage points more likely than male students to be college prepared, given expectations. Among the racial-ethnic groups, East Asians were the most prepared. They were a startling fourteen percentage points more likely than white students to follow through on their expressed college-bound intentions by taking advanced classes (AP exams), college entrance exams, and applying to four-year colleges. About 85 percent of East Asian (Korean, Chinese, and Japanese) students who expected to graduate from college were on the college track in high school. Filipino students were just one percentage point above white students, while Vietnamese, Cambodians, and other

136   From High School to College

Asians were less likely (four to seven percentage points) than their white peers to be prepared, given expectations. Vietnamese had very high aspirations and expectations to finish college, comparable to East Asians; however, they were much less likely, according to our indicators, to be prepared for college, given expectations. In sharp contrast to the pattern among Asian American students, the educational-transition ratios from college expectations to college preparedness were much lower for African American, American Indian, Pacific Islander, and Hispanics (most notably Mexicans) students. Their educational-transition ratios range from 58 to 68 percent among students who expected to graduate from college. The percentages of all students who were college prepared reflect the cumulative toll from lower aspirations, expectations, and preparedness. For example, among all UW-BHS seniors, 51 percent of white students and 56 percent of Asian Americans were college prepared in high school. For East Asian students, 75 percent of high school seniors were prepared. However, only 43 percent of African American, 38 percent of American Indian, and 33 percent of Mexican high school seniors were prepared. First-generation students had lower levels of college preparedness than second- and third-and-higher-generation students—similar to their depressed values for college aspirations and expectations. Second-generation students, however, had an edge—they were about four percentage points more likely to be college prepared, given expectations, than third-andhigher-generation students. If there is a problem getting immigrant-origin students motivated and prepared to go to college, it appears to be limited to the recently arrived first generation. There are two possible interpretations for the below-average ratios of preparedness among disadvantaged minorities (along with men and first-generation immigrants). The first is that they overreported their college ambitions. They said they wanted to graduate from college, but they didn’t really mean it; otherwise, they would have followed through by taking advanced courses and college entrance exams and submitting college applications. The second interpretation is that many of the unprepared students really desired to attend college, but they encountered barriers to information, a lack of guidance, and poor grades that inhibited their actions. This interpretation is similar to Roslyn Mickelson’s concept of concrete attitudes that are informed by constraints.21 Of course, these two interpretations are not mutually exclusive; there are probably both internal motivations and external constraints that shape behavior. We lean toward the second interpretation for several reasons. Recall that the disadvantaged minorities had, in general, lower college aspirations and expectations. This means that the universe of students exposed to being college prepared (limited to those with college aspirations and expectations) represents a smaller and possibly more

The College Pathways Model   137

selective (or determined) fraction of minority students than the comparable group from the majority population. Of course, it is possible that many of the minority students who aspired and expected to graduate from college exaggerated their chances, but there is no obvious reason why their motivations would be fundamentally different from their majority peers. However, their situations—classrooms and social environment— may have been different in other ways. As such, the behavioral choices of high school students with similar goals may not be simple, conscious decisions that align attitudes and behavior. For example, students from an advantaged background are more likely to be “tracked” into collegeorientated classes with more skilled teachers and more engaged students than their less advantaged peers.22 Once on the college-bound track, students may simply be conforming to expectations of their parents, peers, and teachers by signing up to take the SAT and applying for college. If a student makes a decision not to apply to college, they may face considerable “push-back” from parents and peers. Minority and first-generation students may be less likely to have college-educated parents and other adult mentors to keep them on track.

Enrollment in a Four-Year College The surest path to college graduation is to enroll in a four-year college right after high school. But there are other possible paths to college graduation. Some students take a “gap year”—a year off for travel or for selfdiscovery. Others enter the military or work in civilian jobs for several years before entering college. Many more begin in community college and then transfer to a four-year college. These alternatives can be rewarding and many students who pursue them do eventually graduate from college. However, the odds of college graduation are much higher for students who begin their post–high school career by starting college right away. This pattern is evident in virtually all research on the question. Our primary indicator of college enrollment is based on matched records of UW-BHS students with college-enrollment records from the NSC. As noted in chapter 3, enrollment in a four-year college is understated in NSC relative to the UW-BHS follow-up survey. However, the off-setting gain is that we avoid the missing data problem of the UW-BHS follow-up survey—about 10 percent of the original sample of UW-BHS high school seniors were not contacted in the follow-up survey, a group much less likely to attend college. The fourth column of the middle panel of table 4.1 shows the conditional probabilities of enrollment in a four-year college among high school seniors who are college prepared (and who also aspire and expect to be a college graduate). Overall, about 66 percent of college-prepared students were enrolled in a four-year college one year after high school

138   From High School to College

graduation. This educational-transition ratio can be contrasted with the 32 percent of all UW-BHS high school seniors who enrolled in a four-year college the year after high school (see left panel). Another way to compare these figures is to observe that 51 percent of all seniors left the collegebound path in high school, and another 17 percent of all seniors dropped off the college track by not enrolling in college (or one-third of the 49 percent of high school seniors who were college prepared). With the College Pathways Model as our framework, the objective is to explain why only 66 percent of college-prepared seniors enrolled in college. Because of the smaller and more selective sample of college-prepared students, we might expect to see a narrowing of disparities between gender, racial-ethnic, and immigrant-generation groups as students move along the college pathway, as the less ambitious, motivated, and organized students fall by the wayside. For example, only 46 percent of men compared to 53 percent of women “survived” the cumulative attrition of college aspirations, expectations, and preparedness (see the percentage columns in the left-hand panel of table 4.1). Nonetheless, women out­ performed men, though the gap is only a modest three percentage points in the educational-transition rate to college enrollment. In contrast to the persistence of a slight gender gap at all stages of the pathways model, our data reveal an apparent reversal in some of the racial and ethnic disparities in the transition to four-year college enrollment relative to earlier stages. The advantage of Asians and of East Asians in college intentions and preparedness disappears in the transition from college preparedness to four-year college enrollment. Overall, about 71 percent of college-prepared white students enrolled in a four-year college one year after high school. All other racial-ethnic groups were considerably less likely to enroll in a four-year college than whites, given comparable levels of preparedness. There was an even larger gap between immigrant groups (both first and second generation) and long-resident natives (third-and-higher generation) in the transition from preparedness to college enrollment. Almost 70 percent of college-prepared high school students in the third-and-higher generation went on to attend a four-year college, compared to 63 percent in the second generation and 50 percent in the first generation. Later in this chapter, we evaluate the quality of college enrollment data from the NSC with a more expansive index of college enrollment. This inquiry is partially motivated by the unexpected low transitions to college enrollment among Asian American students. But first, let’s consider the final stage of the process from college enrollment to college completion.

Completing College The final step in the College Pathways Model is the transition from college entry to college completion. By college entry and college completion, we mean enrolling in a four-year college and graduating with a bachelor’s

The College Pathways Model   139

degree within seven years of high school graduation, respectively. The rightmost column in the first panel of table 4.1 shows that 30 percent of all UW-BHS seniors graduated college within seven years of high school graduation, including those who detoured along the College Pathways Model. In contrast, the educational-transition ratio from college enrollment in a four-year college to college completion within seven years is 75 percent. There are, of course, students who begin at a community college and then transfer to a four-year college, but the majority of young adults who receive a bachelor’s degree begin their studies at a four-year college. The educational-transition ratios in table 4.1 are based on data for students who followed a linear progression through the College Pathways Model. The curriculum and requirements in American universities and colleges are organized so that a bachelor’s degree can be completed in four years of continuous study. However, the average time to complete a college degree is five years.23 The seven-year graduation rate includes most degrees that are completed, including those that are delayed because of transfers, part-time attendance, and other interruptions. The complement of the college completion rate is the college dropout rate. There are significant gender and racial-ethnic (but not immigrantgeneration) disparities in the college completion rate—the educational transition from college enrollment to college graduation. These disparities persist even though the students who enroll in a four-year college right after high school are selective, including only the most successful students who have survived all the earlier transitions. Women are more likely to complete college than men—77 percent to 72 percent, respectively. Recall that women outperformed men at every prior stage of the process. All other things being equal, the smaller proportion of men who entered college should have had an edge over their female peers because prior attrition removed more of the less qualified (or motivated) men than women. The gender gap, however, favored women at every step along the way. Racial and ethnic gaps in college completion were wider than those observed in high school. With the exception of Vietnamese students, Asian students were less likely to complete college than white students, but the gaps were relatively modest—five to ten percentage points. Dis­advantaged minorities were much more likely to drop out of college than whites. Only about 61 to 63 percent of African American, American Indian, and Mexican youth who enrolled in a four-year college earned a degree in seven years, compared to 78 percent of comparable white students. The disparity between immigrant generations is considerably less than those between racial and ethnic groups. First-generation youth who experienced lower educational transitions at earlier stages of the process were just as likely to complete college if they were able to enroll in a four-year institution.

140   From High School to College

Comparing College Enrollment Rates Between NSC Data and the UW-BHS Follow-Up Survey In table 4.2 we compare alternative estimates of college enrollment based on NSC records with estimates based on our own one-year UW-BHS follow-up survey. Table 4.2 follows a similar format to table 4.1, with percentages of all high school seniors in the left-hand panel and educationaltransition ratios in the middle panel. Two sets of figures appear under the panels showing percentages and educational-transition ratios. The first column, labeled “NSC only,” shows the percent enrolled in a four-year college of all the UW-BHS seniors (and the educational-transition ratio from college prepared to college enrolled) based on the NSC data. These are the same figures as those presented in table 4.1 except that the universe of students is limited to the 8,065 seniors who were reinterviewed in the UW-BHS one-year follow-up survey. The second column, labeled “NSC or UW-BHS,” counts every student who was reported to be college enrolled in either source. Most of the discrepant cases were students who reported to be college enrolled in the follow-up survey but not in NSC records (see table 3.4). The broader definition of college enrollment (when both sources are used) significantly increases the proportion of all UW-BHS seniors who enrolled in a four-year college from 34 to 43 percent. The impact on the transition from being college prepared to college enrollment is even larger, jumping from 67 to 82 percent. Our focus here is not so much on the differences between the two estimates but on the patterns of disparities by gender, race-ethnicity, and immigrant generation for the two estimates. While the gender gap is basically the same, there is a major difference in the transition from college preparedness to college enrollment among Asian American students. With the more expansive measure of college enrollment, Asian American students are no longer at a disadvantage in this transition compared to white students. For example, the gap of 17 percentage points between East Asian and white students (55 percent compared to 72 percent) in the educationaltransition ratio to college enrollment in the NSC data disappears completely in the combined NSC and UW-BHS measure of college enrollment (86 percent for East Asians compared to 85 percent for whites). The other group whose disadvantage is significantly reduced is first-generation immigrants. Among other groups, the numbers shift upwards, but the dis­ parities remain about the same. The NSC records appear to underestimate college enrollment among East Asian, Vietnamese, and first-generation immigrants significantly. For other groups, the patterns of inequalities are remarkably similar in both the NSC and in the combined NSC and UW-BHS samples.

8.7

11.4 10.3 7.8

7.3 8.3 7.1 13.5 23.9 6.2 12.0 6.9 4.2 7.4 7.0 9.2

8.6 8.5

8.5

% Point Change

67

51 63 70

72 59 58 57 55 52 55 62 66 51 61 66

65 68

67%

NSC Only

82

76 80 84

85 73 73 80 86 68 79 72 75 65 80 84

81 83

82%

NSC or UW-BHS

15.4

25.5 16.7 13.5

12.8 14.5 15.0 22.9 30.9 16.0 24.0 10.3  9.0 14.0 19.5 17.8

15.5 15.0

15.2

% Point Change

Educational-Transition Ratios from College Prepared to College Enrolled

7,535

947 1,496 5,092

4,998 1,010 127 1,322 472 225 251 231 143 162 272 174

3,542 4,523

8,065

Total

3,753

385 819 2,549

2,517 413 40 728 349 75 121 116 67 57 87 73

1,591 2,324

3,915

College Prepared

UW-BHS Seniors in the Follow-Up Survey

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: “All UW-BHS seniors” includes all students in the one-year follow-up survey with nonmissing data on college aspirations, expectations, and preparedness. The educational-transition ratio is based on the number of UW-BHS seniors who enrolled in a four-year college relative to the number of college-prepared UW-BHS high school seniors. “NSC only” includes students reported to be enrolled in a four-year college in the NSC. “NSC or UW-BHS” includes students reported to be enrolled in four-year college in the NSC or the UW-BHS follow-up survey. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

44

35

45 35 26 46 67 24 39 41 35 27 29 38

34 47 45

38 27 19 33 43 18 27 34 31 19 22 29

Primary race-ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic

40 45

43%

NSC or UW-BHS

22 37 37

31 37

Immigrant generation First generation Second generation Third-and-higher  generation Total

34%

All students

Gender Male Female

NSC Only

Percent College Enrolled of All UW-BHS Seniors

Table 4.2    Alternative Estimates of Enrollment in a Four-Year College by NSC and UW-BHS Follow-Up Survey Data

142   From High School to College

Although we rely heavily on the NSC data for measures of college enrollment and completion (as explained in chapter 3), the discrepancy for Asian Americans and first-generation immigrants is a serious problem. Accordingly, we will compare the NSC estimates of college enrollment with a broader measure of college enrollment based on combining data from the NSC records and the UW-BHS follow-up survey.

Comparing the Complete College Pathways Model and the Linear-Progression Model George Box, the famous statistician, once remarked that all models are wrong, but some are useful. A useful model is a parsimonious abstraction of the complexities of reality but also one that captures the basic principles of a process. It allows the analyst to see the outline of the forest instead of many individual trees. For purposes of this study, we propose a simplified version of the College Pathways Model that assumes that college graduation is a linear progression from college aspirations to college expectations to college preparation to enrollment in a four-year college to college completion. This simplified model does not account for every case—there are some students who fall off the college-bound track but manage to get back on at a subsequent stage. The question is not whether the linear-progression model is accurate in all cases, but whether it is useful. In other words, do the gains from simplicity offset the loss of information from tracking every transition? Table 4.3 addresses this question by comparing all college graduates with the subset of college graduates who followed the linear-transition steps described in figure 4.3. The comparison is measured for the total population of UW-BHS high school seniors and for ascriptive subgroups defined by gender, race-ethnicity, and immigrant generation. The top row of table 4.3 compares the actual seven-year graduation rate of 30.4 percent among UW-BHS high school seniors, with the predicted graduation rate of 22.9 percent if only graduates who followed a linear path were counted. The actual graduation rate of 30.4 percent includes students who followed every possible trajectory of alternative (downward and upward) pathways in figure 4.2. The complete College Pathways Model is comprehensive, but it does not distinguish between common and rare trajectories. The linear-progression version has the opposite virtue—it identifies the five most common pathways of the complete model, including continuity on the college-bound track and the four major “exit ramps” that lead to failure. Its limitation is that it misclassifies 7.5 percent of UW-BHS respondents, namely the students who took a detour from the collegebound track but still managed to earn a degree from a four-year college. In table 4.3, we take a closer look at the differences between the complete College Pathways Model and the linear-progression version by

23.0 32.1 33.1

15.0 24.0 25.6

21.8 13.6

30.9 18.3 8.0 8.1 7.4

9.1 4.7

8.0 5.3 2.6

7.1 7.7

7.5%

Via Alternative Pathways

100 100 100

100 100

100 100 100

100 100 100

100%

Total

 9 13 12

 9 25

11 17 13

13 11

12%

Enrolled in Four-Year College

78 59 60

69 63

62 54 75

63 62

63%

Enrolled in Two-Year College

13 27 28

22 13

27 29 13

24 26

25%

Not Enrolled

Enrollment Status in Year Following High School for Alternative-Pathway College Graduates

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: All students includes UW-BHS high school seniors with nonmissing data on college aspirations, expectations, and preparedness. College graduation is based on the number of UW-BHS high school seniors who graduated from college in seven years based on the National Student Clearinghouse administrative records. Linear progression means includes students who passed sequentially through each stage of the College Pathways Model. Alternative Pathways includes students who graduated from college but did not follow through each stage of the College Pathways Model.

Immigrant generation First generation Second generation Third-and-higher  generation

27.2 13.4  8.4

19.8 25.5

22.9%

Via Linear Progression

35.2 18.7 11.0

26.9 33.2

Gender Male Female

Primary race-ethnicity White African American Amer. Ind. and   Pac. Islander Asian American Hispanic

30.4%

All students

Total

Percent College Graduates

Table 4.3    Percent of College Graduates Among UW-BHS High School Seniors via Linear Progression and Alternative Pathways

144   From High School to College

gender, race-ethnicity, and immigrant generation. Because of the small numbers of observations, this tabulation combines some of the smaller groups into combined categories, including American Indian and Pacific Islanders, Asian Americans, and all Hispanics (Mexican Americans and Other Hispanics). The first column in table 4.3 shows the percent of college graduates of the complete sample of UW-BHS high school seniors by gender, raceethnicity, and immigrant generation—the same figures as those presented in the first panel of table 4.1. The next two columns show two components of the overall college graduation rate: those who followed the linearprogression route and those who followed an alternative trajectory. In every case, the majority (typically about 70 to 75 percent) of college graduates followed the linear-transition sequence. Although it is possible to recover from low ambitions, lack of preparedness, and nonenrollment in a four-year college, the probabilities of recovery from setbacks are very low. Recall that the probabilities attached to the upward arrows in figure 4.2 were generally less than 10 percent. Although the majority of students follow a linear progression, there is considerable significance in the cumulative number of students who follow an alternative pathway to college graduation. In every group, about 3 to 9 percent of all high school seniors will graduate from college via an alternative pathway. The unconventional roads to college graduation could be labeled as “second chances”—pathways that allow for a recovery from a departure from the conventional model. How do they do it? In the right-hand panel of table 4.3, we classify the alternative-pathway students into three groups based on their enrollment status in the year after high school: (1) enrolled in a four-year college, (2) enrolled in a twoyear college, and (3) not enrolled. Most of the alternative-pathway graduates, upwards of 60 percent in most groups, were students in a two-year college right after high school. There are variations in the percentages by gender, race-ethnicity, and immigration generation, but the similarities are more striking than the differences. These figures help to explain a common misperception about the role of community colleges as an avenue to college graduation. The odds of eventually receiving a college degree among all students who begin their postsecondary studies at a two-year college are fairly low, much lower than students who begin at a four-year institution. Yet a substantial share of all college graduates in our study first enrolled in a two-year college. This apparent anomaly is explained by the very large number of students who enrolled in community colleges. The reality is that many ambitious students face real barriers to enrollment in a four-year college right after high school.24 The fact that some students do succeed in getting a college degree does not mean that it is an easy or a high-probability pathway.

The College Pathways Model   145

The rows in table 4.3 show the estimates for the percentage of UW-BHS high school seniors who followed the linear and alternative pathways to college graduation, by gender, race and ethnicity, and immigrant generation. The percentages of high school seniors who graduated from a four-year college varied widely across these three ascriptive groups. The figures in the first column of table 4.3 are the same as those of the last column of the first panel of table 4.1 (except for the differences in raceethnicity categories). The second column shows these patterns for students who followed the linear-progression model. The linear-progression model captures most—70 to 75 percent—of the actual college graduation rates. For the most part, this model works well; it captures most college graduates and most of the major lines of educational inequality. The alternative-trajectories component is, however, an important route to college graduation for all groups. In absolute terms, the groups that had the highest percent of college graduates via linear progression also had the highest percentages via alternative pathways; for example, females and whites. Disadvantaged groups (Pacific Islanders and American Indians, African Americans, and Hispanics) who had below-average levels via linear progression also had low levels via alternative pathways. This means that “second-chance” college graduates were more common among advantaged groups. There are a couple of exceptions to this pattern—namely, Asian Americans and first-generation immigrants, who showed a higher degree of resilience in bouncing back from adversity. The right-hand panel in table 4.3 shows the distribution of alternative pathways among college graduates (7.5 percent of all UW-BHS high school seniors) by their enrollment status in the year following high school. A small fraction—slightly more than 10 percent—were enrolled in a four-year college right after high school. These are students who took a detour earlier in the College Pathways Model but still managed to get back on track and graduate from college within seven years. These cases, however, are rare. The largest single component of graduates who took alternative pathways is students who enrolled in two-year colleges but still managed to receive a bachelor’s degree within seven years. This route was taken by 60 to 70 percent of alternative-pathway graduates (about 4 to 5 percent of all UW-BHS seniors). Finally, about 25 percent of alternative-pathway graduates (about 1 to 2 percent of UW-BHS seniors) were not enrolled at all in the year after high school. These students likely went to college after working or serving in the military for a year or two. These patterns are similar across all gender, race-ethnicity, and immigrantgeneration groups. Asian and first-generation students seem to have made better use of the community college route to a college degree than other groups.

146   From High School to College

Decomposition of the College Graduation Gaps by Gender, Race-Ethnicity, and Immigrant Generation by Stages of the College Pathways Model The method of demographic decomposition can estimate how much each stage of the College Pathways Model accounts for intergroup gaps (dis­ parities) in college graduation rates by gender, race-ethnicity, and immigrant generation (using the same methods that produced table 2.3 in chapter 2). The results of this exercise are presented in table 4.4, which summarizes the complex patterns of relative group differences in educationaltransition ratios. The first column of table 4.4 shows the actual intergroup (gender, race-ethnicity, and immigrant generation) gaps in the percentage of college graduates among all UW-BHS high school seniors. These figures are based on the sample of 8,719 UW-BHS high school seniors with nonmissing data for college aspirations, expectations, and preparation. We identify the gender gap (female versus male), four race-ethnicity gaps (white versus black, white versus Asian, white versus Hispanic, and white versus American Indian or Pacific Islander), and two immigrant-generation gaps (third-and-higher versus first, and third-and-higher versus second). With only one exception (third-and-higher generation versus second generation), all of these gaps are large enough to be statistically significant and substantively interesting. There is a wide range in the magnitude of inequality between groups. The actual (observed) gender gap in college graduation is only six percentage points, with female high school seniors more likely to graduate from college in seven years than their male counterparts (33.2 percent minus 26.9 percent from column 1 of table 4.3). Race-ethnicity gaps in college graduation are much larger, with white students about sixteen to seventeen percentage points more likely to receive a college degree than African American and Hispanic students, and twenty-four percentage points more likely to graduate from college than the combined American Indian and Pacific Islander population. There is a small gap between whites and Asians of about four percentage points that seems to be due to the underreporting of Asian college enrollments in the NSC data. There is a moderately large immigrant-generation gap of ten percentage points between first-generation students and third-and-highergeneration students. The columns of table 4.4 show the absolute and relative share of the total (observed) gaps in college graduation rates attributable to the five stages of the linear-progression version of the College Pathways Model and an additional column for the residual of alternative pathways.

5.70 90% 13.81 84% 5.41 126% 13.97 83% 18.77 78% 10.21 101% 1.57 162%

6.35 100%

16.49 100%

4.30 100%

16.86 100%

24.20 100%

10.09 100%

.97 100%

Total

.71 7% .30 31%

-1.16 -120%

2.78 11%

.88 9%

1.51 6%

1.78 11%

.67 15%

-1.86 -43% 2.10 12%

1.43 9%

.45 7%

Expect/ Aspire

.07 0%

1.05 16%

Aspire

-1.32 -136%

2.51 25%

2.32 10%

3.38 20%

-.87 -20%

3.00 18%

1.90 30%

Prepare/ Expect

2.31 238%

6.08 60%

4.79 20%

2.79 17%

5.75 134%

4.62 28%

.97 15%

Enroll/ Prepare

1.43 148%

.04 0%

6.98 29%

3.41 20%

1.73 40%

4.60 28%

1.32 21%

Complete/ Enroll

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Note: Based on the sample of UW-BHS high school seniors with nonmissing data on college aspirations, expectations, and preparedness.

Female-male gap    Absolute gap   Relative share White-black gap    Absolute gap   Relative share White-Asian gap    Absolute gap   Relative share White-Hispanic gap    Absolute gap   Relative share White-American Indian and   Pacific Islander gap    Absolute gap   Relative share Third-first generation gap    Absolute gap   Relative share Third-second generation gap    Absolute gap   Relative share

Total PercentagePoint Gap

Among Students Who Followed the Linear-Progression Pathway

Share of Total Gap Attributable to Differences in Educational Transitions in College Pathways Model

-.60 -62%

-.12 -1%

5.43 22%

2.89 17%

-1.11 -26%

2.68 16%

.65 10%

Alternative Pathways

Table 4.4    Decomposition of Intergroup (Gender, Race-Ethnicity, and Immigrant-Generation) Gaps in College Graduation Rates by Stages in the College Pathways Model

148   From High School to College

Every stage of the College Pathways Model, including alternative pathways, contributes something to the total gender gap of a little more than six percentage points. Recall that women had slightly higher college aspirations than men and were slightly more likely to translate these aspirations into expectations, expectations into preparedness, preparedness into enrollment at a four-year college, and enrollment into college completion (table 4.1). The decomposition results show the impact of each transition rate on the overall gender gap in college graduation. Men had slightly lower college aspirations and expectations than women, but attitudes only accounted for 1.5 percentage points of the 6-point gap. About two-thirds of the total male-female gap in college graduation (4.2 of the total 6.3 points) was due to behavioral differences—men were less likely to prepare for college while in high school, less likely to enroll in a four-year college even if they were prepared, and less likely to complete college among those who enrolled in a four-year college. These differences in educational transitions may stem from a common underlying difference. One frequently mentioned hypothesis in the literature is that there are small, but significant, differences between men and women in attentiveness and distractibility that advantage women in following through on the demands of formal schooling.25 Whatever the sources of contemporary gender differences in academic performance that favor females, they were masked in earlier times by societal practices (discrimination) that favored boys over girls. The gender gap is, however, small compared to racial-ethnic differences in college graduation. White high school seniors in the UW-BHS sample were sixteen percentage points more likely to graduate from college in seven years than their African American and Hispanic peers, twenty-four points more likely than American Indians and Pacific Islanders combined, and four points more likely than Asian students. There are some common threads of explanation for the much lower college graduation rates of the disadvantaged minorities. In general, the most important factors are lower college enrollment rates and lower college completion rates, followed by lowered college preparation rates in high school. These three factors account for 60 to 70 percent of the lower minority college graduation rates relative to white students. Low aspirations account for about 2 and 1.5 percentage points of the lower college graduation rates of Hispanics and the combined American Indian and Pacific Islander population, respectively, relative to whites, while black students had levels of college aspiration comparable to white students. The transition from college aspirations to expectations adds another 1.5 to 3 percentage points to the college graduation gap between whites and minorities. These are not trivial differences, but they are relatively modest. High college aspirations are ubiquitous in American adolescent culture, perhaps reflecting the ethos of the American dream and

The College Pathways Model   149

an awareness of the significance of a college degree for upward mobility. Disadvantaged minorities share these beliefs and understandings, but at a somewhat lower rate than white students. Getting on and staying on the college track in high school—taking the right courses, taking the SAT or ACT exam, and applying to a four-year college—pose slightly larger obstacles for disadvantaged minorities and account for about three percentage points of the overall gap. These data do not tell us how much the relative lack of college preparation on the part of minority students is due to student decisions or school practices. Perhaps capable minority students are less likely than white students to be urged to take advanced courses or reminded about deadlines by high school counselors. The last two factors in the College Pathways Model—enrolling in four-year college right after high school and persisting in college once enrolled—are the most important factors accounting for the gaps in college graduation between whites and disadvantaged minorities. The alternative-pathways component primarily reflects the trajectory of beginning at a two-year college, transferring to a four-year college, and completing a degree. Community colleges might be considered a back door to getting a college degree for those who lack the preparation and financial resources to enroll in a four-year college right after high school. It is interesting that the alternative-pathways component accounts for about three percentage points of the lower college graduation rates of African American and Hispanic students (relative to white students) and more than five percentage points of the lower graduation rate of American Indian and Pacific Islander students. The decomposition of the gap between Asians and whites in college graduation rates is complicated by measurement problems. In looking at national data in chapter 2, we found that native-born Asian Americans had closed the educational gap with white students in the 1950s and are now more likely to graduate from high school, more likely to enroll in college, and more likely to graduate from college than white students. However, among UW-BHS high school seniors matched with NSC enrollment records, whites are four percentage points more likely to graduate from college. The problem is that Asian college enrollment is underestimated in NSC records at about twice the level of other groups (see table 4.2). Despite this problem, the decomposition in table 4.4 provides an important insight into the Asian educational advantage in college graduation rates, namely that Asian students are successful in graduating from college primarily because of their experiences in high school. In high school, East Asian and Vietnamese students are much more likely than white students to be on the college-bound track, possessing higher college aspirations and expectations and better levels of preparation. While Cambodian students have lower levels on these transition rates, this is

150   From High School to College

entirely due to their lower socioeconomic background (see chapter 5). Filipino and other Asian students are about as likely as white students to be on the college-bound path in high school. Asian students are also advantaged relative to white students in terms of using two-year colleges as a stepping stone to a college degree. The decomposition of the gaps in college graduation by immigrant generation may also be affected by the selective undercoverage of college enrollment of the first generation in NSC records. Table 4.4 shows that college graduation rates are about the same for second-generation and third-and-higher-generation students, but first-generation students are about ten percentage points behind. Both first- and second-generation students appear to have a deficit relative to the third generation in the transition to college enrollment. The first generation has a bit of a newcomer disadvantage in terms of slightly lower values for college aspirations and college preparation. In contrast, the second generation has higher values on both dimensions. Why are Asians and immigrants able to take advantage of American high schools to get on the college-bound track compared to other minority students? This important question is addressed in subsequent chapters of this volume.

Conclusions In this chapter we identified five key educational transitions on the road to college graduation. These five transitions represent the College Pathways Model—college aspirations, college expectations, college preparation, college enrollment, and college completion. These steps are conceptualized and measured as sequential and cumulative steps that students must traverse on the path to college graduation. Let’s begin with the conclusions then reconsider the value and the limitations of models in (social) science and for the study of disparities in college graduation, in particular. Educational attainment, like many other social phenomena such as careers and family formation, is a temporal process that may extend over many years. However, research into these phenomena often looks at an overall summary of the process in question or perhaps at one stage along the way. This sometimes leads to disparate findings, as different factors may not be equally important at all states of the process. The problem is particularly acute for the study of education, especially if the reasons for student success at one stage are not the same at later stages. At first glance, the study of educational attainment appears to be logical and straightforward. The final level of educational attainment is simply the sum of a series of steps along a survival curve from primary school through secondary school to higher education. This cumulative structure of education follows the organizational nature of schools and

The College Pathways Model   151

the maturation of human beings from childhood to adolescence to adulthood. Perhaps, the symmetry of form and progression has deflected attention from understanding the relative significance of steps along the pathway to college graduation. Problems arise from every angle. Everything seems to affect students— home and family life, the nature of schools, teachers and peers, and even the unknowable potential of students to learn and perform well on periodic assessments. In addition, relationships can and do change over time, especially given the highly intercorrelated relationships among all these factors. The time dependency of schooling and the life course of students mean there will be substantial correlations from different stages of the process. For example, outcomes from first grade may be highly predictive of college success, but these associations may not be causal, as many aspects of early schooling may reflect antecedent conditions. Moreover, changes in family circumstances, neighborhood and schooling contexts, and individual development over the course of ten to fifteen years certainly filter the influences of early life-course events in a variety of ways. Perhaps it is not too surprising that these complex factors are reduced to a simpler narrative based on ideology, fragmentary knowledge, or the availability of data. Awareness of these issues and the potential risks associated with them does not guarantee more objective or cumulative research. The virtue of a model is that it makes explicit the essential argument as well as underlying assumptions. The College Pathways Model is very simple, but this is a necessity if questions are to be posed and empirically addressed. The five stages proposed here are largely borrowed from the prior research literature. For example, college aspirations and expectations have been a central feature of the life-course literature on educational attainment for decades. The orientations of students toward their educational futures is assumed to be the internalized motivation (or lack of one) from family socialization and the influences of significant others. Preparation in high school is a major determinant of post–high school educational outcomes and is influenced by social background, school practices, and how different groups are streamed and encouraged. The steps of college enrollment and college completion represent critical transitions that have attracted considerable interest from the policy world as well as academic researchers. In the balance of this volume, the College Pathways Model will be used to organize empirical analyses of the impacts of family SES, cultural forces, student employment, participation in extracurricular activities, and between-school differences on disparities in college graduation by gender, race-ethnicity, and immigrant generation. In addition to introducing our analytical model, this chapter provided an initial assessment of how much each stage of the model accounted for intergroup differences in

152   From High School to College

college graduation rates. Several interesting and important findings have emerged from this preliminary analysis. College aspirations, at least those expressed by high school seniors, are high, with three-fourths of students aspiring to graduate from college. Aspirations vary moderately by gender, race-ethnicity, and immigrant generation, but these variations only account for a small share (generally less than 10 percent) of between-group disparities in college graduation. College expectations are highly correlated with aspirations. Indeed, only about 11 percent of students who aspire to graduate from college do not expect that they will get the degree, and the drop-off from aspirations to expectations is associated with disadvantaged status. Among disadvantaged students (American Indian and Pacific Islanders, Hispanics, and African Americans) there is a much lower level of college preparation in high school relative to preparation among white students. In theory, high school resources, such as college preparatory courses and advice from counselors, are available to all students. Moreover, the fact that Asian American students and second-generation students are more college prepared than white students with comparable levels of college ambitions suggests that high school resources are probably more available to those who seek them out. The lower levels of college preparation among males (versus females) and first-generation students (relative to third-and-higher generation) indicate that student characteristics may be just as important as institutional factors. Lower levels of student achievement (GPA) may be an important intervening variable between social background characteristics and college readiness in high school. The results of the decomposition analysis show that the major reasons for lower college graduation rates among disadvantaged groups are low transition rates to four-year college enrollment, even among college-ready students, and lower college completion rates among students who enrolled in a four-year college. We also discovered that students who graduated college by a nonstandard path—primarily by first enrolling in a two-year college and transferring—are most common among advantaged groups. White students as a group were more likely than most other groups to find their way through an alternative pathway. We also found that Asians and first-generation immigrants were able to improve their chances to get a college degree through an alternative pathway. These issues are addressed in more depth in the successive chapters.

Chapter 5 Social Origins and College-Pathway Transitions

I

chapter, we move from description to explanation, with an analysis of the determinants of gender, racial-ethnic, and immigrantgeneration inequality for each of the five steps to college graduation in the College Pathways Model. The explanatory focus is on social origins—a broad concept that encompasses socioeconomic status (SES) and family background. We also consider the mediating roles of academic performance (GPA) and encouragement from significant others. GPA and encouragement have important direct effects on educational outcomes, but our focus here is on their indirect role as mediators of ascriptive inequality. The descriptive data, presented in prior chapters, revealed varied patterns of educational inequality by gender, race-ethnicity, and immigrant generation. About 33 percent of women in the UW-BHS sample of high school seniors received a bachelor’s degree, about six percentage points higher than their male peers—roughly the same gender gap that is found in national data. The gender gap favoring women in college graduation is a stunning reversal of the historical pattern. In contrast, there is an unsurprising familiarity to the persistent inequality in college graduation rates between whites and disadvantaged minority groups, including African Americans, Hispanics, American Indians, and Pacific Islanders. In chapter 4, we reported that more than one-third (35 percent) of white students in the UW-BHS sample received a BA or BS degree, compared to 19 percent of African American, 18 percent of Mexican American, 13 percent of American Indian, and only 9 percent of Pacific Islander youths. These patterns of racial and ethnic inequality in college graduation rates among students from the Pacific Northwest mirror those from national level data. According to the national data reported in chapter 2, college graduation rates for minority students rose, especially during the n this

153

154   From High School to College

era of educational expansion during the 1950s and 1960s. However, white and Asian American students made even more rapid progress, and racial and ethnic gaps in college graduation rates are now wider than ever (see figure 2.3). Asian Americans have achieved exceptional levels of education attainment—national data show that recent cohorts of native-born Asian Americans are twice as likely to have graduated from college as their white peers (see chapter 2). Our UW-BHS data, with more specific classifications among Asian Americans, show considerable heterogeneity across Asian ethnic groups. East Asian (Koreans, Chinese, and Japanese) and Vietnamese students have above-average levels of college graduation, while other Asian nationality groups fare less well. First-generation Americans—those who were born outside the United States—are much less likely to graduate from college than long-resident Americans. The second generation—native-born children of immigrants—has nearly caught up to third-and-higher-generation Americans. Because the UW-BHS data are based on high school seniors, our analysis does not include the very selective stream of Asian immigrants who were educated abroad or came to the United States to attend college or graduate school. In chapter 4 we postulated a sequential College Pathways Model, with college graduation as the outcome of a cumulative series of steps, beginning with college aspirations—the expressed desire among high school seniors to get a college degree. The second step is having expectations— aspirations tempered by reality as respondents consider whether they truly expect to graduate from college. Aspirations and expectations are attitudes, which are likely to guide behavior—expectations more so than aspirations. The third step is preparing for college in high school by taking AP courses, taking the SAT or ACT college entrance exams, and applying to a four-year college. The fourth and fifth steps extend beyond high school and are enrollment in a four-year college immediately after high school and, finally, college completion. College completion is measured as having received a bachelor’s degree within seven years of graduating high school. About one in three UW-BHS high school seniors graduated from college within seven years after high school graduation. The overwhelming majority (75 percent) of those who graduated from college followed the five sequential steps in the College Pathways Model of expressing college aspirations, expecting to realize their aspirations, preparing for college in high school, enrolling in a four-year college, and completing college within seven years. Since departure from this sequential pathway leads to a low probability of college graduation, most of our analysis focuses on these five steps that lead to college graduation. At the end of this chapter, however, we also consider the factors that predict success along the alternativepathways route.

College-Pathway Transitions  155

One of the major conclusions of the decomposition analysis in chapter 4 was that actions are more important than attitudes. Differential aspirations and expectations generally account for about 10 to 20 percent of the overall gap in college graduation rates between ascriptive groups. Attitudes do matter, but behavior matters more, especially being college prepared in high school, enrolling in a four-year college right after high school, and finally getting through college in a timely fashion. But what background factors explain the inequality in each of these transitions by ascriptive characteristics—gender, race-ethnicity, and immigrant generation? To address this question, we test the classic ideas from the educational stratification research literature, informed by the logic of the life course. Specifically, a series of multivariate equations are estimated to test how family background and differential resources might explain observed gender, race-ethnicity, and immigration generation disparities at each stage of the College Pathways Model—the conditional probabilities of aspirations, expectations, preparation, enrollment, and completion. Conditional probabilities are the individual-level expression of educationaltransition ratios, introduced in chapter 4. The likelihood of each transition is estimated for those in the prior state—those actually exposed to the risk of the transition. The introduction of independent (explanatory) variables in sequential and cumulative series of equations is based on the analytical framework sketched in figure 5.1. The logic, and many of the hypotheses embedded in figure 5.1, is drawn from the Wisconsin model, informed by subsequent studies of educational stratification.1 The dependent variables in the figure are the stages of the College Pathways Model—expressed as individual-level conditional probabilities akin to the aggregate educational-transition ratios. The analysis in this chapter is organized into five sections, paralleling the sequential stages leading to college graduation. We also add a sixth section to analyze the factors that lead to an alternative pathways route to college graduation.

Models of Educational Stratification There are two sets of exogenous variables in figure 5.1: the ascription variables of gender, race-ethnicity, and immigrant generation; and eight social-origins variables, which represent family background and SES. Between the sets of exogenous predictors and the college-pathways outcomes are two intervening variables: academic performance, as measured by self-reported GPA, and encouragement from family and friends. We do not make assumptions about causal priority between the two intervening variables—the double-headed arrow in the figure suggests that there are reciprocal influences.

156   From High School to College Figure 5.1    Hypothesized Model of Ascription, Social Origins, High School GPA, Encouragement, and College-Pathway Transitions Ascription Gender Race-ethnicity Immigrant generation

High school GPA College pathway Encouragement

Social origins Father and mother education Father and mother employment Father and mother SEI Home ownership Family structure

Source: Author’s compilation. Note: SEI = socioeconomic index.

This model reflects many of the major hypotheses of the Wisconsin model, segmented assimilation theory, and other theoretical perspectives in the educational-stratification literature. Our objective is not to present a new theory but to test empirically the major hypotheses that reappear in almost all theories of educational stratification. The central hypothesis tested in this chapter is that ascriptive inequality in college graduation rates is primarily a function of social class or SES. Closely related are the hypotheses that advantaged social groups support their children’s educational success through significant-other influence (SOI) or encouragement and by coaching them to study and work hard in school (the mediating role of GPA—the proxy for effort). This basic framework is elaborated in successive chapters to test additional hypotheses. We begin with an overview of the five models or equations that will be analyzed in this chapter.

Model 1: The Baseline Model The first model to be estimated only includes gender, race-ethnicity, and immigrant generation as independent variables. The coefficients show baseline differences (inequality) in each step of the College Path­ways

College-Pathway Transitions  157

Model. Since the outcomes are dichotomous variables, we estimate logistic regression equations that yield odds ratios (exponentiated logistic regression coefficients) for each independent variable. The odds ratios are relative to the omitted category of each independent variable: females relative to males, racial-ethnic minorities relative to whites, and first and second generation relative to the third-and-higher generation. The odds ratios from the baseline model are slightly different from the educational-transition ratios presented in chapter 4. First, intergroup differences are presented as odds ratios relative to one group (the omitted population). More importantly, the baseline logistic regression model includes all three ascriptive variables simultaneously, so that the effects are net of the other ascriptive variables in the baseline model. Since the gender composition is approximately the same for all race-ethnicity groups and immigrant-generation groups, there is little difference in the gender effects in a bivariate model or in the baseline model with all three ascriptive variables. However, immigrant-generation composition varies widely across racial and ethnic groups—most Asian Americans are first- and second-generation immigrants, while almost all African Americans and American Indians are third generation and higher. Hispanics have an intermediate immigrant-generation composition. Purging the race-ethnicity and immigrant-generation coefficients of each other allows for a clearer estimate of each dimension.

Model 2: The Role of Family SES and Family Structure The first “package” of covariates consists of family-background variables: parental education, employment, occupational attainment, home owner­ ship, and family structure. Although these variables represent multiple dimensions, they share a common thread of family SES and other attributes of social origins. There is a long-standing sociological hypothesis that much of the inequality by race and ethnicity reflects social class and the differential distribution of resources by family of origin.2 Accordingly, testing the social class hypothesis is a logical first step in our search for an explanation for the substantial variation in educational outcomes across race-ethnicity groups and immigrant generations.3 The social-origins hypothesis, however, is often much less relevant to gender inequality given that male and female respondents have the same family backgrounds. There are, however, slight differences in social origins by gender in the UW-BHS sample of high school seniors because of the underrepresentation of males in the senior class of high school. Since male students are slightly more likely than female students to have dropped out of high school, the UW-BHS male senior sample is more selective of higher-status families than is the female sample.4

158   From High School to College

Model 3: The Role of Academic Performance as a Mediating Variable The next model adds an index of academic performance—the student’s self-reported GPA in high school. As reported in chapter 3, GPA is measured by the response to the question, “In general, what grades do you get?”—with eight response categories ranging from “mostly As” to “mostly below Ds.” The ordinal-scale responses were transformed into a continuous variable with a range from 1.83 to 3.62 (see appendix table 3.A9 online).5 The GPA metric values were recalibrated to reflect the actual distribution of high school grades for students in each self-reported response category from the UW-BHS survey. As discussed in chapters 1 and 3, there are two standard interpretations of the high school GPA. The first is as a measure of an individual’s ability to learn, which has both an innate and an environmental component. There is substantial variance in the inherent ability to learn among students, some of which is heritable. The direct effect of GPA on educational outcomes reflects social influences as well as biological potential, since parents (and others) are more likely to reward high-ability children with more attention and stimulation. Nature and nurture are intertwined. All observations and measurement of academic performance, even during the preschool age range, reflect both innate and social influences. Our focus here is on the second interpretation of GPA—as a potential explanation of between-group differences. As noted earlier, there is no evidence that between-group differences in academic achievement are rooted in biological differences between social groups, defined by race and ethnicity, immigrant generation, or social origins. Although there is some speculation that there might be differences along gender lines in the propensity to learn, we precede with the assumption that differential socialization and other social influences are more consequential. Therefore, our approach is to analyze intergroup differences in educational outcomes that are mediated by GPA as a result of the social influences of families and communities that promote the educational success of their children. Parents, siblings, and other persons begin to train or coach their children from infancy onwards through play, talking and reading to them, and creating a stimulating social environment. These patterns certainly vary by SES and family background (social origins) and they are also likely to vary by ethnicity and immigrant generation. These social influences on children affect educational outcomes, largely through academic performance. Although there are certainly patterns of favoritism by teachers, grades are generally awarded to students based on their performance on exams and classroom exercises. The differences in academic performance among gender, race-ethnicity, and immigrant-generation

College-Pathway Transitions  159

groups are learned behavioral patterns. Some students work harder than others—they pay attention in class, do their homework, and make efficient use of time. The behaviors—labeled effort—are strongly influenced by social factors that vary between groups defined by social origins and ascription. Indeed, the primary objective of socialization in many families and cultures is to prepare students to do well in school. In addition to family members, friends and teachers play an important role in training students to expend more effort to get better grades in school. Based on this logic, we posit that GPA is endogenous to the ascriptive and social-origins variables in figure 5.1. This assumption allows us to estimate the shares of the total effects of ascription and social origins that are mediated by GPA—primarily because parents train their children’s ability to learn and succeed in school. It is possible that some fraction of the influence of family SES on GPA may be due to inherited ability, but the research literature suggests that this pathway is relatively small compared to socialization and other social influences.

Model 4: The Role of Encouragement by Significant Others as a Mediating Variable One of the most important findings of the proponents of the Wisconsin model of educational attainment is that much of the impact of family background on attainment is mediated through the influence of significant others.6 Family, friends, and teachers communicate powerful expectations to children and adolescent students through their words and actions. Encouragement is hypothesized to vary across families and communities, and different levels of importance are placed on achievement, specifically academic achievement.7 These norms may have roots in the cultures brought with recent immigrants or perhaps reflect long-standing religious and cultural divisions that are passed along from generation to generation. Although boys and girls share the same parents, they may differ in their socialization and family experiences. These gendered differences may produce variations in educational ambitions among daughters and sons from the same families.8 We label the influence of significant others as encouragement. The encouragement index is based on the sum of father’s, mother’s, siblings’, friends’, favorite teacher’s, and an adult mentor’s college expectations (“the most important thing for you to do after high school”). We do not posit causal priority between GPA and encouragement; we consider both to be endogenous to the ascriptive groups and family background. Our analytical approach is to measure the independent and shared explanatory power of GPA and encouragement on educational outcomes. Model 3 includes only GPA, model 4 includes only encouragement, and model 5 includes both.

160   From High School to College

Model 5: Saturated Model with Ascription, Social Origins, Academic Performance, and Encouragement by Significant Others While we consider academic performance and encouragement independently in models 3 and 4, they are, in fact, interdependent. Just as taller youth are more likely to be encouraged to play basketball, academically inclined students may receive more support from family, peers, and teachers to do well in school. However, encouragement and support may also lead to higher grades. The reasoning is that encouragement leads to increased motivation, which in turn leads to more effort and better grades. This pathway is theoretically important if academic achievement is rooted in racial-ethnic and immigrant-generation cultures that operate through family and peer encouragement.

Explaining Ascriptive Inequality in College Aspirations: The First Step of the College Pathways Model There is substantial literature on the educational aspirations of high school seniors, both as a major predictor of college enrollment and college completion9 and as an important phenomenon in its own right.10 During adolescence, the influences of family and schooling are solidified as students begin to develop a realistic sense of their future lives. There are many subsequent life-course events that will alter adolescent ambitions, but a snapshot of young adults and their educational goals just before leaving high school provides an initial benchmark of process of educational stratification. As noted in chapter 3, the question on college aspirations—“How far would you like to go in school?”—elicits overly optimistic perceptions of student prospects for higher education. More than seven out of ten seniors in the UW-BHS sample aspire to graduate from college. Differences in college aspirations played a significant, but relatively modest, role in explaining ascriptive inequality in college graduation (see table 4.4). Table 5.1 shows complete results for models 1 through 5 with coefficients for each independent and mediating variable on college aspirations. Given the detailed nature of these results, the tables for all other dependent variables in this chapter are only presented in online appendix tables.11 The results of table 5.1 are summarized in figure 5.2, which shows the effects of gender (females relative to males), race-ethnicity (each group relative to white), and immigrant generation (first and second generations relative to third-and-higher generation) for models 1, 2, and 5. The odds ratios (shown in vertical bars) in figure 5.2 and the

College-Pathway Transitions  161 Table 5.1    Logistic Regression of the Probability of College Aspirations for UW-BHS High School Seniors Exponentiated Odds Ratios Relative to the Omitted Category for Each Variable Ascriptive variables Gender (male omitted)  Female Race-ethnicity (white omitted)   African American   American Indian   East Asian  Cambodian  Vietnamese  Filipino   Other Asian   Pacific Islander  Mexican   Other Hispanic Immigrant generation   (third omitted)   First generation   Second generation Social-origins variables Father education (HS or less  omitted)   Some college, no degree   College degree or above   No father or DK Mother education (HS or less   is omitted)   Some college, no degree   College degree or above   No mother or DK Father SEI (not working, NA,   DK omitted)   Father employed   Father SEI (employed × SEI) Mother SEI (not working, NA,   DK omitted)   Mother employed   Mother SEI (employed × SEI) Home ownership   (renting omitted)   Family owns home

Model 1 Model 2

Model 3

Model 4

Model 5

1.25***

1.29***

1.01

0.97

0.82

0.93 0.60** 3.52*** 0.74 1.84** 1.33 1.16 0.67* 0.67** 0.81

1.39*** 0.97 3.40*** 1.47* 3.11*** 1.25 1.45 0.96 1.10 1.01

1.68*** 1.04 3.36*** 1.53* 2.42*** 1.34 1.48 1.13 1.29 1.19

1.19 0.96 2.85*** 1.01 1.97*** 1.05 1.22 0.79 0.93 0.89

1.41*** 1.02 2.92*** 1.09 1.70** 1.13 1.27 0.92 1.09 1.03

0.66*** 1.15

0.90 1.24**

0.85 1.22*

0.74** 1.23*

0.71** 1.21*

1.35*** 2.13*** 1.09

1.31*** 1.95*** 1.10

1.22** 1.81*** 1.19

1.20* 1.73*** 1.19

1.27*** 2.26*** 1.02

1.22** 2.09*** 1.07

1.21** 2.05*** 1.17

1.18 1.95*** 1.19

0.87 1.009***

0.90 1.008***

0.82 1.008***

0.85 1.007***

0.78* 1.007***

0.80 1.007***

0.81 1.006**

0.81 1.007**

1.19*

1.12

1.09

1.06

(Table continues on p. 162.)

162   From High School to College Table 5.1    (Continued) Exponentiated Odds Ratios Relative to the Omitted Category for Each Variable Model 1 Family structure (not intact omitted) Lives with biological or   adoptive parents GPA Encouragement index Constant McFadden’s adjusted   R-squared BIC index Number of observations

Model 2

Model 3

Model 4

Model 5

1.16*

1.07

1.03

0.97

0.03*** 0.145

1.60*** 0.18*** 0.199

2.70*** 1.53*** 0.01*** 0.226

8,386 8,719

7,873 8,719

7,617 8,719

3.53*** 2.85*** 0.015 9,545 8,719

0.74** 0.093 8,879 8,719

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample of 8,719 UW-BHS students includes all respondents with nonmissing data on college aspirations, college expectations, and college preparedness. DK = don’t know. NA = not available. BIC = Bayesian information criterion. SEI = socioeconomic index. GPA = grade point average. HS = high school. UW-BHS = University of Washington-Beyond High School project.

remaining figures in this chapter are benchmarked relative to the omitted categories (males, whites, third-and-higher generation), which are set at 1.0. The discussion and interpretation in the results in this chapter are based on the online tables as well as the summary graphs. The odds ratios in model 1 are similar to the bivariate percentages presented in chapter 4. The first coefficient in table 5.1 shows that women are 25 percent more likely than their male peers to aspire to graduate from college. Recall from table 4.1 that the observed gender difference in college aspirations was about four percentage points. This gap accounted for about one percentage point of the six-point gender gap in college graduation (table 4.4). The slight rise in the gender gap from model 1 to model 2 (from 1.25 to 1.29) is due to the male sample having a slightly higher SES composition than the female sample (because of the gender gap in dropout rates). One of the most interesting findings in table 5.1 is that the gender gap in college aspirations is no longer significant in model 3 (when GPA is added), model 4 (when encouragement is added), and model 5 (when both GPA and encouragement are included as covariates). Perhaps

1 2 5 Female

1.29 1.25

3.40

2.92

1 2 5 East Asian

3.52

1 2 5 Vietnamese

1.84

3.11

1 2 5 Filipino

1 2 5 Cambodian

1.47

1 2 5 Other Asian

1 2 5 American Indian

0.60

1 2 5 Pacific Islander

0.67

1 2 5 Mexican

0.67

1 2 5 Other Hispanic

0.71

1 2 5 1st generation

0.66

1.241.21

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 African American

1.391.41

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on table 5.1. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

Figure 5.2    Logistic Regression of College Aspirations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors

Regression coefficient

164   From High School to College

the families, friends, and social networks of women are more nurturing, positive, and supportive of attending college than those of men. Alternatively, females may receive more encouragement because they tend to do better in school (higher GPA). Both encouragement and GPA have highly significant direct effects on aspirations to graduate from college. The odds ratio for encouragement in model 5 is approximately the same as in model 4, but the GPA coefficient is reduced by about 25 percent from model 3 to model 5. These results suggest that part of the GPA effect on gender inequality in college aspirations is mediated by encouragement. The female edge in college aspirations is probably due to both factors: females are doing better in high school and also receive more encouragement, independent of their grades. In contrast to the modest difference in educational aspirations between men and women, there are much larger racial and ethnic differentials. Almost all analysts emphasize the role of differential social origins (social class) as a key reason for the lower educational attainments of minorities and new immigrants.12 Recall that the odds ratios in table 5.1 are expressed relative to the reference categories (set to 1.0) of white students and third-and-higher-generation students for minorities and immigrants, respectively. Racial and ethnic differences in college aspirations in the baseline model (model 1) as shown in table 5.1 mirror the descriptive bivariate patterns from chapter 4. There are no significant differences in the likelihood of college aspirations between white, African American, Cambodian, Filipino, Other Asian, and Other Hispanic high school seniors. East Asians and Vietnamese have higher levels of college aspirations than white students. However, there are several minority groups—American Indians, Pacific Islanders, and Mexicans—that report significantly lower aspirations to complete college than their white peers. The decomposition analysis in chapter 4 showed that differences in college aspirations play a significant, but relatively small, role in accounting for the lower college graduation rates of American Indians, Pacific Islanders, and Hispanics. Model 2, which includes five family-SES and background variables as covariates, shows that the lower college aspirations of disadvantaged race and ethnic groups are, in fact, a function of differential social origins. One of the most important findings is that the observed parity in college aspirations between white and black students in model 1 is a function of the counterbalancing forces of relatively poorer family backgrounds and much higher underlying college aspirations among black students. If we hold the variables of family SES and family structure constant, black students are 39 percent more likely to aspire to complete college than white students (see model 2). These high ambitions

College-Pathway Transitions  165

are “suppressed” in model 1 because black students are more likely to come from homes with lower parental SES and a lower prevalence of two-parent families. There is a widely discussed claim, expressed in the writings of John Ogbu and his colleagues, that the descendants of involuntary immigrants, African Americans in particular, are less interested in social mobility and have a lower commitment to schooling than students from the majority population.13 According to this interpretation, minority students discourage their high-achieving peers by labeling studying and doing well in school as “acting white.”14 Not only does our research find the opposite—black students, after adjusting for familial resources, have higher “underlying” college aspirations than white students—but we find that these patterns are largely explained by more encouragement to attend college among African Americans (see the decline in the African American positive coefficient from model 2 to model 4 in table 5.1). Our findings are consistent with a larger body of research, which finds that black students have high educational ambitions and a high regard for successful students.15 Higher college aspirations among African American students are evident, even though they have, on average, lower grades (GPA) than white students. Holding all other measures constant (social origins, encouragement, and GPA), black students are 40 percent more likely to aspire to graduate from college than are white students (model 5). These findings, however, need to be seen in a broader context that is somewhat less optimistic. First, the UW-BHS sample consists of high school seniors. Because a higher proportion of black students than white students drop out of high school before their senior year, the UW-BHS sample of black students may be more selective of black (than white) students on characteristics that predict academic success. Second, the higher level of college aspirations among black youth in model 2 is net of social origins (SES and family structure). However, in the actual population (model 1), black students face considerable barriers to higher educational aspirations due to their impoverished family backgrounds, living in single-parent families, and poorer academic performance. In spite of these qualifications, we find that high college aspirations provide considerable evidence of resiliency among black students. The observed lower levels of college aspirations among other dis­ advantaged groups—American Indians, Pacific Islanders, and Mexicans— are completely eliminated in model 2 when we adjust for social origins (family background). As noted, this does not mean that the low college aspirations displayed by these groups in model 1 are not real; they are, however, rooted in their impoverished family backgrounds, not in the

166   From High School to College

cultures and orientations of minority groups. It seems that beliefs in the American dream of social mobility and high college aspirations are widespread among American youth of all ethnic backgrounds with similar socioeconomic backgrounds. There is a high odds ratio for East Asians, and to a lesser extent for Vietnamese, in model 1. The inclusion of family-background variables in model 2 alters, only slightly, East Asian college aspirations, but it boosts the aspirations of Vietnamese youth to approximately the same level of East Asian students. With family background and SES held constant in model 2, there is a statistically significant positive effect of Cambodian ethnic origins on college aspirations. The Cambodian students in the UW-BHS sample are primarily the children of impoverished refugee families with low parental educational levels (see appendix table 3.A6, available online).16 Despite wide variations in immigrant generation and family social origins, almost all Asian American national origin groups exhibit extraordinarily high college aspirations. High aspirations are, however, only the first step in the long journey. Other transitions will prove to be more consequential in explaining inequality in college graduation. The high college aspirations of Asian students appear to be partially due to above-average levels of encouragement from significant others (compare coefficients in model 2 and model 4). Asian American students are more likely than their peers to report that their families and friends think that going to college is the most important thing to do after high school. Except for Vietnamese, there is no evidence that higher GPA is an important determinant of Asian American college aspirations (compare model 2 with model 3 and model 4 with model 5). The expression of support for their children’s education by Asian American parents, especially by working-class immigrants, may not follow the conventional middleclass-white model of frequent conversation and unconditional love. Some research suggests that Asian families put high pressures on their children to succeed, which may lead to high ambitions, but this also leads to heightened levels of depression and anxiety.17 The interpretation of the differences in educational aspirations by immigrant generation is very similar to the prior discussion of racial and ethnic groups. The lower college aspirations of first-generation students are primarily a function of family background (see the shift from model 1 to model 2). There is a new twist, however; the gap in college aspirations between the first generation and third-and-higher generation widens from model 2 to model 4. The first generation is “boosted” because of the encouragement from significant others. Holding encouragement constant in model 4 shows that the first generation actually has low college aspirations, net of all other variables. It may be that recent arrivals to the United States have not quite been socialized in the American “college for all” culture. In the observed population, this net negative effect is cancelled out

College-Pathway Transitions  167

by the very positive encouragement of family and friends experienced by first-generation adolescents. The second generation, in contrast to the first generation, has higher college aspirations than their third-and-higher-generation peers, net of social origins. Since first- and second-generation students come from similar households (children of immigrants), the difference in their college aspirations may simply be due to the increased exposure to American culture, facilitated by the language fluency and social contacts that come with longer residence in the United States. The positive second-generation effect on college aspirations is consistent with the Asian American pattern. American-born children of immigrant parents are slightly more educationally ambitious than third-and-highergeneration Americans. The magnitude of this pattern is small, but it is consistent with many studies on immigrant optimism and secondgeneration advantage.18 Our primary goal in this chapter is to describe gender, racial-ethnic, and immigrant-generation disparities in the transitions of the College Pathways Model and to explain these differences. To summarize the findings for college aspirations: underachieving minorities, including firstgeneration immigrants, are primarily handicapped by their lower social origins (relative to white students). This pattern is sometimes labeled as the “inheritance of poverty.”19 However, minorities report more encouragement from significant others than do white students, which compensates (at least partially) for their impoverished family background. Some of the higher levels of encouragement are linked to doing well in school (higher GPA), but most of it is independent of academic performance. Males have slightly lower college aspirations than their female peers. The lower aspirations of males are completely explained by the higher GPA and encouragement reported by female students (relative to males). Almost all of the social-origins variables—parental education status, parental employment, parental occupational status, home ownership, and family structure—have significant total effects on college aspirations (model 2), with parental socio­economic variables (educational and occupational status) having the strongest impact. The effects of family SES on college aspirations are largely direct, while the smaller effects of family structure and homeownership are mediated by encouragement and GPA. Independently of the social-origins and ascription variables, encouragement and GPA have huge direct effects on college aspirations. The incremental explanation of college aspirations by GPA (measured by McFadden’s adjusted R-squared) is 5.2 points (model 3 compared to model 2), but only 2.7 points is net of encouragement (model 5 compared to model 4). The total impact of encouragement is much larger, 9.6 points (model 4 compared to model 2), and 8.1 points is independent of GPA (model 5 compared to model 4).

168   From High School to College

Explaining Ascriptive Inequality in College Expectations: The Second Step of the College Pathways Model The next step on the pathway to college graduation is student expectations to graduate from college. The difference in the survey questions that measure college aspirations (“How far would you like to go in school?”) and expectations (“Realistically speaking, how far do you think you will get in school?”) is small. However, there is a wide gap in the meaning of the two concepts. When asked about expectations, respondents are asked to consider the world they actually live in, rather than the world they might wish for. The result is a small but significant slide down the scale from dreams to reality. About three-fourths of UW-BHS students aspire to graduate from college, but only two-thirds actually expect to do so (table 4.1). In terms of educational-transition ratios (or conditional probabilities), only 90 percent of those with aspirations to complete college expect to do so (table 4.1). Social science theory suggests that the translation of aspirations into expectations is a function of two factors: awareness of barriers that might obstruct college completion and confidence that barriers can be overcome. The costs of higher education present a major barrier for many low-income students. Many students, especially those from minority and working-class backgrounds may fear that enrolling in a major university far from home will put them in an unwelcoming environment where they may have difficulty finding friends and social support. Many students with weak high school grades may worry about their ability to meet the academic challenges of completing college. Self-confidence plays a role in shaping expectations that one will be able to overcome the internal and external hurdles in the transition from high school to college graduation. The pattern of odds ratios of college expectations (conditional on college aspirations) in the baseline model in figure 5.3 (and in appendix table 5.A1, available online)20 is generally similar to the odds ratios for college aspirations. Female high school seniors are more likely than males to expect to complete college, given aspirations, just as they were more likely to aspire to complete college. The absolute difference is relatively small but statistically significant. The modest difference in social origins between men and women (as discussed earlier) means that the net gender gap actually widens a bit from model 1 to model 2. The disadvantages of underachieving racial and ethnic groups relative to whites are larger in magnitude and more pervasive for college expectations than college aspirations. The Asian American advantage is limited to East Asians and is smaller in magnitude. First-generation youths have significantly lower expectations of completing college than

1 2 5 Female

1.231.27

1 2 5 African American

0.53

1 2 5 American Indian

0.36

1 2 5 East Asian

1.44

1 2 5 1 2 5 Cambodian Vietnamese

0.37

2.51

1 2 5 Other Asian

1 2 5 Pacific Islander

0.27

0.36 0.41

0.69

1 2 5 Mexican

0.41

1 2 5 Other Hispanic

1 2 5 1st generation

0.75

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.64

1.11

1 2 5 2nd generation

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A1. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 5.3    Logistic Regression of College Expectations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Aspiring UW-BHS High School Seniors

Regression coefficient

170   From High School to College

third-and-higher-generation youths, while second-generation students have neither a deficit nor an advantage in expectations to complete college. Most, but not all, of the lower expectations of disadvantaged minorities are due to differences in social origins—that is, poorer socioeconomic conditions and less family stability. These findings provide support for Stephen Morgan’s claim that the educational expectations of minority groups differ from aspirations. 21 Minorities may wish to complete college—they share in the belief in the American dream of getting ahead via education. But when asked if it will “realistically” happen, minorities begin to scale back their dreams. The lower level of confidence displayed by minority youth is, by and large, rooted in socioeconomic realities, not in ethnic cultures. In model 2, with family background included, most of the white-minority gaps are substantially reduced and are no longer statistically significant. For African Americans, American Indians, Cambodians, Filipinos, and Mexicans, there is a widening of the gap in college expectations (relative to whites) from model 2 to model 4. The differences are modest and not always significant, but they suggest that the expectations (confidence) of minorities to go to college would be considerably lower were it not for the above-average levels of encouragement from family and peers. High school GPA lowers the college expectations of minority youth (see change in odds ratios from model 4 to model 5 in online appendix table 5.A1), but this is partially counterbalanced by ethnic differences in positive encouragement to attend college. Asian Americans have higher college aspirations than white students, especially net of social origins. However, the translation of these high aspirations into expectations is limited to East Asians (and Vietnamese, after adjusting for social origins). The confidence among East Asian and Vietnamese students that their college aspirations will be realized remains in the odds ratios of subsequent models, but the coefficients slip back and forth across statistical significance in subsequent models. Immigrant generation is only slightly associated with college expectations. The first generation has lower college expectations, but this deficit is entirely mediated by lower social origins in model 2. Pacific Islanders have lower college aspirations than whites, but it is entirely a function of their lower SES. However, their low expectations cannot be explained by any of the covariates in these models. For example, in the baseline model 1 (with only gender and immigration generation as covariates), Pacific Islanders were 73 percent less likely than their white peers to expect to graduate from college among those who aspired to do so. In model 5, with all covariates, including social origins, GPA, and encouragement, held constant, this figure is still 59 percent below that of their white peers. The situation of Pacific Islanders is extreme, but all minority groups have lower expectations than aspirations (relative

College-Pathway Transitions  171

to white students). Even among the overachieving Asian groups (East Asian and Vietnamese), their edge (relative to white students) is less for expectations than aspirations. It seems that all students, net of their socioeconomic backgrounds, share the desire to obtain a college degree—buying into the abstract ideal of the American Dream. However, the follow-up question concerning expectations is a reality check. Minorities (with the exception of Asian Americans) are more likely than white students to have doubts that they will graduate from college.

Explaining Ascriptive Inequality in College Preparation: The Third Step of the College Pathways Model The third step in the College Pathways Model is preparation for college while in high school. Our index of college preparedness is based on responses to three questions on whether the student is taking AP courses, taking college entrance exams (SAT or ACT), and applying to four-year colleges by the spring of the senior year of high school. Students who performed at least two of these three actions were coded “1” (college prepared), and other students were coded “0.” We assume that these behaviors are also indicators of other actions that would make a student college ready, such as taking a rigorous curriculum of college prep courses. Following the logic of prior analyses, figure 5.4 (and online appendix table 5.A2)22 shows the logistic regression equations predicting the conditional probability from college expectations to college preparation. The sample for the transition to college preparation is limited to the subset of students who aspire and expect to graduate from college. Female students are 31 percent more likely to be college prepared than are male students, among those expecting to get a college degree. The gender gap rises to 35 percent in model 2, which is net of social origins. The female advantage along the College Pathways Model is significant at each stage of the process. The gender gap in college preparation is “explained” in model 3, which includes GPA (academic performance) as a covariate. Men receive lower grades than women, even among those expecting to graduate from college. But among men and women with comparable grades, there is no gender difference in college preparation. There is no evidence that young women are any less likely to take rigorous high school courses than their male peers.23 Females are slightly more likely to report encouragement to attend college, but gender differences in encouragement do not make a statistically significant contribution to the gender differences in college preparation (compare models 2 and 4 in appendix table 5.A2). East Asian students who expect to go college are more college prepared than their white peers, and white students are more college prepared

1 2 5 Female

1.311.35

1 2 5 African American

0.65

1.45

1 2 5 American Indian

11

0.53

3.13

3.28

1 2 5 East Asian

2.97

1.85

1 2 5 1 2 5 Cambodian Vietnamese

1.64 1.65

1 2 5 Other Asian

1 2 5 Pacific Islander

1 2 5 Mexican

1 2 5 Other Hispanic

0.73

0.63

1 2 5 1st generation

0.56

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.60

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A2. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

Figure 5.4    Logistic Regression of College Preparedness on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Expecting UW-BHS High School Seniors

Regression coefficient

College-Pathway Transitions  173

than historically disadvantaged minorities (African American, American Indian, and Mexican American). First-generation students are significantly less likely to be college prepared than third-and-higher-generation students. In patterns very similar to those for college aspirations, the lower levels of college preparation among disadvantaged minorities are entirely a function of differential social origins (compare race and ethnic groups from model 1 to model 2). What are often perceived as racial-ethnic differences in responding to opportunities to get on the college track in high school are actually differences in social class and family background. The only group with a statistically significant net below-average rate of college preparation in model 2 is first-generation students, which might be explained by their recent arrival. The advantage of East Asians in college preparation in model 1 broadens to include Cambodians and Vietnamese in model 2. There are also movements in the same direction among Filipinos and other Asians, though the gaps are not statistically significant. In other words, holding SES constant, all Asian groups appear to be more college prepared than their white peers. Some of the Asian edge is due to higher levels of encouragement (see change from model 2 to model 4). Asian students are much more likely to be first and second generation, but the positive Asian effect appears to be largely unrelated to immigrant generation. First-generation students are much less likely to be college prepared and, surprisingly, this pattern is largely unrelated to any other measured covariate. This is in contrast to college aspirations and expectations. The disadvantage displayed by first-generation youth was attributable to their lower levels of social origins. Second-generation students are not statistically different from third-and-higher-generation students in their propensity to be college prepared. There is no evidence that any minority group is less likely to be college prepared, net of social origins. Although high school grades (GPA) have a huge direct impact on college preparation and do mediate the female edge, GPA does not mediate the Asian advantage, except for Vietnamese. As noted earlier, African American students are less likely to be college prepared than white students due to their lower SES. However, model 3 shows that African Americans are actually more likely to be college prepared net of GPA. The low GPA of black students is not explained by their social origins. The covariates in these models do not exhaust all potential explanations for differential college preparedness. GPA and encouragement have strong direct effects on preparedness, but they do not mediate the impact of social origins. GPA does explain (mediate) the advantage of two ascriptive groups—females and Vietnamese—but other gaps remain in model 5. An important hypothesis is that some ethnic and community cultures “push” students by expressing high levels of encouragement. There seems

174   From High School to College

to be some support for this hypothesis, especially among Cambodians. The net effect of Asian overrepresentation in the final saturated model indicates that other forces might be at work—perhaps teachers and peers pull many Asian students into college-preparation activities regardless of their backgrounds or grades.24 Teachers and families may also be providing some support for the relatively small number of promising African American students to get ready for college, even if their grades are not the highest. Another important conclusion from appendix table 5.A2 is the strong and pervasive direct influence of social origins (family background) on college preparedness. Consistent with the findings for college aspirations and expectations, students from advantaged families are more likely to be college prepared in high school than their peers from disadvantaged homes. In most cases, the positive effects of family background are direct and not mediated by academic performance or encouragement. Most importantly, we find that gaps in college aspirations, expectations, and preparation between whites and non-Asian minority groups are primarily due to disparities in social origins. Encouragement does have a modest net (of social origins and GPA) impact on college preparedness—there is an increment of 1.1 points in pseudo R-squared from model 3 to model 5 in appendix table 3.A3 (online).25 On the other hand, academic performance (GPA) has a huge impact on college preparedness—as evident in the jump of 9.1 points in pseudo R-squared from model 2 to model 3. This is probably an indicator of the direct effect of ability on educational outcomes that is largely uncorrelated with any other measured variable.

Explaining Ascriptive Inequality in College Enrollment: The Fourth Step of the College Pathways Model Enrollment in a four-year college upon high school graduation is a key step along the pathway to college graduation. Almost three-quarters of college-prepared UW-BHS high school seniors who enrolled in a fouryear college immediately after high school received a college degree, compared to 10 percent of those who did not (figure 4.2). Being prepared to attend a four-year college does not guarantee that a student actually enrolls in one—fewer than two out of three college-prepared students enroll in a four-year college within a year of graduation. Differences in rates of college enrollment, net of prior steps in the College Pathways Model, is one of the most important factors that contributes to the whiteblack gap and the white-Asian gaps in college graduation. It is also a key reason for inequality in college graduation rates by immigrant generation.

College-Pathway Transitions  175

These figures on the transition to a four-year college are based on the match of UW-BHS seniors with records on college enrollment from the NSC. However, as noted in chapter 4, the measurement of the transition to enrollment in college was underestimated in NSC records. As a check on the validity of the NSC-based estimates of college enrollment, we created an alternative transition rate to enrollment in a four-year college based on the combination of NSC records and the UW-BHS oneyear follow-up surveys. The alternative index yielded an enrollment rate among UW-BHS seniors that was nine percentage points higher (and an educational-transition ratio from college preparedness to college enrollment fifteen percentage points higher) than from the NSC data alone (see table 4.2). Despite the differences in magnitudes, the patterns in the data sets were similar by gender, race-ethnicity, and immigrant-generation patterns with one major exception. The college enrollments for Asian American students, especially East Asian students, are grossly under­ estimated in NSC records. As a check on the robustness of our findings, we replicate the analysis of the transition from college preparedness to four-year college enrollment in figure 5.5 (based on NSC alone) and figure 5.6 (based on the combined NSC and UW-BHS data). More detailed results are presented in appendix tables 5.A3 and 5.A4 (available online).26 Net of social origins, the gender differential in college enrollment (among college-prepared students) is only about four percentage points (table 4.1) or about 16 percent to 24 percent when expressed as odds ratios (figures 5.5 and 5.6). As with earlier stages in the College Pathways Model, the female advantage in college enrollment is mediated by the fact that females are more academically successful in high school, as indicated by their GPA. Differential encouragement is also related to the gender gap in the transition to college enrollment, but its impact is small compared to GPA. (These more detailed findings are based on appendix tables 5.A3 and 5.A4.) Figures 5.5 (based on NSC data alone) and 5.6 (based on combining NSC and UW-BHS estimates of college enrollment) show somewhat different patterns of racial and ethnic inequality in the transition from college preparedness to college enrollment. The baseline model shows that college-prepared minority groups (including all Asian populations) and first-generation immigrants are 30 to 50 percent less likely than white students to enroll in a four-year college right after high school. Figure 5.6, with the expanded definition of college enrollment, confirms the basic story of minority disadvantage, but with two major exceptions—East Asians and Vietnamese are no longer disadvantaged. These are the two groups whose college enrollment was underestimated in NSC records along with low-SES groups.27 Not all of these minority differences in college enrollment are statistically significant, in part because of the small sample sizes for some populations. Each stage of the College Pathways Model filters out those

1 2 5 Female

1.16

0.63

0.72

1 2 5 African American

0.53

1 2 5 American Indian

0.52

0.64 0.61

1 2 5 East Asian

0.62

1 2 5 1 2 5 Cambodian Vietnamese

1 2 5 Other Asian

0.57

1 2 5 Pacific Islander

0.52

1 2 5 Mexican

1 2 5 Other Hispanic

0.60

0.57

1 2 5 1st generation

0.54

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.43

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A3. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 5.5    Logistic Regression of Enrollment in a Four-Year College (NSC Only) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors

Regression coefficient

1.24

1 2 5 Female

1.20

0.62

1 2 5 African American

0.45

1 2 5 American Indian

0.48

1.47

1 2 5 East Asian

1.45

1 2 5 1 2 5 Cambodian Vietnamese

0.54

0.60 0.55

1 2 5 Other Asian

1 2 5 Pacific Islander

0.37

0.52

1 2 5 Mexican

1 2 5 Other Hispanic

0.70 0.63

1 2 5 1st generation

0.61

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.59

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A3. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

Figure 5.6    Logistic Regression of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors

Regression coefficient

178   From High School to College

who dropped off the college-bound pathway or the linear-progression assumption. Only 4,033 of the original UW-BHS sample of 8,719 students “survived” the sequential filters of college aspirations, expectations, and preparation to be at risk (eligible) for the transition from college preparedness to college enrollment. For some of the minority groups, the sample size of UW-BHS students at risk of college enrollment is fewer than one hundred students. There is an important difference in the net effects of social origins on college enrollment in model 2 of the online appendix tables 5.A3 and 5.A4. In contrast to the strong direct effects of social origins on college aspirations, expectations, and preparedness, social origins do not have direct effects on the transition to college enrollment in appendix table 5.A3 (based on the NSC data alone). However, this finding is not supported with the broader measure of college enrollment in appendix table 5.A4 (based on both NSC and UW-BHS data). We think that the broader measure of college enrollment in appendix table 5.A4 offers a more credible estimate of the significance of social origins. Although the direct effects of social origins on college enrollment are understated in the NSC data, the mediating effect of social origins on disparities in college enrollment by race-ethnicity and immigrant generation is roughly similar in both appendix tables. These findings suggest that social origins continue to be significant, but their influence is stronger on outcomes measured in high school than the subsequent ones in the College Pathways Model. For several minority groups, there is a disadvantage (relative to white students) in making the transition from being college prepared to enrollment in a four-year college, even when SES and family background are held constant (see model 2 in figures 5.5 and 5.6). Social origins do play an important role in depressing college enrollment rates of disadvantaged minorities (compare model 1 and model 2), but the gaps in college enrollment remain after the social origins are included as covariates in the model. The lower transition rates to college, net of social origins, remain statistically significant for some groups, including African Americans, Filipinos, and Pacific Islanders. The overall pattern is similar for other minority groups. In contrast to college aspirations and preparation— where social origins fully explained the barriers faced by disadvantaged minorities—college enrollment is a much steeper climb for minorities. Encouragement and GPA have statistically significant direct effects on four-year college enrollment. However, the total impacts of encouragement and GPA are somewhat less for the transition to four-year college enrollment than for the first three stages of the College Pathways Model. Lower high school grades are part of the reason for the lower rates of transition to college for college-ready African American and Pacific Islander students, but social origins and other barriers are also important. Recall

College-Pathway Transitions  179

our prior interpretation that differences in GPA between ethnic groups are not inherited but more likely the result of differential family coaching and training in school-specific skills. The success of East Asians and Vietnamese in transitioning to college might be due to encouragement and grades, but the evidence for this is weak. The low college-transition rate of college-prepared Filipino students is a bit of a surprise, and none of the variables included here offer any clue to explain it. The relationship between immigrant generation and four-year college enrollment shows a familiar pattern. As with the transition to college preparedness, first-generation students who are college ready are less likely to enroll in a four-year college than comparably prepared third-andhigher-generation students. This difference is partially attributable to the lower social origins of first-generation students. There is no difference between second- and third-and-higher-generation students.

Explaining Ascriptive Inequality in College Completion: The Fifth Step of the College Pathways Model Completing college, even for those with high ambitions and solid preparation, is a hard slog.28 There is a minimum of four years of course taking, exams, term papers, and other requirements. Expectations are considerably higher in college than in secondary school, and requirements are less flexible. In addition to the academic hurdles, many students from disadvantaged backgrounds have to face high (and rising) costs of college. Tuition costs have more than doubled (in constant dollars) from 1980 to 2010, which puts increased pressures on students of modest means.29 Many college students have to juggle work and familial commitments along with their studies, which often delays college graduation. For example, the average time to a college degree (for students who do not take time off) is fifty-five months, with a higher average time to a degree among transfer students and non-full-time enrollees.30 Here we measure college completion as the proportion of college entrants (in a four-year college) who receive a BA or BS degree in seven years. Less than 50 percent of four-year college entrants receive a college degree within four years of completing high school, but this figure rises to 75 percent after seven years. These college completion rates are similar to national trends in college completion.31 Our analysis relies on NSC-based estimates of college enrollment and college completion. The underestimation of college enrollment in NSC records should not affect the rate of college completion (transition from four-year college enrollment to college graduation), since both the numerators and denominators are from the same source. As noted in chapter 4, there are wide disparities in rates of seven-year college completion (hereafter

180   From High School to College

referred to as college completion) among students who entered a four-year college right after high school. These disparities in college completion among four-year-college entrants account for one-quarter to one-third of the gender and racial-ethnic gaps in college completion among the UW-BHS universe of high school seniors (see table 4.4). In figure 5.7 (and in online appendix table 5.A5),32 we search for potential clues to the reasons for the lower college-completion rates among disadvantaged groups. The determinants of college completion in figure 5.7 (and in appendix table 5.A5) look more similar to the patterns predicting college enrollment in figure 5.6 than to the earlier steps in the College Pathways Model. The most obvious similarity is that the widespread minority disadvantage in the transition from college entry to college completion cannot be explained by differential social origins. Much like the transition to fouryear college enrollment, only some of the effects of a student’s social origins are statistically significant—in this case, the mother’s college degree and father’s occupational status (see appendix table 5.A5). Our interpretation is that the effects of family and other background characteristics are more influential when adolescents are still in high school. When students leave home, the influence of their social origins remains, but the impact is somewhat less. It should also be noted that the much lower sample sizes at this point in the College Pathways Model also lowers the likelihood of statistical significance. The gender gap in college completion is relatively modest—five percentage points in table 4.1 and a 32-percent difference in the odds ratio in model 2 in figure 5.7. Although these magnitudes are modest in absolute terms, the gender gap in college completion is consequential in explaining the overall gender gap in college-graduation rates (21 percent of the observed gap; see table 4.4). Because more women enroll in college than men (following the pattern at every stage of the College Pathways Model), we might expect that the smaller male sample would be more selective at this stage than the female counterpoint. However, the male-female gap in college completion is largely explained by the female edge in high school GPA among four-year college entrants. The odds ratios show that minorities, in general, are less likely to complete college than whites. However, not all of the coefficients are statistically significant, perhaps because of small sample sizes. The samples are smaller because the college completion analysis is limited to college entrants. African American, American Indian, Cambodian, Filipino, and Hispanic college entrants are about 40 percent to 60 percent less likely to earn a college degree in seven years than are white college entrants. Pacific Islander college entrants are about 80 percent less likely to complete a degree than comparable white students. As with college enrollment (but not with race-ethnicity differences in college aspirations, expectations, and preparation), the lower rates of college completion among minorities are

1.32

1 2 5 Female

1.28

0.60 46

1 2 5 African American

0.44

1 2 5 American Indian

0.42

0.68 0.67

1 2 5 East Asian

0.66

1 2 5 1 2 5 Cambodian Vietnamese

2.55

0.49

1 2 5 Other Asian

0.280.30

1 2 5 Pacific Islander

0.19

1 2 5 Mexican

0.52

1 2 5 Other Hispanic

0.460.50

1 2 5 1st generation

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.53

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A5. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 5.7    Logistic Regression of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Entrant UW-BHS High School Seniors

Regression coefficient

182   From High School to College

not completely due to differences in family background and SES. There is a modest narrowing of the gap from model 1 to model 2 (which includes the social-origins variables) and, in some instances (for example, American Indian and Filipino) the differences are no longer statistically significant. However, these changes are relatively small in absolute terms. African Americans have one of the lowest college-completion rates (or highest college dropout rates). Black students who survived the gauntlet from college aspirations to enrollment in a four-year college are 55 percent less likely to complete college than their white peers. The African American disadvantage in college completion narrows a bit when controlling for social origins and high school GPA, but it is still lower than that of whites (though statistical significance fluctuates from model to model). The small numbers of Vietnamese students who enrolled in a four-year college are much more likely to complete college than are white students. Given the problems of using NSC records to measure East Asian college enrollments, we have doubts about the reliability of the negative coefficient on college completion. Although this transition is only measured with NSC data (the ratio of graduates to college entrants), there is still the potential of differential mismeasurement. In contrast, the low rate of college completion among Filipino students in figure 5.7 is consistent with low transition from college preparedness to college enrollment in figure 5.6. This is an important finding that should be replicated with national level data.

Alternative Pathways to Completing College Only 30 percent of all UW-BHS high school seniors earned a college degree within seven years following high school graduation. Three-quarters of the college graduates (or 22.5 percent of all high school seniors) followed a linear progression—successfully crossing each sequential stage in the College Pathways Model. Nonetheless, 7.5 percent of all high school seniors earned a bachelor’s college degree through an alternative pathway. These individuals diverged from the college-bound track but managed to navigate their way back to a four-year college and to complete college. Although the probability of getting back on the college-bound pathway is relatively low (see the conditional probabilities of upward movement in figure 4.2) at every stage in the process, the absolute number who beat the odds is significant. Moreover, they represent an interesting group of students who have been able to find their way to graduate from college even after falling off of the linear progression route. Some of these alternative-pathway college graduates did not express college ambitions in high school and a few did not prepare for college in high school. However, the majority of alternative-pathway college graduates found their way to a four-year college after enrolling in a technical or two-year community college after high school. Financial considerations,

College-Pathway Transitions  183

or possibly other personal problems, may have led many of these students to first enroll in a community college or to delay entry altogether.33 Recall that about 73 percent of students who enroll in a four-year college right after high school will graduate and earn a degree in seven years, as opposed to the 10 percent among students who do not enroll in a four-year school (see figure 4.2). As such, one can argue that alternative-pathway college graduates represent a particularly highly motivated and resourceful group of students who were able to beat the odds. Table 5.2 shows a more nuanced portrait of the role of community college as an alternative pathway to a college degree. The sample for this table is the 5,888 students (about two-thirds of the UW-BHS sample) who were not enrolled in a four-year college in the year right after high school. The first panel in table 5.2 shows the enrollment status in the year after high school of the UW-BHS seniors who were not enrolled in a four-year college. About 39 percent were enrolled in a two-year college and 61 percent were not enrolled in any college—presumably most were working. The next panel shows the percent of students who earned a college degree conditional on their enrollment status in the year after high school. About one in five (18 percent) students who enrolled in a two-year college managed to obtain a college degree. College completion among students who did not enroll in any college immediately after high school is a relatively rare event—only 5 percent of this group completed a college degree within seven years. The idea of taking a “gap year” before continuing on to college is an option, but relatively few students not enrolled in any college right after high school manage to earn a degree in seven years. The overall college graduation rate of 10 percent of students not enrolled in a four-year college right after high school (the upward sloping arrow in figure 4.2) is a weighted average of two groups—those who enrolled in a community college and those not enrolled in any school in the year after high school. Of these two pathways, community college is more likely to lead to graduation (18 percent versus 5 percent). However, the odds of graduating from college are four times higher for students who are enrolled in a four-year college than those enrolled in community college (73 percent to 18 percent). There are only modest variations by gender, race-ethnicity, and immigrant generation in the patterns of enrollment in community college, and in the likelihood of earning a degree conditional on community college enrollment. Women hold slight advantages over men in the alternative pathways to college graduation, similar to their modest edge in every stage of the linear progression through the College Pathways Model. Female students are four percentage points more likely than men to be enrolled in a community college (41 percent to 37 percent). They are also about one percentage point more likely to earn a four-year college degree

61%

59 63

59 63 68 66

Total

Gender Female Male

Race-ethnicity White African American American Indian East Asian

Not in Any College

41 37 32 34

41 37

39%

In TwoYear College

100 100 100 100

100 100

100%

Total

3,381 890 112 295

3,138 2,750

5,888

(N)

Enrollment Status in Year After High School

6 3 0 7

5 4

5%

Not in Any College

19 10  8 21

18 17

18%

In TwoYear College

11  6  3 12

11  9

10%

Total

Percent College Graduate by Enrollment Status in Year After High School

113 18 0 14

98 65

163

Not in Any College

264 34 3 21

233 175

408

In TwoYear College

377 52 3 35

331 240

571

Total

Graduated from Four-Year College (N)

2,006 563 76 194

1,851 1,729

3,580

Not in Any College

1,375 327 36 101

1,287 1,021

2,308

In TwoYear College

3,381 890 112 295

3,138 2,750

5,888

Total

Enrollment Status in Year After High School (N)

Table 5.2    Alternative Pathways to College Graduation of UW-BHS Seniors Who Did Not Enroll in a Four-Year College Right After High School, by Gender, Race-Ethnicity, and Immigrant Generation

43 41 39

34

66

42 60 48 32 25 26 34

57 59 61

59 40 52 68 75 74 66

100

100 100 100

100 100 100 100 100 100 100

544

808 1,020 3,516

200 194 165 117 142 239 153

1

2 6 5

2 3 7 5 1 0 3

14

18 18 18

17 30 14 21  8 17  8

 5

 9 11 10

 8 19 10 10  3  5  5

2

11 35 115

2 2 6 4 1 0 3

25

64 76 243

14 35 11 8 3 11 4

27

75 111 358

16 37 17 12 4 11 7

361

462 599 2,158

117 77 85 79 106 176 101

183

346 421 1,358

83 117 80 38 36 63 52

544

808 1,020 3,516

200 194 165 117 142 239 153

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Note: The sample of 5,888 UW-BHS students includes all respondents who did not enroll in a four-year college right after high school and with nonmissing data on college aspirations, college expectations, and college preparedness.

Immigrant generation First generation Second generation Third-and-higher  generation Missing

Cambodian Vietnamese Filipino Other Asian Pacific Islander Mexican Other Hispanic

186   From High School to College

than men among both those not enrolled (5 to 4 percent) and those who enrolled in a two-year college (18 to 17 percent). Only one minority group has established a relatively successful pathway to college graduation through community college enrollment. About 60 percent of Vietnamese students who were not enrolled in a four-year college attended a community college—and 30 percent of these Vietnamese students (who began in a community college) earned a college degree in seven years. These figures are much higher than any other group. Most other minorities in this category were much less likely to enroll in a community college (among those not attending a four-year college): only 37 percent of African Americans, 32 percent of Native Americans, 25 percent of Pacific Islanders, and 26 percent of Mexican Americans in the UW-BHS high school senior sample. The college graduation rate of white students who enroll in a two-year college right after high school is only 19 percent. Still lower are the comparable college graduation rates of 17 percent for Mexican Americans, 10 percent for African Americans, and 8 percent for both American Indians and Pacific Islanders. The opportunity for a second chance at a college degree via community college does not seem to be a road well traveled by most disadvantaged groups. We now turn to a formal analysis of the determinants of college graduation via alternative pathways as we did the other steps in the College Pathways Model. The structure of the analysis follows the precedent of the linear progression of the College Pathways Model with five models or equations: (1) baseline model with only gender, race-ethnicity, and immigrant generation; (2) baseline-model variables plus social origins; (3) baseline-model variables plus social origins and GPA; (4) baselinemodel variables plus social origins and encouragement; and (5) saturated model with ascriptive, social origins, GPA, and encouragement variables. Selected results (models 1, 2, and 5) are shown in figure 5.8, and the complete results are presented in appendix table 5.A6 online.34 Before looking at disparities by gender, race-ethnicity, and immigrantgeneration via alternative pathways to college graduation, we should note the strong direct effects of parental education (especially the father’s), home ownership, and family structure (shown in appendix table 5.A6). Our initial expectation was that community college and other alternative pathways to a college degree would be a means to success for individuals and groups who lacked the resources and connections to stay on the linear progression route. In other words, “second chances” (an alternative trajectory) would be most common among groups that fared poorly through the linear progression route of the College Pathways Model. The results in model 2 in online table 5.A6, however, show the opposite. An advantaged background helps students to stay on track to become a college graduate through the College Pathways Model, but these resources

1.37

1 2 5 Female

1.27

0.71

1 2 5 African American

0.52

0.300.32

1 2 5 American Indian

0.21

1 2 5 East Asian

2.46

1 2 5 1 2 5 Cambodian Vietnamese

2.43

3.53

1 2 5 Other Asian

1 2 5 Pacific Islander

0.280.29

1 2 5 Mexican

0.38

1 2 5 Other Hispanic

0.44

1 2 5 1st generation

1 2 5 2nd generation

Interpreting coefficents in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

1 2 5 Filipino

0.22

Model 1: Gender, race-ethnicity, and immigrant generation only Model 2: Model 1 plus social origins Model 5: Model 2 plus ecouragement index and GPA

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 5.A6. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

Figure 5.8    Logistic Regression of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors Who Did Not Enroll in a Four-Year College

Regression coefficient

188   From High School to College

are also absolutely critical in helping students to get back on track should they slip along the way. Following the linear-progression model (figures 5.2 to 5.7), women have higher college aspirations than men and are more likely to prepare for college in high school, enroll in a four-year college right after high school, and complete college—each stage of the College Pathways Model. The differences are not huge, but they are consistent and cumulative in their impact on the gender difference in college graduation rates. Figure 5.8 shows that the female edge extends to students who did not enroll in four-year college right after high school. Among this universe, women are 27 percent more likely to graduate from college than are men (see model 1). This figure rises to 37 percent in model 2, with social origins held constant. Given that more men drop out of high school than women, the UW-BHS male sample is a bit more selective than the female sample. Therefore, the gender gap in model 2 is probably a more accurate measure of the female edge over their male counterparts with equivalent family backgrounds. The primary mediating factor in women’s relative success in the community-college route to a college degree is academic performance in high school, as measured by GPA. This variable serves to explain why women are more likely than men to enter college (among the college prepared), more likely to complete college (among entrants), and more likely to find an alternate pathway (among those not enrolling in a fouryear college after high school). These are three different outcomes among three different populations. The finding that superior academic prowess in high school is the common factor does not explain why women earn higher grades in high school than male students. The race-ethnicity coefficients in figure 5.8 show that only Vietnamese have an edge relative to white students via the alternative-pathways trajectory. Vietnamese students display unusually high levels of resourcefulness in obtaining a college degree even if they did directly follow the standard (linear-progression) College Pathways Model. Although Vietnamese students tend to come from disadvantaged families, they often had comparable or higher college aspirations, expectations, and preparation relative to their white peers. When SES and family background are held constant, Vietnamese had even higher outcomes than white students on all transitions (except for enrollment in a four-year college). The increase in the Vietnamese odds ratio from model 1 to model 2 (from 2.4 to 3.5) in figure 5.8 suggests that Vietnamese from poorer socioeconomic backgrounds were particular adept in making effective use of community colleges. Among Vietnamese students who did not attend a four-year college, 60 percent enrolled in a two-year college compared to 40 percent of white students in the same category. Vietnamese students who began at a community college were also more likely to earn a college

College-Pathway Transitions  189

degree within seven years than similarly enrolled white students (30 percent to 19 percent, see middle panel of table 5.2). Quite a few minority groups, especially African American, American Indian, Pacific Islander, Mexican, and Other Hispanic students, were much less likely than white students to find a successful alternative path to college completion. Differential family resources, as measured in model 2, reduced part of the minority disadvantage, but most of original gaps were still significant in all models. If there is a “back-door” path to college graduation in American society, it seems that whites are more likely to benefit than are minorities (with the exception of Vietnamese). First-generation high school seniors who did not enroll in a four-year college right after high school were slightly less likely to complete college than their third-and-higher-generation counterparts, though this finding was not statistically significant. This gap is largely explained by the lower family SES composition of newcomers to American society. There is no statistically significant difference between second and third-and-highergeneration students in their likelihood of getting a college degree through an alternative pathway. The American educational system allows for “second chances” (and even third and fourth chances). It is possible to drop out of high school and return to finish an alternative high school equivalency certificate in community college. Students can also fail college and then start all over again at any point in time. Low-income students can begin at a community college and then transfer to a four-year college. Accounts of famous Americans who recovered from early academic failure with a second chance are not only stories of individual resilience but also of a system that allows for comebacks and flexibility. While there are major channels of opportunity for students from all backgrounds, our findings show that this applies more to white students than to minorities.

Conclusions A generation ago, lower levels of college enrollment and graduation among women and minorities were thought to be rooted in low aspirations.35 If this was ever true, it is no longer an accurate description of the educational stratification in American society. Almost every high school senior aspires to some sort of postsecondary schooling at some point in their lives. About three-fourths of UW-BHS high school seniors express a desire to complete a four-year college degree (BA or BS). This figure is similar to those reported in national surveys of high school students.36 There are only modest differences in college aspirations between men and women and across racial and ethnic groups. With the reversal of the gender gap in education, young women are now more likely to aspire to complete college than men, but the gap is small, only three or four

190   From High School to College

percentage points. Immigrants are only slightly less likely to express a wish to graduate from college than long-resident Americans (thirdand-higher generation), and the children of immigrants hold slightly higher college aspirations than the third-and-higher generation. There are slightly wider racial and ethnic gaps in college aspirations than by gender and immigrant generation, but they are, all in all, relatively modest. More than two-thirds of American Indian, Pacific Islander, and Mexican origin youth report that they aspire to graduate from college. Black and white high school seniors are right at the overall average— about 75 to 76 percent. College aspirations for Asian American students are even higher, and over 90 percent of East Asians aspire to graduate from college. The belief in the American dream, and the understanding that a college degree is the ticket to upward mobility, is ubiquitous if not universal. Given that college aspirations are universally high, they do not play a major role in explaining the vast inequality in college-graduation rates by gender, race-ethnicity, and immigrant generation. If there were no between-group variation in college aspirations, the measured differences in college graduation rates would only decline, on average, about 10 percent (or less). The slightly lower levels of college aspirations reported by some disadvantaged racial and ethnic groups and immigrants are entirely explained by family socioeconomic background. This is not true, however, for the higher desires for college graduation of Asian Americans and second generation immigrants. The belief that college is the gateway to opportunity is widespread among Asian and second generation high school students regardless of their social origins.37 The same conclusion almost holds for differences in college expectations—a concept that was measured by adding “realistically speaking” to the college-aspirations question. About 89 percent of high school seniors who aspired to complete college expected to reach their objective, but this pattern of drop-off from aspirations to expectations is correlated with the location of students in the socioeconomic hierarchy. For example, disadvantaged minorities including African Americans, American Indians, Cambodians, Filipinos, Pacific Islanders, and Mexican Americans, were less likely than white students to think their aspirations would be realized. These gaps are entirely a function of their socio­economic position. In model 2 in figure 5.3, the race-ethnicity gaps (and those among immigrants) in college expectations are dramatically reduced and no longer statistically significant for most disadvantaged minorities. For this reason, we do not consider the racial and ethnic gaps in college expectations to simply be products of cultural socialization or community ambitions between groups. Awareness of one’s circumstances and limited options is the primary reason why some groups are more likely to report more modest expectations than aspirations.

College-Pathway Transitions  191

In contrast to aspirations and expectations, differences in high school preparation for college among students who expect to graduate from college are a major reason for inequality in college graduation by gender, race and ethnicity, and immigrant generation. We measure high school preparation for college by three criteria: taking AP tests, taking the SAT or ACT, and applying to a four-year college while in high school (two of these criteria suffice to have a student coded as college prepared). These indicators have a broader interpretation of college preparation than coursework and other actions that put students on a trajectory toward “college readiness.” Only 68 percent of UW-BHS high school seniors who expect to graduate from college are college prepared. Between-group variations in college preparation explain about 20 percent of overall gaps in college graduation. College preparation—being on the college track—is closely intertwined with a student’s academic performance. Students who enter high school with above-average grades are directed by teachers, guidance counselors, and parents into honors classes and courses in math, science, and other areas that will meet the criteria for college admission. On the other hand, students who are struggling may be tracked to a general curriculum that is more vocational than academic. The association is shown with the addition of GPA to the prediction of college preparation in model 3 in the bottom panel of appendix table 5.A2—the adjusted R-squared score rises from 7.0 percent to 16.1 percent. Our interpretation, presented in more detail in chapter 1, acknowledges that a significant share of inter-individual variation in educational outcomes is due to academic performance (indexed by GPA), which is a product of innate differences in ability and the social feedback loops in parenting and the social environment. Regardless of the role of ability as an important explanation of inter-individual variation in educational outcomes, there is no empirical or theoretical support for assuming that intergroup differences in schooling are caused by innate differences in ability. The mediating role of GPA as an explanation of inequality in educational outcomes between ascriptive groups is a result of “training”— differential family investment and coaching that leads to improved levels of performance. This interpretation is consistent with the finding that social origins explain most of the gaps between disadvantaged racial-ethnic groups and whites in college preparation. The advantages in college readiness for Asian Americans are partially a function of encouragement. Student GPA packs a punch in general, but it is not the primary reason for intergroup differences. The largest components of the ascriptive gaps in college graduation are products of differences in college entry (among those who are prepared) and college completion (among four-year college entrants). These two factors contribute to about 40 to 50 percent of the lower achievement of

192   From High School to College

disadvantaged racial and ethnic groups and about one-third of the lower male college graduation rate. Interestingly, Asian Americans who experience advantages in earlier stages of the College Pathways Model are less advantaged in the transition to four-year college enrollment. There is tentative evidence that some Asian groups may encounter barriers to college enrollment and completion. Family socioeconomic background “explains” nearly all of the lower levels of college aspirations, expectations, and preparedness among disadvantaged minorities and first-generation immigrants. This is less true, however, for the transitions to college enrollment and college completion. This is also true for several of the Asian American communities, with the notable exception of East Asian and Vietnamese students. There is some reduction in the majority-minority gaps from model 1 to model 2 in figures 5.6 and 5.7, but the minority disadvantages do not disappear. Not all the lower rates of college entry and completion (odds ratios below 1) are statistically significant, which is in part an effect of small sample sizes. The overall pattern is, however, one of lower rates of college entry and completion that is not explained by any of the measured covariates in our model. This pattern is also evident for first-generation immigrants in terms of college entry, but not for college completion. The problems faced by disadvantaged minorities in college entry and college completion are also evident in the alternative-pathways trajectory. In general, white students who fall off the college bound track are more likely to graduate from college than are minorities in a similar condition. The possibility of attending a low cost community college and then transferring to a four-year college is open to all students, and some students from every background are able to take advantage of this alternative route to college graduation. But in general, white students and students from advantaged backgrounds are much more likely to find their way to college graduation than other groups even if they do not enroll in a four-year college right after high school. The only Asian group that is able to narrow the inter-ethnic gap in college graduation through alterative pathways (community college) is the Vietnamese. The below-average rates of college graduation via the alternative-pathways route (not enrolling in four-year college right after high school) by most minorities is partially explained by family SES, but most of the net gaps in college graduation with white students are not explained by any of the variables in our models. The importance of social origins, the set of variables representing social class and other family background variables, cannot be overstated. Students from families with more resources, as measured by parental education, father’s occupation, home ownership, and family structure do much better than students from disadvantaged backgrounds in preparing for college while in high school. Minority students and first-generation immigrants tend to have fewer resources and begin the trek to college far

College-Pathway Transitions  193

behind third-and-higher-generation white students. Controlling for social origins “explains” virtually all of the racial and ethnic differences in college aspirations, expectations, and preparation. The last two steps of the College Pathways Model—enrollment in a fouryear college and completing college once begun—are the most important factors in accounting for lower college graduation rates among disadvantaged minority groups. Family resources are important but do not tell the whole story of lower transition rates of African American, American Indian, Pacific Islander, and Hispanic students. Perhaps other intangible factors that are not measured here may explain why white students are more likely to continue on the college path beyond high school. These might include social networks of family members or other adults who have been to college, assistance in applying for loans or other sources of financial support to attend college, or better academic preparation. Asian Americans, who were overachievers in high school, have a more complicated record in enrolling in four-year colleges and obtaining a degree. East Asian students continue to do well, and Vietnamese students stand out for their educational attainments despite their lack of family resources. But other Asian students, including Cambodians and Filipinos, have lower rates of college enrollment and college completion than expected. Perhaps the Vietnamese students (and first-generation immigrants) who do enroll in college are very selective in terms of intangible characteristics. The Vietnamese in the UW-BHS sample are primarily those (or the children of those) who arrived in the 1980s known as the “boat people,” who were not as highly educated or as closely connected to the American government as the first wave of Vietnamese refugees who arrived in 1974 and 1975. Many boat people were attacked by pirates at sea and had to endure years in refugee camps before being allowed to settle in the United States. The memories of these horrific experiences may have hardened the determination of many Vietnamese immigrants to become successful in their adopted homeland and to push their American-born children to do the same. In the next chapter, we push beyond social origins to explore cultural variables that might help us interpret the variations in the College Pathways Models by gender, race-ethnicity, and immigrant generation.

Chapter 6 A Closer Look at the Role of Culture in Explaining Educational Transitions

C

ulture is one those words that is widely used in scientific and popular discourse but with a variety of meanings.1 Students entering college far from home are told to anticipate “culture shock.” New leaders often express their intent to change the “culture” of an organization or business. Immigrants to the United States have often been told to shed their traditional culture if they wish to “become American.” Societies, social classes, and ethnic groups are only some of the groups thought to have cultures that include language, religion, customs, norms about appropriate behavior, and values that guide understandings and behavior. The varied meanings of culture in popular discourse arise, at least in part, from its role as an expansive, often imprecise concept in the social sciences. One textbook definition of culture is a society’s (or group’s) “symbol systems and the information they (the symbols) convey.”2 This particular text goes on to explain that symbol systems include oral and written language as well as “body language” that is conveyed by gestures and facial expressions that are widely understood by members of the society or group. The information that is conveyed by symbols includes basic knowledge about the physical, biological, and social world as well as an ideology of beliefs, values, and norms that justify and guide human action. Other texts and scholarly works might offer different definitions of culture, but the broad inclusiveness of the concept would be similar. In addition to the problem of breadth—what is not included within the all-inclusive rubric of culture?—other aspects of the concept have fueled intellectual debates and misunderstandings. For example, the origins of a society’s culture and the relationships between social institutions and culture are only dimly understood. Culture is certainly related to

194

Explaining Educational Transitions   195

the technological development of a society and the material conditions of life of a group. Agricultural and industrial societies differ in terms of population density, the proximity of kin, and the family as an economic unit—all of which shape how people think and behave. For example, a recent study found that rice-growing areas are much more likely to be cooperative and hold communal values than people who live in wheatgrowing regions within the same country.3 The other major question is how cultures evolve and change. Although cultures are generally considered stable and slow to change, a number of vivid examples show that cultures can change dramatically in response to environmental conditions. Jared Diamond describes the tragic encounter between the aggressive Maori, with a culture of war-making and dominance, and the peaceful Moriori peoples, with a tradition of resolving disputes peacefully, in the Chatham Islands in the South Pacific in 1835.4 The Maori and the Moriori shared a common ancestry but developed quite different cultures within a few centuries of separation. Many similar examples from recent history can be cited, including the change from militarism to pacifism of the Japanese nation following World War II. Culture is molded, in large part, by political, economic, and demographic conditions, but there is also a built-in reproduction of culture through intergenerational socialization and institutions that reward conformity and punish those who deviate from tradition. Cultural continuity—identified without reference to its origins or supporting institutional fabric—is often given as the answer to any question about differences between societies and groups. Why do some societies progress and others don’t? Why are some people the rulers and other people the ruled? In popular discourse, and in some branches of social science, these differences (and many more) are assumed to be “explained” by religion, language, national origin, or any other marker of culture that varies between societies and groups. Since cultures can and do change, explanations of differences in behavior based on culture contain an element of circularity.5 The uncritical use of cultural explanations has a sordid history as rationalization for conquest and dominance. For example, European colonial powers often justified their rule on the basis of claims that natives (and their cultures) were inferior and not capable of self-governance and participation in democratic political practices. At one time, cultural differences were thought to reflect inherent “racial” qualities that no amount of education or experience could modify.6 Although race, as a biological concept, has lost much of its credibility, cultural interpretations as justifications for differential rights are still widespread in many societies. At various times and places in American history, African Americans, Asians, Jews, and Catholics were restricted from certain jobs, neighborhoods, schools and universities, and social clubs by dominant groups. Minorities were disparaged as disloyal, uncivilized, untrustworthy, lacking norms of proper

196   From High School to College

conduct, or having cultural traits that made others (“us”) uncomfortable. In almost every society around the globe, nomadic peoples who refuse to give up their traditional culture and settle down are dispossessed of their lands and resources. The cultures and character of “outsider” groups in almost all modern societies—including the Roma in Europe, middleman minorities dependent on small scale trade, and immigrants—are generally disparaged by the dominant group, and the outsiders are not accorded equal rights and citizenship. The other analytical problem with defining the concept of culture is the exclusive focus on between-group variation and lack of attention paid to within-group variation. Groups are rarely monolithic—there will always be persons who deviate from prescribed norms that are identified with a group’s culture and institutions. Moreover, as societies change and modernize, there is usually a trend towards increasing within-group heterogeneity. For example, religion has historically been considered an important predictor of fertility, divorce, and political affiliation (among other outcomes) in the United States because of shared values, kinship ties, and intergenerational socialization. Although religious differences have not disappeared, other factors, including education and economic differentiation, have created greater within-group variation in cultural outlooks and behavior. If families and other primary groups are to remain the major influence on the intergenerational transmission of values, there must be high levels of homogamy and homophily within groups and spatial segregation between groups. In this chapter, our aim is to consider cultural attributes as potentially important mediating variables of educational disparities between ascriptive groups defined by gender, race-ethnicity, and immigrant generation. In contrast to many other studies, however, we consider cultural explanations within an analytical framework that anticipates many of the weaknesses of culture as a post hoc interpretation of between-group outcomes. The first step is to formulate and carefully measure cultural attributes as systems of meaning and values that are hypothesized to motivate behavior. Drawing upon the research literature, we specify a series of specific variables that are grouped under the rubrics of cultural contexts, cultural orientations, and cultural expressions. In chapter 5, we included academic ability (GPA) and significant-other influences (encouragement) as intervening variables between ascription, social origins, and educational outcomes. In this chapter, GPA and encouragement are embedded into a more elaborate formulation of the role of cultural factors. Cultural factors may mediate the influence of the ascriptive variables of gender, race-ethnicity, and immigrant generation, but also of social origins (SES and family background). Using survey data, cultural factors are measured as individual-level characteristics; this allows us to measure the direct impact of attitudes and orientations on educational

Explaining Educational Transitions   197

outcomes. Our primary focus, however, is on cultural attitudes, orientations, and behaviors that may explain (mediate) intergroup differences in educational outcomes. Social science research on the role of cultural values as the explanation for intergroup inequality has been very controversial. The lack of a scientific consensus on how to define and to reliably measure cultural factors has been part of the problem, but the more troubling issue has been the uncritical use of cultural explanations in ideological and political struggles. In the 1960s, Oscar Lewis and Daniel Moynihan popularized the phrase culture of poverty to describe the situation of Puerto Ricans (Lewis) and African Americans (Moynihan) in New York and other large cities.7 Lewis and Moynihan, in their independent works, were careful to distinguish between their description of cultural values as nonadaptive and the analysis of the structural forces that produced cultural patterns. They each argued that the experiences of American minorities were shaped by discrimination and the lack of opportunity. These structural factors, in turn, led to broken families, hyper-unemployment, and the loss of ambition among young men. Without strong families and the educational credentials necessary for upward mobility, minorities were increasingly trapped in a cycle of poverty as maladaptive values had become entrenched in the cultural fabric of disadvantaged communities. The concept of the culture of poverty was often adopted by political conservatives, who ignored Lewis’s and Moynihan’s analyses of structural factors and simply invoked the phrase to reinforce their claims that welfare and other government programs to reduce poverty were ineffective and wasteful.8 The “culture of poverty” thesis soon became a political football, as illustrated with the frequent mention of “welfare queens” by then presidential candidate Ronald Reagan, to claim that governmental welfare programs would invariably be abused by the undeserving poor. A substantial body of literature has been published on the culture of poverty and the role of culture in explaining differences in socioeconomic outcomes between groups.9 Many of these writings are analyses of the polemical debates over culture. There is much to criticize on all sides. Those analysts or policy makers who focus solely on cultural differences often ignore the historical context of how limited opportunities, segregation, and discrimination have molded values, attitudes, and behaviors that may reinforce existing obstacles to upward mobility. However, it is important to acknowledge that intergenerational socialization and community organization (or disorganization) may reproduce maladaptive orientations and behaviors. Not every invocation of culture is an attempt to blame the victim (or to justify the success of achievers). There is likely to be a myriad of complex relationships between social institutions, opportunity structures, and cultural orientations that shape and condition intergenerational social mobility. The key point is that structure

198   From High School to College

(shorthand for social structure and institutions) and culture are not mutually exclusive concepts or explanations of motivated behavior.10 The challenge for social science in general, and for our research agenda here, is to find ways to blend structural and cultural analysis to interpret differences in socioeconomic outcomes. Examples abound of creative research that seeks to identify and measure the role of cultural factors as both mediating variables and significant independent causal influences on socioeconomic outcomes. In his celebrated 1967 ethnographic study of the lives of African American men who spent their days and nights hanging out on the titular Tally’s Corner, Elliot Liebow found that appearances could be deceiving.11 Although the day-to-day lives of the jobless men on Tally’s Corner appear to epitomize the culture of poverty, drifting from one dead-end job to another and fathering out-of-wedlock children with multiple women, Liebow’s indepth interviews revealed that most street-corner men had middle-class values. They desired stable employment and wanted to settle down in conventional marriages with children. But the limited opportunities in inner cities did not offer employment and wages to realize their aspirations. The apparent culture of poverty of street-corner men was a consequence, not the cause, of their plight. After publishing the magnum opus The American Occupational Structure in 1967, Otis Dudley Duncan and his colleagues broadened the basic model of social stratification to include cultural orientations as intervening variables that might explain differential socioeconomic attainment.12 They found relatively little empirical support for the claims that the “need for achievement,” the “Protestant ethic,” and other cultural values were major causal or intervening variables in the process of intergenerational race and ethnic stratification in American society.13 Although not specifically focused on explaining race and ethnic differences, the Wisconsin school of social stratification expanded the basic model of intergenerational social stratification (the Blau-Duncan model) to include intervening cultural factors that are communicated through socialization and the social influences of parents, friends, and teachers (considered to be “significant others”) on educational and occupational attainment.14 One of the key findings of the Wisconsin school is that the internalized expectations of significant others (as perceived by students) is the critical pathway that mediates much of the impact of socioeconomic origins on educational aspirations and attainment. However, research that extended the Wisconsin model to analyze racial and ethnic inequality reported more ambiguous results. In their important 1976 study, Alejandro Portes and Kenneth Wilson found little empirical support for the widespread belief that the lower educational attainment of African Americans is due to a lack of high aspirations and unsupportive significant others.15 Alan Kerckhoff argued that the Wisconsin model overemphasized the role of socialization as the primary intervening mecha-

Explaining Educational Transitions   199

nism and neglected processes of “allocation”—institutional practices, such as tracking in schools and disciplinary procedures, that limited the educational opportunities of disadvantaged minority students.16 The results of our study provide support for both the socialization and allocation perspectives of educational stratification. High educational aspirations and expectations of college graduation are necessary (but not sufficient) prerequisites for college graduation. Very few students with low educational ambitions graduate from college. Moreover, the encouragement index has very strong direct effects on college preparation in high school and enrollment in a four-year college (see chapter 5). However, the lower educational attainments of disadvantaged minorities (African Americans, American Indians, Pacific Islanders, and Mexicans) are not explained (mediated) by the lower levels of encouragement they receive. In fact, minority groups appear to receive more encouragement to attend college than do white students. Stephen Morgan’s review of the status attainment literature reminds us that students’ survey responses about their future educational attainment may reflect perceptions of limited opportunities as much as family socialization.17 The inclusion of the transition from educational aspirations to expectations in this study was meant to capture this distinction. Although the lack of encouragement is not a major reason for lower educational outcomes of disadvantaged minorities, encouragement and academic achievement do play an important mediating role in explaining gender inequality and in the above-average outcomes of Asian Americans in the early educational transitions in high school. These complex and inconsistent findings (in our research and in the literature more broadly) may represent an inadequate conceptualization of the role of early family life on educational outcomes. Adding variables representing socialization and significant others’ influences is certainly an important step in building a more realistic model of educational stratification, but there is room for a more ambitious approach to bringing culture into models of educational attainment. The inclusion of GPA (academic performance) and encouragement (significant others’ influences) in chapter 5 drew upon the pioneering work of the Wisconsin school and other classic studies of education attainment of young adults.18 The model was intended to be a first-order approximation of the variety of processes that occur between childhood and the completion of college. In this chapter, we present a more elaborate formulation of potential cultural influences and mechanisms that may mediate the role of ascription and social origins on educational outcomes. Our aim is to provide a more comprehensive picture of how cultural values are conditioned by family background, internalized into attitudes, and expressed in action between ascriptive groups defined by gender, race-ethnicity, and immigrant generation. But first, it is necessary to take a close look at the conceptualization, measurement, and relationships among a broader

200   From High School to College Figure 6.1    A Model of Cultural Context, Cultural Orientations, and Cultural Expressions Cultural context Parental communication and support Parents know friends Encouragement index Family expects college Friends’ college plans

Cultural expressions Homework hours GPA (effort)

Cultural orientations School is important Locus of control (self-efficacy) Source: Author’s compilation.

array of cultural attributes. Our proposed framework for the role of cultural influences is presented in figure 6.1 under three clusters representing cultural context, cultural orientations, and cultural expressions. The first cluster in figure 6.1 is labeled cultural context and includes a number of social factors that are likely to “make culture” by influencing the values and attitudes of children and adolescents towards schooling and the likelihood of college graduation in particular. We identify aspects of the family and environment that are (1) antecedent to decisions about college and (2) “pro-college” or at least supportive of high educational aspirations. Since the survey questions about these background factors are measured contemporaneously with outcomes, we cannot assert temporal priority with certainty. The survey questions, however, about family and peer influences are “logically prior” and are likely to be highly correlated with temporally prior influences. These family and peer influences are not measured directly but are reflected through the perceptions of the UW-BHS high school students. For our purposes, the perceived influences by students are actually preferable to direct measures of communications from family and friends. For example, parents may think that they encouraged their child to attend college, but unless the child “hears” the message, it is unlikely to have an impact. Each of the following five components of perceived cultural context are reported by students: • Parental communication and support • Parents know friends

Explaining Educational Transitions   201

• Encouragement index • Family expects college • Friends’ college plans The first three variables are indexes or constructions of multiple survey questions (see appendix table 6.A1, available online).19 Each item was coded (or reverse coded) so that higher values on a scale from 0 to 3 are hypothesized to positively influence educational attainment. For multipleitem scales, the index is coded at the simple average (mean) of each item, excluding missing items. Parental communication and support is an index constructed from six questions in the UW-BHS senior survey. The first two survey questions asked student respondents how often they had discussed “school activities” and “going to college” with their parents. The other four items asked whether students agreed (or disagreed) with statements about the quality of their relationships and interactions with parents (support, source of advice, frequent in-depth conversations, and parental disapproval). Although most of the items did not directly measure parental support for college, we assumed that students with close ties with their parents would be likely to feel supported to attend college. The second index, parents know friends, was based on two questions that asked if the student’s parent(s) knew (1) the student’s closest friends and (2) the parents of the student’s closest friends. These questions are related to James Coleman’s concept of social capital that is expressed in the collective supervision (or watchfulness) of adolescents by a community or neighborhood.20 The encouragement index is the sum of the number of significant others who think that the most important thing for the student respondent to do after high school is to go to college. The significant others included in the index are: father, mother, brothers and sisters, friends, and an adult “whose advice you value.” Because our objective is to measure prior familial and social contexts, we excluded the item that asked about encouragement from a favorite teacher. Moreover, a teacher’s encouragement is more likely to be influenced by the student’s academic performance—an attribute that is conceptually if not empirically independent from encouragement from family and peers. Another survey question that touched on a similar concept asked if the student’s family always expected her or him to attend college ( family expected college). The final item, friends’ college plans, was a single survey question that asked how many of the student’s school friends were planning to go to a four-year college. The five-category scale was based on responses that ranged from “none or some” (coded 0) to “most or all” (coded 4). These five variables are posited to be significant influences that motivate and encourage students to believe that college is the right step after high

202   From High School to College

school. The underlying assumption is that authoritative parenting, strong social support, high expectations, and positive peer models socialize adolescents to adopt high educational aspirations and to follow through with behaviors that are likely to increase their likelihood of success. The next cluster of variables in figure 6.1 is labeled cultural orientations to represent conscious awareness of the importance of schooling and the responsibility for taking charge of one’s life. The underlying idea of cultural orientations is that if students are exposed to positive cultural contexts, they are more likely to hold pro-education values and express attitudes that conform to the high expectations of others. We have identified two variables to tap cultural orientations. The first, school is important, is based on responses to a question (in the UW-BHS survey) that asked the student if she or he agreed with the statement, “How well I do in school is an important part of who I am as a person.” The reverse-coded scale ranged from 0 (strongly disagreed) to 3 (strongly agree). Overall, about 30 percent of UW-BHS students strongly agreed with this statement and another 40 percent agreed. The second variable, locus of control (or self-efficacy), is a well-established concept in social psychological literature21 and is sometimes described as a “sense of personal control.”22 The standard locus of control scale is based on five standard items that tap whether the respondent feels a sense of control over events and plans or whether everything is up to fate or luck. Persons with an above-average sense of self-efficacy (an internal locus of control) are more likely to think that hard work and planning are necessary for future success. Self-esteem is another widely analyzed trait that is often posited to be integral to school success and college ambitions.23 However, our preliminary research found that self-esteem was much less consequential than having an internal locus of control.24 Many students expressed a positive sense of themselves, but we saw little net relationship between self-esteem and any of the educational transitions in this study. The third cluster of intervening variables in figure 6.1 is cultural expressions, which are meant to capture behavioral patterns or “culture in action.” To do well in school and to enroll in college, students need to put in time and effort. The two empirical indicators of cultural expressions are self-reported GPA and homework hours. As noted in chapter 3, GPA, or academic performance, is partially a product of innate ability, but our focus here is on the socially determined component that might mediate differences in socialization, coaching, and training between ascriptive groups by gender, race-ethnicity, and immigrant generation. The socially determined component might be reflected in behaviors such as attending class regularly, completing homework, paying attention in class, and preparing for exams. These dimensions of GPA might be labeled as effort, which could be conditioned by family and community characteristics, admonitions, and other aspects of groups that share common backgrounds and social interaction.

Explaining Educational Transitions   203

In table 3.6, we estimated to what degree the gaps (by gender, raceethnicity, and immigrant generation) in GPA (actual and self-reported) could be accounted for by intergroup differences in test scores. We postulated that the residual—variations in GPA unexplained by test scores— was a proxy for effort. Although a crude indicator, the results were highly suggestive of the assumption that academic performance purged of test scores reflected motivations and culturally shaped behaviors that led to school success. For example, women are better able to translate their “abilities” (test scores) into school grades (GPA) than men. This outcome is consistent with other patterns showing that women, relative to men, have fewer disciplinary problems, spend more time doing homework, and more strongly agree with the statement, “Doing well in school is an important part of who I am.” We conclude that effort (actual GPA minus predicted GPA) is a very plausible index of culturally expressed actions that are likely to lead to higher education. The only problem is that effort can only be measured for high school students in District 1, for which school records are available. In the next section, we present correlations among all the cultural variables. The correlation of .66 between effort and self-reported GPA suggests that GPA is probably a pretty good proxy for effort. Self-reported GPA is available for the full UW-BHS sample of high school seniors. We have also included homework hours (outside of school) as a second behavioral measure of cultural expressions, or “culture in action.” The variable homework hours is measured by responses to a survey question that asked, “Overall, about how much time do you spend on homework each week outside of school?” Careful readers will notice that we have taken the two intervening variables identified in chapter 5, GPA and encouragement, and packaged them into a more ambitious formulation of how families and communities potentially shape educational outcomes through socialization, parental coaching, peer influences, and social processes. Our formulation of cultural variables in figure 6.1 is clearly a first approximation of the effect of culture on educational outcomes. We do not claim that all aspects of culture are presented here or are captured by the measured variables. Our aim is to present a preliminary model that allows for empirical assessment of the indirect (mediation) role of cultural variables and also for the direct effects of cultural context, cultural orientations (attitudes), and cultural expressions (behaviors) on educational transitions in the College Pathways Model.

Correlations Among Cultural Variables Table 6.1 presents the bivariate correlations among all the cultural variables, including context, orientations, and expressions, as outlined in figure 6.1. Because these variables are a mix of indexes and single items, and are generally measured on ordinal scales, we are looking at the signs of correlations and rough orders of magnitude. The first observation is that

.19* .39*

.18* .27* .18*

Cultural expressions Homework hours Self-reported GPA Effort

1.00 .46* .30* .21* .23*

Cultural orientations School is important Locus of control

Cultural context Parental comm and support Parents know friends Encouragement index Family expected college Friends’ college plans

Comm and Support

.07* .10* .08*

.12* .20*

1.00 .21* .13* .17*

Know Friends

.21* .30* .25*

.19* .18*

1.00 .41* .35*

Encouragement

.19* .23* .16*

.16* .10*

1.00 .25*

Expect College

.26* .32* .18*

.07* .18*

1.00

Friends’ Plans

.19* .34* .34*

1.00 .18*

School Important

Correlation Matrix

.17* .27* .17*

1.00

Locus of Control

1.00 .30* .18*

Homework

1.00 .66*

Self GPA

Table 6.1     Correlations Among Cultural Variables (Context, Orientations, and Expressions) for UW-BHS High School Seniors

1.00

Effort

8,573 8,645

8,386 8,577 3,635

Cultural orientations School is important Locus of control

Cultural expressions Homework hours Self-reported GPA Effort 8,354 8,542 3,613

8,527 8,598

8,600 8,586 8,530 8,546

Know Friends

8,377 8,570 3,626

8,555 8,627

8,632 8,558 8,576

Encouragement

8,319 8,506 3,626

8,608 8,639

8,640 8,520

Expect College

8,337 8,524 3,610

8,520 8,589

8,591

Friends’ Plans

8,320 8,506 3,624

8,639 8,639

School Important

8,386 8,577 3,656

8,713

Locus of Control

8,390 8,342 3,533

Homework

8,581 3,610

Self GPA

3,659

Effort

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample of UW-BHS students includes all respondents with nonmissing data on college aspirations, college expectations, college preparedness, and each cultural variable. GPA = grade point average. UW-BHS = University of Washington-Beyond High School. WASL = Washington Assessment of Student Learning. Effort = Self-reported GPA - predicted GPA. Predicted GPA = 2.07 + .00646 (WASL math score) + .00613 (WASL reading score).    The coefficients in the prediction equation were obtained from: GPA (self-reported) = a + b1(WASL math score) + b2 (WASL reading score) + µ    Predicted GPA is only available for students in district 1. *p < .05

8,648 8,600 8,628 8,576 8,587

Cultural context Parental comm and support Parents know friends Encouragement index Family expected college Friends’ college plans

Comm and Support

Number of Observations for Each Correlation

206   From High School to College

all of the cultural variables are positively, but only moderately, correlated. UW-BHS high school seniors who have a high (or low) value on one of the measured dimensions of context, orientations, and expressions are likely to have high (or low) values on other dimensions. The correlations among the cultural-context cluster of variables vary from .1 to .4. Parental communication and support (frequent interaction of students and their parents) is highly correlated (.46) with parents know friends (friends of the student and also their parents). Parents who are engaged with their child in one dimension are, not too surprisingly, engaged in other dimensions. The other relatively high correlation among the cultural context variables is the association (r = .42) between encouragement index and family expected college (“my family always expected me to go to college”). Encouragement index is also moderately highly correlated (.30) with parental communication and support and (.35) with friends’ college plans (“most or all friends are planning to attend a four-year college”). The two dimensions of cultural orientations—school is important (“how well I do in school is an important part of who I am as a person”) and locus of control—are only weakly correlated at .17. The relatively low correlation between locus of control and school is important may be due, in part, to the generality of locus of control compared to the specificity of school achievement as a part of self-identity. Locus of control is more highly correlated (.39) with parental communication and support. Perhaps a high level of positive interaction with parents provides adolescents with a sense of confidence and a feeling of being in charge of their lives. The variables representing cultural expressions are moderately correlated. The highest correlation, at .66, is between GPA (self-reported grades) and effort (self-reported GPA minus predicted GPA based on test scores). Effort is constructed to capture overperformance or underperformance in school relative to test taking. Effort is our attempt to tap the theoretical concept of striving—highly motivated behavior which is often hypothesized to vary by gender, race and ethnicity, and immigrant generation. For the purposes of this study, effort would be the preferred variable to test the hypothesis that socialization (from family, community, and peers) mediates educational disparities by ascription and social origins. However, effort can only be measured for students in District 1, about 40 percent of the total UW-BHS sample. Given this limitation, we use self-reported GPA as a proxy for effort. The high correlation (.66) between the two variables (in District 1) indicates that our assumption is partially justified. GPA actually has higher bivariate correlations with the other cultural variables than effort. This pattern suggests that GPA is tapping other dimensions in addition to effort. For example, the high correlation of GPA with parental communication and support may be predictive of “coaching” that improves academic performance. Similarly, other societal patterns such as homogenous friendships or tracking (by ability or test scores) may contribute to

Explaining Educational Transitions   207

the higher correlations between GPA and parents know friends and friends’ college plans than with effort. There is a positive, but modest, correlation (.30) between GPA and homework hours (outside of regular school hours)—the other measure of cultural expression. Note that the association is even lower (.16) between effort and homework hours. The standard assumption is that those who put in extra study time will get higher grades (or higher than they would be expected to get). While there is a positive relationship, it is not as strong as the conventional wisdom would suggest. Indeed, it is possible that homework hours are partially endogenous to ability. Some very capable students may be able to do well in school without investing much time or effort.25 Nonetheless, we consider homework hours and self-reported GPA as good indicators of the expression of between-group cultural practices (parental coaching, student effort) that may mediate the impact of ascription on educational outcomes. The underlying model, expressed in figure 6.1, assumes that cultural context (from family and peers) has a strong influence on cultural orientations (beliefs of students), and that both domains have strong influences on cultural expressions (hard work represented by homework hours and academic performance represented by GPA [effort]). The model is confirmed by the overall pattern of statistically significant positive coefficients in table 6.1. Perhaps the correlations would be a bit higher with more reliable measurement of these complex social psychological attributes. All of the correlations are in the expected direction, but the socializing influences of parenting and peers on beliefs and behaviors should not be exaggerated. The associations might be best described as moderate, with correlations from .2 to .3. Cultural dimensions are related to one another, but they are not a tightly wound system, much less a deterministic one.

Associations Between Ascription and Cultural Context, Orientations, and Expression Before examining the role of cultural variables in the explanation of pathways to college graduation, we examine the connections between ascription and cultural patterns. Culture is produced and reproduced through intergenerational socialization in families and communities, as well as through recurring social interaction with neighbors, friends, coworkers, and participants in formal and informal associations. To the extent that the three ascriptive variables—gender, race-ethnicity, and immigrant generation— capture important social divisions that correspond with socializing and interacting groups and communities, we might be able to identify the structural roots of differential cultural influences. Table 6.2 shows a cross-classification of the three ascriptive groups down the rows and the ten variables representing cultural contexts, cultural (Text continues on p. 212.)

2.06

1.98 2.11

2.11 2.05 2.01 1.86 1.99 1.66 1.71 1.97 1.86 2.03 2.03 2.00

1.86 2.00 2.13

8,648 0 to 3

Total population

Gender Male Female

Race-Ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic

Immigrant generation First generation Second generation Third-and-higher generation

(N) Range

x¯

-0.12* -0.05* 0.16*

0.13* 0.00 -0.01 -0.15* -0.03* -0.12* -0.11* -0.02* -0.05* -0.01 -0.01 -0.02

0.11*

r

Communication and Support

8,600 0 to 3

1.57 1.70 1.80

1.78 1.81 1.63 1.58 1.64 1.52 1.43 1.70 1.53 1.73 1.70 1.68

1.66 1.80

1.74

x¯

-0.08* -0.02* 0.10*

0.06* 0.03* -0.02 -0.09* -0.03* -0.05* -0.07* -0.01 -0.04* 0.00 -0.01 -0.01

0.09*

r

Parents Know Friends

8,632 0 to 5

4.02 3.88 3.71

3.69 3.58 3.33 4.12 4.22 3.92 4.36 4.03 3.88 3.76 3.65 3.65

3.50 3.93

3.74

x¯

0.07* 0.04* -0.02

-0.04* -0.04* -0.03* 0.11* 0.08* 0.02 0.07* 0.32* 0.01 0.00 -0.01 -0.01

0.14*

r

Encouragement Index

Cultural Context

8,640 0 to 3

2.41 2.46 2.29

2.28 2.29 1.99 2.56 2.62 2.36 2.65 2.57 2.55 2.42 2.19 2.28

2.27 2.37

2.32

x¯

0.04* 0.08* -0.05

-0.07* -0.02 -0.05* 0.13* 0.09* 0.01 0.07* 0.05* 0.04* 0.02 -0.03* -0.01

0.06*

r

Family Expected College

8,591 0 to 4

1.84 2.17 2.04

2.05 1.67 1.65 2.24 2.58 1.68 2.13 2.32 2.13 1.73 1.45 1.89

1.84 2.12

2.00

x¯

-0.04* 0.05* 0.04*

0.05* -0.09* -0.03* 0.07* 0.10* -0.04* 0.02 0.04* 0.01 -0.03* -0.07* -0.01

0.09*

r

Friends’ College Plans

Table 6.2    Mean Values of Cultural-Context, Cultural-Orientations, and Cultural-Expressions Variables and Correlations with Gender, Race-Ethnicity, and Immigrant Generation Among UW-BHS High School Seniors

2.00

1.82 2.14

1.92 2.08 1.81 2.16 2.06 2.30 2.19 2.10 2.28 2.25 2.17 2.05

2.22 2.08 1.95

8,639 0 to 3

Total population

Gender Male Female

Race-Ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic

Immigrant generation First generation Second generation Third-and-higher generation

(N) Range

x¯

0.09* 0.04* -0.07*

-0.11* 0.04* -0.03* 0.08* 0.02 0.06* 0.04* 0.02 0.05* 0.04* 0.04* 0.01

0.19*

r

School Is Important

8,713 0 to 3

1.93 2.08 2.13

2.12 2.08 2.05 1.98 2.03 1.90 1.88 2.06 1.97 1.99 2.07 2.05

2.03 2.13

2.09

x¯ r

-0.13* 0.00* 0.14*

0.10* -0.01 -0.01 -0.11* -0.03* -0.07* -0.09* -0.01 -0.04* -0.03* 0.00 -0.01

0.11*

Locus of Control

Cultural Orientations

8,390 0 to 10

3.57 3.34 2.93

3.03 2.47 2.49 3.60 3.64 3.12 4.40 3.17 3.50 3.00 2.79 2.61

2.52 3.42

3.02

x¯

0.07* 0.05* -0.04*

0.00 -0.08* -0.02* 0.09* 0.05* 0.01 0.09* 0.01 0.02* 0.00 -0.02 -0.02*

0.16*

r

Homework Hours

8,581 1.72 to 3.62

2.99 2.99 2.98

3.01 2.77 2.85 3.04 3.10 2.88 3.16 2.99 2.97 2.81 2.83 2.86

2.86 3.05

2.97

x¯

0.02 0.02* 0.02*

1.06* -0.15* -0.03* 0.07* 0.07* -0.03* 0.07* 0.01 0.00 -0.04* -0.05* -0.03*

0.19*

r

Self-Reported GPA

 3,659 -2.35 to 1.74

0.21 0.00 -0.01

0.01 -0.09 -0.13 0.12 0.02 0.04 0.39 -0.01 0.07 -0.05 0.03 0.00

-0.15 0.13

0.01

x¯

Effort

0.14* 0.00 -0.05*

0.00 -0.08* -0.04* 0.10* 0.01 0.02 0.16* 0.00 0.02 -0.02 0.01 0.00

0.28*

r

(Table continues on p. 210.)

Cultural Expressions

8,648

3,849 4,799

5,277 1,165 133 1,399 494 240 264 241 160 173 296 205

Gender Male Female

Race-Ethnicity White African American American Indian Asian American   East Asian   Cambodian   Vietnamese   Filipino   Other Asian Pacific Islander Mexican Other Hispanic

Comm and Support

Total population

Table 6.2    (Continued)

5,254 1,149 133 1,392 494 240 263 240 155 173 296 203

3,821 4,779

8,600

Know Friends

5,262 1,166 133 1,399 494 240 264 241 160 173 295 204

3,841 4,791

8,632

Encouragement

5,260 1,161 134 1,407 497 243 265 242 160 173 299 206

3,854 4,786

8,640

Expected College

Cultural Context

5,242 1,157 131 1,392 491 240 262 240 159 172 294 203

3,826 4,765

8,591

Friends’ Plans

5,261 1,167 134 1,401 496 240 264 242 159 173 298 205

3,852 4,787

8,639

School Important

5,306 1,178 136 1,413 499 243 266 243 162 173 300 207

3,887 4,826

8,713

Locus of Control

Cultural Orientations

Number of Observations for Each Correlation

5,140 1,116 127 1,362 483 234 258 235 152 166 286 193

3,708 4,682

8,390

Homework Hours

5,236 1,152 133 1,393 492 240 263 238 160 172 292 203

3,819 4,762

8,581

SelfReported GPA

Cultural Expressions

2,198 559 66 567 150 143 146 72 56 67 122 80

1,582 2,077

3,659

Effort

1,013 1,578 5,405

1,012 1,573 5,398

1,013 1,578 5,396

1,016 1,577 5,389

1,005 1,566 5,379

1,010 1,573 5,396

1,020 1,585 5,440

981 1,522 5,267

1,009 1,571 5,363

386 641 2,350

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample of UW-BHS students includes all respondents with nonmissing data on college aspirations, college expectations, college preparedness, and each cultural variable. GPA = grade point average. UW-BHS = University of Washington-Beyond High School. WASL = Washington Assessment of Student Learning. Effort = Self-reported GPA - predicted GPA. Predicted GPA = 2.07 + .00646 (WASL math score) + .00613 (WASL reading score).    The coefficients in the prediction equation were obtained from: GPA (self-reported) = a + b1(WASL math score) + b2 (WASL reading score) + µ    Predicted GPA is only available for students in district 1. *p < .05

Immigrant generation First generation Second generation Third-and-higher generation

212   From High School to College

orientations, and cultural expressions arrayed across columns. In each cell are two summary indicators: the mean value of the cultural variable for each ascriptive group, and the correlation of each cultural variable (scored as a scale) with each ascriptive group (scored as a 0, 1 binary variable). The two summary indicators tell a similar story for each relationship. The major finding from table 6.2 is that there is not a consistent alignment of cultural traits by ascriptive groups. It seems that the race-ethnicity and immigrant-generation groups specialize in different cultural traits. We begin with a discussion of well-defined cultural patterns by gender and then turn to the much wider variation in cultural traits by race-ethnicity and immigrant generation. Women express more proschooling attitudes and behaviors; this is what might be expected based on their academic performance in school relative to their male peers. Somewhat more surprising is that women also report more positive values on the cultural-context variables than men. Given that men and women are reared in the same families (which are assumed to share some common values and attitudes), why do women report higher proportions of engaged, supportive, and encouraging parents relative to their brothers? One plausible interpretation is that the parents create different social milieus for their daughters than their sons. Perhaps parents, especially mothers, speak more with their daughters than with their male offspring and are more likely to get to know their daughters’ friends. Females may also receive more encouragement (from parents and friends) to go to college because they are already on track to do so with higher grades. Regardless of the precise mechanism, female students, on average, report much more support from families and peers to attend college than do male students. Cultural variations by race-ethnicity and immigrant generation reveal more complex and nuanced patterns. Relative to other racial-ethnic groups, white students in the UW-BHS sample report above-average levels of parental communication and support and are slightly more likely to have friends who plan to attend a four-year college. However, the reported levels of encouragement (encouragement index) and parental college expectations ( family expected college) by white students are a bit below average. White students express an above-average sense of control over their lives (internal locus of control) and get above-average grades (GPA), but no more than would be expected based on their measured test scores. White students are about average in terms of their reported number of homework hours and are a bit below average in proschooling attitudes (school is important). A closer look shows that the majority of white students agreed with the question, “How well I do in school is an important part of who I am as a person,” but they were slightly less likely to “strongly agree” with the statement. There is heterogeneity among Asian students, with East Asians and Vietnamese at one end of the continuum and Cambodians at the other

Explaining Educational Transitions   213

end. But compared to other racial and ethnic groups, Asian Americans are more similar to each other than they are to other racial-ethnic groups. For example, Asian students rank higher in proschool attitudes, GPA and effort (overperforming relative to their test scores), and amount of time spent on homework. Asian students also report higher levels of encouragement, experience higher family expectations to attend college, and have more friends who plan to go to college. However, their family cultural environments are not entirely positive. Asian students express below-average levels of parental communication and support. In contrast to Coleman’s hypothesis of above-average levels of social capital among Asian immigrants, the parents of Asian students are less likely to know their friends and the parents of their friends.26 Asian students are also less likely to feel a sense of personal control over their lives (locus of control). These findings are consistent with other research which finds that Asian students experience less supportive families and are more likely to report depression and low self-esteem.27 Immigrant students, especially the first generation, report having similar cultural contexts, orientations, and expressions as Asian students. Immigrant parents are relatively unengaged in the lives of their adolescent children (as measured by parental communication and support) but, nonetheless, hold high expectations for their children’s educational attainment. First-generation students report above-average levels of proschool attitudes and time spent doing homework. The GPAs of first-generation immigrant students are about average for the population but are considerably higher than would be predicted by test scores alone. Third-and-higher-generation students are the opposite— they report high levels of parental support but below-average expectations and encouragement for college. They are less invested in schooling (school is important), do less homework (homework hours), and do not perform up to their potential ability (effort). With some exceptions, other minority groups—including African American, American Indian, Pacific Islander, Mexican, and Other Hispanic students—have a less supportive cultural context and report lower levels of academic performance. The differences are not huge, but they show a consistent pattern. Disadvantaged minorities have fewer high school friends planning to go to four-year colleges, have less parental communication and support, and have slightly lower levels of family encouragement to attend college. The one exception to this pattern is with proschooling attitudes; minorities are more likely to report that they “strongly agree” that doing well in school is very important to their identity. The question that remains, however, is how do these cultural variables fit with the patterns of progression up the educational ladder (the College Pathways Model) and the socioeconomic determinants of college aspirations, expectations, preparation, enrollment, and completion?

214   From High School to College

Cultural Influences on Transitions in the College Pathways Model So far, we examined the relationships among the cultural variables and whether cultural repertoires are “embedded” in particular ascriptively defined communities. From table 6.1, we learned that cultural context influences but does not determine cultural orientations (beliefs) and cultural expressions (behaviors). Parental engagement and encouragement appear to push students to express more proschool attitudes and to work harder in school. But the correlations are generally modest—around .2 or .3. Having strong family support (a favorable cultural context) is more common for white and Asian students than for other minority groups, but there was not a monolithic model of what supportive families do. White students are more likely to have engaged parents who provide emotional support and frequent communication and who know their friends. White students are also more likely to have friends who are going to four-year colleges. Asian students, in general, report that their parents were less likely to be engaged and supportive, but they were much more likely to feel encouraged and expected to go to college. Asian students were also much more likely to have peers going to college. Other minorities, with some exceptions, had much lower levels of family support, had fewer friends going to college, and were less likely to be encouraged and expected to go to college. However, African American and other disadvantaged minorities report above-average proschool attitudes. But do these cultural characteristics predict educational success? In general, the cultural variables have positive associations with educational achievement. This conclusion is based on appendix table 6.A2, available online, which shows educational-transition ratios for each of the nine cultural variables (five cultural-context, two cultural-orientations, and two cultural-expressions variables).28 These bivariate relationships show that high school seniors who are reared in supportive and encouraging families, express proschool attitudes, and invest in and do well in school are more likely to advance up the educational ladder than those who do not. There are, however, wide variations in the strength of cultural influences on different rungs of the ladder. Parental influences (measured by the variables parental communication and support and parents know friends) are more important on the first three transitions along the college pathway (measured in high school) than on the transitions to college enrollment and college completion. There are also signs of thresholds, with fairly modest effects of additional cultural resources for students in the middle to higher range. However, cultural deficits matter—students with the lowest levels of cultural resources are much less likely to stay on the college track.

Explaining Educational Transitions   215

In terms of getting on the college track in high school, encouragement and family expectations make a huge difference. Or to put it the other way around, students of whom little is expected are not likely to prepare for college in high school. At the higher end of the College Pathways Model, encouragement and family expectations have only modest effects on college enrollment and completion. The one cultural-context variable that substantially influences college enrollment and completion is friends’ college plans (the number of friends who are planning to attend a fouryear college). There may be some reverse causation in this relation­ship because college-bound students may choose to befriend peers with similar plans. Nonetheless, we believe that peer influences are an important mechanism that supports and reinforces college planning and enrollment. Cultural orientations (attitudes) and cultural expressions (behaviors) have consistently positive effects at each step on the educational ladder. The attitudinal variables, which tap proschool attitudes (school is important) and internal locus of control, are positively associated with college entry and completion, but the effects are small and are primarily consequential in the lower range of the variable. The measures of working hard and doing well in school—homework hours, GPA, and effort—have larger and more pervasive effects at all transitions. Thus far, the results show that cultural variables are moderately correlated with each other, have modest ties with ascription, and have modest consequences for ascent up the educational ladder. These results, however, need to be integrated with all the background variables introduced in chapter 5. The question is whether the cultural variables, outlined in figure 6.1, play a role in explaining differential outcomes by gender, raceethnicity, and immigrant generation, net of social origins.

Multivariate Analysis of the College Pathways Model with Intervening Cultural Influences Figure 6.2 is our hypothesized model of how cultural variables mediate and add explanatory power to educational outcomes in the College Pathways Model. In this model, the ascriptive variables (gender, raceethnicity, and immigrant generation) and social origins are considered exogenous to the cultural variables. We assume that the socialization of children across race-ethnicity (and immigrant-generation) groups are captured, at least in part, by the variables measured within the rubrics of cultural context, cultural orientations, and cultural expression. Similar to the analysis in chapter 5, changes in net gender, race-ethnicity, and immigrant-generation odds ratios across models allow for an interpretation of how cultural variables mediate disparities in educational outcomes.

216   From High School to College Figure 6.2    A Revised College Pathways Model with Cultural Context, Cultural Orientations, and Cultural Expressions Ascription Gender Race-ethnicity Immigrant generation

Cultural context

Cultural orientations

College-pathway transitions

Cultural expressions

Social origins

Source: Author’s compilation.

The logistic regression models presented in this chapter follow the analytical strategy introduced in chapter 5. The first two models are the same—only ascription variables are included in model 1, followed by model 2, which includes the ascription and social-origins variables. Then, in cumulative fashion (following the sequential logic of figure 6.2), cultural-context variables are added in model 3, cultural-orientations variables in model 4, and cultural–expressions variables in model 5. These analyses address two major questions. First, does culture (as measured here) mediate the inequality in educational outcomes between populations defined by ascription variables and social-origins variables (socioeconomic status and family background)? Second, do cultural variables have direct effects on educational outcomes that are independent of the background variables? Because of the volume of results, the complete results of the five models for the first step in the pathway—college aspirations—are presented in a table in this volume (table 6.3) and the balance (appendix tables 6.A3 to 6.A7) are available online.29 The bulk of the most important findings are summarized in figures 6.3 to 6.8 in this volume. The figures contain only the effects (odds ratios) of the three ascriptive variables (gender, race-ethnicity, and immigrant generation) for three of the five models presented in the tables:

Explaining Educational Transitions   217

models 2, 3, and 5. Model 2 is the baseline model with the ascriptive and social-origins variables as independent variables. Model 3 adds the cultural-context variables, and model 5 includes all three sets of cultural variables. Our interpretation draws upon the summary results in the figures and also from the more detailed results presented in the online tables.

College Aspirations Before examining the role of cultural variables as potential mediators of ascriptive inequality on educational transitions, it is important to acknowledge the very significant direct impact of the cultural variables on college aspirations. This impact is shown in the net effects of cultural context, cultural orientations, and cultural expressions, which are introduced as covariates in models 3, 4, and 5, respectively, in table 6.3. The introduction of the five cultural-context variables in model 3 more than doubles the variance explained in college aspirations from model 2 to 3 (9.3 percent to 22.8 percent) and the Bayesian information criterion (BIC) model-fit statistic declines by more than 1,200 points. Parental communication and support has a net positive impact on aspirations, but its influence is via student attitudes and behavior (see change in odd ratios from models 4 and 5). In contrast, the effects of encouragement index, family expected college, and friends’ college plans are direct—in other words, they are not mediated by student attitudes and behaviors. Cultural orientations (school is important and locus of control) and cultural expressions (GPA and homework hours) have very strong and significant positive effects on college aspirations. A peculiar finding is that the bivariate positive effect of parents know friends on college aspirations becomes negative when other culturalcontext variables are included in model 3. Recall that the parents know friends variable is positively correlated (.46) with parental communication and support (see table 6.1). We suspect that high levels of parents know friends, net of other aspects of parental engagement, may be indicative of high-risk situations—for example, when parents are worried that peer influences are pulling their children in the wrong directions. Another important finding from table 6.3 is that a substantial share of the significant effects of social origins on college aspirations is mediated by the cultural-context variables (see changes from model 2 to model 3). This pattern, which is quite common in many of the tables in this chapter, means that socioeconomic advantages elevate their children’s aspirations by providing a favorable cultural context. We still see highly significant direct effects of parental education and occupational status, which are not explained by the three clusters of cultural variables. In contrast to the “either-or” story—whether it is really all about structure or culture— considerable overlap in explanatory variance exists in addition to the direct effects of both culture and structure on college aspirations.30 (Text continues on p. 221.)

Social-origins variables Father educ (HS or less omitted)   Some college, no degree   College degree or above   No father or DK

Ascriptive variables Gender (male omitted)  Female Race-ethnicity (white omitted)   African American   American Indian   East Asian  Cambodian  Vietnamese  Filipino   Other Asian   Pacific Islander  Mexican   Other Hispanic Immigrant generation (third omitted)   First generation   Second generation 1.39*** 0.97 3.40*** 1.47* 3.11*** 1.25 1.45 0.96 1.10 1.01 0.90 1.24**

0.93 0.60** 3.52*** 0.74 1.84** 1.33 1.16 0.67* 0.67** 0.81 0.66*** 1.15

1.35*** 2.13*** 1.09

1.29***

Model 2

1.25***

Model 1

1.17* 1.65*** 1.20

0.78* 1.17

1.22* 0.97 2.26*** 0.92 1.52* 0.91 0.97 0.76 1.06 0.91

0.96

Model 3

1.18* 1.69*** 1.18

0.80 1.18

1.20* 1.02 2.34*** 0.90 1.57* 0.89 0.94 0.76 1.03 0.90

0.90

Model 4

1.14 1.61*** 1.17

0.71** 1.12

1.42*** 1.05 2.43*** 0.94 1.24 0.98 1.00 0.83 1.17 1.06

0.77***

Model 5

Exponentiated Odds Ratios Relative to the Omitted Category for Each Variable

Table 6.3    Logistic Regression of the Probability of College Aspirations Among UW-BHS High School Seniors with Cultural Variables

Cultural variables Cultural context   Parental communication and support   Parents know student’s friends   Encouragement index (minus teacher)   Family always expected college   Friends have college plans Cultural orientations   School is important   Locus of control

Mother educ (HS or less is omitted)   Some college, no degree   College degree or above   No mother or DK Father SEI (not working, NA, DK omitted)   Father employed   Father SEI (employed × SEI) Mother SEI (not working, NA, DK omitted)   Mother employed   Mother SEI (employed × SEI) Home ownership (renting omitted)   Family owns home Family structure (not intact omitted)   Lives with biological or adoptive parents 0.86 1.006** 0.85 1.005* 1.02 0.97

0.87 1.009*** 0.78* 1.007*** 1.19* 1.16*

1.23** 0.79*** 1.43*** 1.39*** 1.49***

1.14 1.75*** 1.14

1.27*** 2.26*** 1.02

1.00 1.51***

1.16*** 1.76***

(Table continues on p. 220.)

0.94 0.83*** 1.40*** 1.35*** 1.40***

0.95

1.03

0.84 1.006**

0.88 1.005**

1.12 1.67*** 1.18

1.04 0.79*** 1.42*** 1.38*** 1.48***

0.98

1.03

0.83 1.005**

0.84 1.006**

1.14 1.78*** 1.17

2.85*** 0.015 9,545 8,719

Model 1

0.74** 0.093 8,879 8,719

Model 2

0.14*** 0.228 7,623 8,719

Model 3

0.05*** 0.236 7,558 8,719

Model 4

0.01*** 0.262 7,327 8719

2.07*** 1.15***

Model 5

Exponentiated Odds Ratios Relative to the Omitted Category for Each Variable

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: The sample of 8,719 UW-BHS students includes all respondents with nonmissing data on college aspirations, college expectations, and college preparedness. DK = don’t know. NA = not available. BIC = Bayesian information criterion. SEI = socioeconomic index. GPA = grade point average. HS = high school. UW-BHS = University of Washington-Beyond High School. *p < .05; **p < .01; ***p < .001

Constant McFadden’s adjusted R-squared BIC index Number of observations

Cultural expressions  GPA   Homework hours

Table 6.3    (Continued)

Explaining Educational Transitions   221

We now turn to the role of cultural variables as mediators of the ascriptive variables, which is summarized in figure 6.3. Chapter 5 showed a relatively simple story with respect to gender, in contrast to the complexity of results for race-ethnicity and immigrant generation. In general, females were more successful in the college-pathway transitions than their male peers because of their superior academic performance in high school. The results in figure 6.3 provide a new wrinkle. In model 3, which only adds cultural context (but not GPA), the gender gap in college aspirations dis­ appears. It seems likely that the friends’ college plans variable plays a critical mediating role (see table 6.3). In the saturated model 5, when GPA and homework hours are included as covariates, women have college aspirations significantly lower than their male peers. Students are likely to form friendships based, at least in part, on their academic orientations. The friends’ college plans variable, which captures part of gender differences in GPA, is sufficient to explain the higher college aspirations of female students. Female students who do not have an above-average GPA and are not friends with many students planning to attend college are less likely to aspire to graduate from college than comparable male peers. Before considering the role of culture in explaining racial-ethnic and immigrant-generation disparities in college aspirations, it is important to recall two of the major conclusions from chapter 5. For disadvantaged racial and ethnic groups, socioeconomic inequality, indexed by social origins, was the primary obstacle that explained their lower college aspirations relative to white students. This pattern is evident in the baseline model 2 in figure 6.3. Net of social origins, there are no significant racial and ethnic disadvantages in college aspirations. In fact, African American and second-generation students register slightly higher college aspirations than do white students and third-and-higher-generation students, respectively. Asian American students, especially East Asians and Vietnamese, are considerably more ambitious than white students. The addition of the cultural-context variables in model 3 shows that a more favorable cultural environment actually favors minorities and immigrants relative to white and third-generation students. The immigrantgeneration gap and some of the Asian American advantages are the result of a more favorable cultural context—primarily higher levels of encouragement, family expected college, and peers going to college. Most of the changes for other racial and ethnic groups are not statistically significant, but we do see a general pattern. Almost all groups have lower college aspirations (relative to whites) when the cultural-context variables are held constant in model 3 relative to model 2. This means that the minority cultural context (net of social origins) is an asset, not a liability. Model 5 does not show that the net effects of cultural orientations or cultural expressions play a significant role in explaining the gaps in college aspirations between disadvantaged minorities and their white peers.

0.77

2 3 5 Female

1.29

1.42

1.22 10

2 3 5 African American

1.39

2 3 5 American Indian

2.26

2.43

2 3 5 East Asian

3.40

1.52

2 3 5 2 3 5 Cambodian Vietnamese

1.47

3.11

2 3 5 Other Asian

2 3 5 Pacific Islander

2 3 5 Mexican

2 3 5 Other Hispanic

0.71

2 3 5 1st generation

0.78

1.24

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on table 6.3. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

Figure 6.3    Logistic Regression of the Probability of College Aspirations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of UW-BHS High School Seniors with Cultural Variables

Regression coefficient

Explaining Educational Transitions   223

There are two important caveats to this interpretation. The detailed results presented in table 6.3 show that cultural context plays a significant role in mediating the positive impact of family SES and family structure on college aspirations. One of the reasons why higher-SES students have higher college aspirations is that high-SES families are more encouraging, supportive, and engaged in the lives of their children. Disadvantaged minorities may have fewer cultural resources, but this is because of their lower socioeconomic backgrounds, not because of their ethnic or community cultures. One final point is to recall the finding from chapter 4 that college aspirations play only a small role in explaining inequality among ascriptive groups in eventual college graduation rates.

College Expectations At first glance, the patterns for college expectations appear very similar to the patterns for college aspirations. For example, neither college aspirations nor college expectations play a major role in explaining inequality in college graduation rates among ascriptive groups (see chapter 4). The limited impact of college expectations is partially due to the narrow range of variation in the variable. Overall, about 90 percent of high school seniors who aspire to graduate from college expect to do so. The male-female gap is only three percentage points, and there are only slight variations across immigrant generations. Wider variations exist, however, among the racial-ethnic communities, ranging from the 70-to-low-80-percent range for disadvantaged minority groups, to the low-90-percent level for whites and Asians. We examine the role of cultural factors as potential mediators of these differences more closely in figure 6.4 and in appendix table 6.A3, available online.31 The model-fit statistics in appendix table 6.A3 show that college expectations are much more strongly tied to family socioeconomic background and to cultural influences than to race-ethnicity and other ascriptive variables. This finding illustrates a recurrent finding that variables of social class (another term for social origins) are generally more highly associated with educational outcomes than are the ascriptive variables of gender, race-ethnicity, and immigrant generation. Only 2 percent of the total variation in the transition to college expectations is explained by ascription, but a full 22 percent is explained in the saturated model, with the lion’s share attributable to social origins and cultural context. Most of the effects of social origins, along with high encouragement, family expected college, friends’ college plans, and GPA are direct. As previously mentioned, women hold modest advantages relative to men at all stages of the College Pathways Model, including having higher college expectations given aspirations. Females grow up in the same families as do men, and there are no major gender differences in socioeconomic

0.76

2 3 5 Female

1.27

2 3 5 African American

2 3 5 American Indian

2 3 5 East Asian

2 3 5 2 3 5 Cambodian Vietnamese

0.60 0.58

2.51

2 3 5 Other Asian

0.37 0.32

2 3 5 Pacific Islander

0.36

2 3 5 Mexican

0.69 0.64

2 3 5 Other Hispanic

2 3 5 1st generation

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

0.55 0.54

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 6.A3. UW-BHS = University of Washington-Beyond-High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 6.4    Logistic Regression of the Conditional Probability of College Expectations on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Aspiring UW-BHS High School Seniors with Cultural Variables

Regression coefficient

Explaining Educational Transitions   225

background and family structure. However, females report a more favorable cultural context (more encouragement, more family engagement, more friends planning to go to college) than males, and women report more positive proschool attitudes, a higher internal locus of control, and higher GPAs than their male peers. These cultural factors “explain” why women have higher expectations, given aspirations, than men. In other words, women are more likely to express confidence that they will achieve their educational goals because they tend to be enmeshed in more supportive environments, express more positive attitudes, and invest more time and effort in schooling outcomes. As we will see, this gender story is repeated in subsequent analyses of college preparation, college enrollment, and college completion. In general, there appears to be more of a problem of low college expectations for minorities (relative to white students) than low college aspirations. Net of social origins, only one Asian group—Vietnamese—has significantly higher college expectations than white students. This advantage is due to a more favorable cultural context among Vietnamese students. Net of SES (and cultural context) many minority groups, including Cambodians, Filipinos, Pacific Islanders, and Mexican Americans, have significantly lower college expectations than white students, given aspirations. There appears to be some residual “minority” disadvantage that persists, especially when the cultural variables are included in models 3 and 5. The gaps are relatively modest, and some are not statistically significant, but there does appear be some evidence that “dashed” college expectations— wanting to go to college, but not expecting to—affects most American minority groups. Although we cannot identify the precise reasons for these dashed expectations, minorities may face more obstacles and have fewer resources (beyond those measured here) than white students.

College Preparation Of the three stages of the College Pathways Model measured in high school (aspirations, expectations, and preparation), college preparedness is the most consequential. College preparation is indexed by several key behaviors—taking AP courses, taking college entrance exams (SAT or ACT), and applying to a four-year college before the spring of the senior year in high school. Only two-thirds of students who both aspired and expected to graduate from college actually followed through by taking two of these three steps. These actions reflect serious efforts to get ready for higher education. In terms of explaining gender and racial-ethnic gaps in college graduation, variations in preparation had considerably more impact than differentials in aspirations and expectations (see table 4.4). Following the precedent of prior college-pathway transitions, figure 6.5 and appendix table 6.A4 (available online)32 show the five models

1.18

0.88

2 3 5 Female

1.35

2 3 5 African American

1.56

2 3 5 American Indian

2.53

2.82

2 3 5 East Asian

3.13

1.85

2 3 5 2 3 5 Cambodian Vietnamese

1.64

2 3 5 Other Asian

2 3 5 Pacific Islander

2 3 5 Mexican

2 3 5 Other Hispanic

2 3 5 1st generation

0.61

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

0.730.73

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 6.A4. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

Figure 6.5    Logistic Regression of the Conditional Probability of College Preparedness on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Expecting UW-BHS High School Seniors with Cultural Variables

Regression coefficient

Explaining Educational Transitions   227

predicting college preparation. As noted in chapter 5, there are strong social-origins effects on college preparation. Variance explained jumps from 2 to 7 percent from model 1 to model 2, when the social-origins variables are added as covariates predicting college preparation. Parental educational attainment, occupational attainment, and home ownership are strong and significant predictors, whose impacts are largely direct. Family encouragement and expectations are important predictors, along with number of friends attending college. Variance explained increases from 7 to 14 percent from model 2 to model 3. GPA appears to be the most important predictor of college preparation—variance explained jumps from 15 to 22 percent from model 4 to model 5. Moreover, GPA serves to mediate the positive effects of proschool attitudes and locus of control on college preparation (see appendix table 6.A4). We have learned that cultural factors, in addition to higher academic performance, explain the gender gap in the college-pathway transitions. High school females are more likely than their male counterparts to be encouraged and expected to go to college by their families and have more friends planning to go to college. They are also more likely to express proschool attitudes and to feel in charge of their destiny (internal locus of control). However, as noted earlier, these factors are not completely independent of high academic performance (GPA). Doing well in school, spending more time doing homework, and preparing for college are much more common among female than male high school students. Females are also more likely to feel that these actions are normative. The significance of the gender gap is not its size—it is fairly modest—but its ubiquity. Female students report an advantage on almost every cultural dimension. As with prior stages of the College Pathways Model, if all cultural and behavioral background variables (including GPA) are held constant with men, women are a bit less likely to be college prepared than men. This means that the female advantage in educational outcomes (including college preparedness) is “explained” by their higher grades, more positive attitudes toward school, and higher proportion of friends planning to go to college relative to their male peers. The race-ethnicity and immigrant-generation deficits in college preparation for a few groups (African Americans, American Indians, Mexican Americans, and first-generation immigrants) are largely a function of family socioeconomic background. In this regard, below-average college preparation levels in high school are more similar to levels for college aspirations than to those for college expectations. There appeared to be some signs of a persistent minority gap in college expectations (given aspirations), but this is not evident for college preparation or college aspirations. The most interesting finding about college preparation is the net Asian advantage, especially for East Asians, Vietnamese, and Cambodians. This

228   From High School to College

pattern is strongly associated with a supportive cultural context (higher encouragement and expectations and more friends planning to attend college). However, the extraordinarily high level of college preparation among East Asians defies our attempts to find any measured covariate as a mediating variable. Over 84 percent of the East Asian high school seniors who aspire and expect to complete college are college prepared— a full fifteen percentage points above the population mean (see table 4.1). Regardless of their individual and familial circumstances, almost all East Asian students are on the college-bound track. First-generation immigrants have a persistent deficit in being college prepared that does not appear to be related to any of the measured background or cultural variables. Perhaps newcomers, especially those who are still English-language learners, are “steered” to more vocational tracks than the college-preparatory track. Self-selection may also play a role here; recall that first-generation students also had somewhat lower college aspirations.

Enrollment in a Four-Year College As noted in prior chapters, the NSC records underestimate the college enrollment of East Asian and Vietnamese students. However, the advantages of using the NSC data matched records are important, including the inclusion of the complete universe of UW-BHS high school seniors (not just the 90 percent that were successfully followed up) and the bridge to college graduation is only available in NSC records. Following the precedent of Chapter 5, we estimate the transition from college preparedness to enrollment in a four-year college for both samples—the NSC matched students alone (figure 6.6 and appendix table 6.A5, available online) and the combined NSC and UW-BHS follow-up estimates of college enrollment (figure 6.7 and appendix table 6.A6, available online).33 With two major exceptions—the higher college enrollment rates of East Asian and Vietnamese students and the significant social origins effects—the overall patterns are similar in figures 6.6 and 6.7. We think the results in figure 6.7 (and appendix table 6.A6) provide a more accurate picture of the impact of Asian heritage and social origins on the transition to college enrollment. Regardless of which estimates are used, the gender gap in the transition from college prepared to college enrollment is a bit smaller than the gender gaps observed in high school (aspirations, expectations, and preparation). The gender gap in college enrollment is statistically significant, but just barely, and disappears entirely when the cultural context variables are added in model 3. Female students are more likely than their male peers to be encouraged and expected to attend college and have more friends planning to go to college. These attributes (which are correlated with GPA) explain the slight female edge.

2 3 5 Female

1.16

0.68

2 3 5 African American

0.63

2 3 5 American Indian

0.59

2 3 5 East Asian

0.640.60

2 3 5 2 3 5 Cambodian Vietnamese

2 3 5 Other Asian

2 3 5 Pacific Islander

2 3 5 Mexican

2 3 5 Other Hispanic

2 3 5 1st generation

0.600.590.57

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

0.520.51

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 6.A5. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 6.6    Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC Only) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors with Cultural Variables

Regression coefficient

2 3 5 Female

1.24

0.71

2 3 5 African American

0.62

2 3 5 American Indian

2 3 5 East Asian

1.45

2 3 5 2 3 5 Cambodian Vietnamese

0.50 0.49

2 3 5 Other Asian

2 3 5 Pacific Islander

0.52 0.51

2 3 5 Mexican

2 3 5 Other Hispanic

0.65

2 3 5 1st generation

0.70

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

0.60

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 6.A6. NSC = National Student Clearinghouse. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

Figure 6.7    Logistic Regression of the Conditional Probability of Enrollment in a Four-Year College (NSC or UW-BHS Follow-Up) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College-Prepared UW-BHS High School Seniors with Cultural Variables

Regression coefficient

Explaining Educational Transitions   231

For disadvantaged minorities, there is a distinct difference between college aspirations and preparation, on one hand, and the transition (from college preparedness) to college enrollment on the other. For aspirations and preparation, the gaps between whites and minorities were entirely a function of differential social origins. For the transition to college enrollment, the disadvantages are not completely explained by social origins or by any measured deficit in cultural context or expressions. These findings are similar in both samples (figures 6.6 and 6.7). The cultural variables have strong and significant effects on the transition to college enrollment. Students with more parental support, encouragement from significant others, friends planning to go to college, internal locus of control, and higher GPA are much more likely to enroll in a fouryear college (net of social origins). With the exception of high school GPA, the cultural variables do not explain the minority disadvantage in the transition to college. And GPA is only important for African American and Pacific Islander students. Lower social origins are part of the reason for the lower transition to college enrollment for disadvantaged minorities, but it does not tell the whole story. Perhaps the social-origins variables measured here do not completely capture some of the economic costs of college, including the willingness to borrow to pay for higher education.34 Other social factors not captured by the family background and cultural variables may also play a role in raising college-going rates for white and third-generation students relative to minority students in similar circumstances. Such factors may include receiving assistance from grandparents and other family members, benefiting from admissions or scholarships based on legacy programs, minor sports athletic financial aid, and social connections that encourage or assist students to get into college. The apparent East Asian deficit in the transition to college enrollment with the NSC sample (figure 6.6) is not evident in the combined NSC and UW-BHS follow-up measure of college enrollment (figure 6.7). The higher college-enrollment transition rates of East Asian and Vietnamese students relative to whites are, however, not statistically significant in figure 6.7 (except in one model). The college-transition rates for other Asian groups—Cambodians, Filipinos, and other Asians—more closely reflect the patterns for disadvantaged minorities than those for East Asian students. These patterns suggest that there may be some general barriers to college that affect all minorities, not just non-Asian minorities. The overall pattern of lower transition to college enrollment is also evident for first-generation immigrants. Appendix table 6.A6 also shows that cultural resources, net of social origins and ascription, are important determinants of the transition to college enrollment. The summary statistic for variance explained jumps from 3.8 percent to 7.4 percent from model 2 to model 3, when the

232   From High School to College

cultural-context variables are added, and by five additional percentage points (from 7.6 to 12.6) from model 4 to model 5 when GPA and homework hours are included. High school students who are highly encouraged, have friends going to college, and receive good grades are more likely to enroll in college, independent of their social origins and other background factors.

From College Enrollment to Completion With few major exceptions, the transition from college entry to completion (figure 6.8) appears very similar to the patterns for the transition from high school to four-year-college enrollment. Females maintain a modest advantage rooted in cultural context and cultural orientations. The presence of college-going friends in high school and a strong internal locus of control appear to be particularly important (see appendix table 6.A7 online).35 Minorities, especially African Americans, Native Americans, Filipinos, Pacific Islanders, and Other Hispanics, seemed to be particularly disadvantaged in completing college. There are, however, two striking exceptions. Vietnamese students are very likely to complete college. Overall, the Vietnamese students in the UW-BHS sample disproportionately enroll in community college, primarily because it is a low-cost alternative to a four-year college. Moreover, the Vietnamese students who get a toehold in a four-year college are more likely to finish than any other group. The exceptional commitment of Vietnamese students to higher education is evident in almost every indicator. The other interesting finding in figure 6.8 is the lack of any disadvantage for first-generation-immigrant students in completing college. There is no real advantage for first-generation students—relative to third-andhigher-generation students—though sometimes the difference is close to statistical significance. This is in sharp contrast to figures 6.6 and 6.7, which showed that first-generation students had a distinct disadvantage in the transition from prepared high school student to enrollment in a four-year college. One possible interpretation is selectivity: fewer firstgeneration students are able to enroll in a four-year college, but those who do are likely highly motivated to be successful. With the exception of the Vietnamese and first-generation-immigrant coefficients, the patterns in figures 6.7 and 6.8 are similar. The consistent pattern is a general minority (relative to white) disadvantage in college entry and college completion. One possible explanation is that the realities of getting into and through college are quite different from the realities of aspiring and preparing for college in high school. Students receive a lot of family and peer support for college planning in high school. Recall that most minorities (especially Asians) had higher levels of family encouragement and college expectations than white students. However,

1.24

2 3 5 Female

1.32

2 3 5 African American

0.600.63

2 3 5 American Indian

2 3 5 East Asian

0.68

3.01

2.48

2 3 5 2 3 5 Cambodian Vietnamese

2.55

2 3 5 Other Asian

2 3 5 Pacific Islander

0.280.290.32

2 3 5 Mexican

2 3 5 Other Hispanic

0.500.52

2 3 5 1st generation

2 3 5 2nd generation

Interpreting coefficients in models Solid black fill = statistically significant above reference population, set at 1.0 Medium gray fill = statistically significant below reference population, set at 1.0 No fill = not statistically significant

2 3 5 Filipino

0.540.48

Model 2: Ascription and social origins Model 3: Model 2 plus cultural context Model 5: Model 3 plus cultural orientations and expressions

Source: The University of Washington-Beyond High School project (Hirschman and Almgren 2016). Notes: This figure is based on appendix table 6.A7. UW-BHS = University of Washington-Beyond High School.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

Figure 6.8    Logistic Regression of the Conditional Probability of College Graduation (in Seven Years) on Gender, Race-Ethnicity, and Immigrant Generation Relative to Males, Whites, and the Third-and-Higher Generation of College Entrant UW-BHS High School Seniors with Cultural Variables

Regression coefficient

234   From High School to College

they face more serious barriers to getting into and through college—both financial and motivational. The leading theory of college completion emphasizes “engagement” with the institution and with other students.36 Minority and first-generation students may become more isolated than white students, finding it harder to establish close friendships and attachments that help to pull them through difficulties along the way.

Conclusions The primary aim of this chapter was to test whether cultural variables (context, orientations, and expressions) played an important role in (re)producing gender, race-ethnicity, and immigrant-generation differences at each stage of the College Pathways Model from college aspirations to college graduation. Our analysis finds an important, but circumscribed, role for cultural variables in creating and maintaining gender, race-ethnicity, and immigrant-generation inequality in educational outcomes from college aspirations to college graduation. While culture appears to matter, its role is much smaller than many proponents generally claim. Students reared in families with a favorable cultural context of support, parental engagement, and encouragement are more likely to report positive cultural orientations—proschool attitudes and internal locus of control (selfefficacy)—and to invest more time in homework and obtain a higher GPA. The correlations among these variables, however, are modest—generally in the range from .2 to .3. Academic and lay arguments on the role of culture in maintaining educational stratification tend to be “either-or”—they are either “very important” or nonexistent. Our finding is that cultural factors work in the expected direction, but they should not be considered a tightly wound system of cultural inputs and behavioral outputs. All of the cultural variables are positively associated with attitudes and preparation for college in high school but are less strongly linked to actually enrolling in and completing college. Furthermore, the relationships were not always linear. For some variables, such as encouragement and proschool attitudes, the impact on college outcomes was strong (negatively) at the bottom of the distribution, but there were only slight differences in the outcomes in the distribution’s middle and higher ranges. The most consequential cultural traits on college entrance and completion were having many high school friends planning to attend college and a high GPA. In the case of gender, cultural variables, particularly those labeled as cultural context, completely “explain” the modest female advantages at all stages of the College Pathways Model. Women are reared in the same households as men; therefore, differences in the family SES and family structure cannot explain why women are more educationally ambitious than men, why female high school students are more prepared for col-

Explaining Educational Transitions   235

lege, or why women are more likely to enroll in college right after high school, and why they are more likely to complete college. The female edge appears to be because women report receiving higher levels of family encouragement and expectations and having more high school friends who are planning to go to college than their male peers. The female advantage is also linked to higher proschool attitudes (“doing well in school is part of who I am”) and a higher internal locus of control. Conceptually, we have modeled GPA as dependent on cultural context and cultural orientations. This is a logical premise, which assumes that behavior—high school GPA—is determined, at least in part, on prior beliefs and experiences. However, there is a problem with this assumption because cultural context and GPA are measured simultaneously in the UW-BHS senior survey. An alternative to our assumption that academic performance in the past (before the senior year) is the primary cause of all the cultural variables measured here. In other words, the superior school performance of women in the past means they are more likely to be encouraged to go to college and to have a greater sense of confidence that their actions will be effective. The cumulative history of female advantage in academic performance has probably created a “cultural environment” that is more favorable for continued female success in college. Asian students display distinct advantage in the first three stages of the College Pathways Model but not in the transitions to college enrollment (given preparedness) and college completion (given college enrollment). The high level of college aspirations, expectations, and preparation of Asian students is not due to their favorable socioeconomic origins, higher levels of family support and engagement, or locus of control. In fact, on each of these criteria, most Asian students have below average resources (East Asians are the exception). However, Asian students do have a “cultural” advantage that explains part of their edge in the high schoolmeasured attributes—namely, the high level of encouragement and family expectations of college attendance. Asian students also report the highest proportion of high school friends planning to attend college. The family and peer environment creates “pressures” to conform to the norms of aspiring and preparing for college. These pressures are the strongest when Asian adolescents are in high school—when they are living with their parents and enmeshed in peer networks. Once students leave high school, they confront a new reality. There is little evidence of the Asian “model student” stereotype in the net transition from being college prepared to college enrollment or from college enrollment to completion. The only exceptions to this pattern are East Asians and Vietnamese. Overall, Asians are still doing pretty well—the percent of Asian students who graduate from college is only a few percentage points below that of white students in our NSC-matched sample

236   From High School to College

(see chapter 4), while census data show that Asian students are much more likely to enroll and graduate from college than white students (see chapter 2). The foundation of the Asian advantage lies in the characteristics measured in high school—Asian high school students are much more likely to complete high school and are much more likely to exhibit high college aspirations, expectations, and preparation, regardless of their social origins. Part of the Asian American advantage is rooted in their cultural context—Asian students report much higher levels of encouragement and expectations to attend college.37 They are also much more likely to have friends who are planning to attend college. These advantages do not entirely disappear in the transitions to college enrollment and college completion, but they are not as influential as they were for the high school–based outcomes. Furthermore, some Asian groups—Cambodians and Filipinos—appear to encounter lower transition rates that are similar to disadvantaged minorities. If culture is a central element of the gender story and plays a significant (albeit limited) role for Asian students, what impact does culture have in explaining the lower levels of educational attainment of disadvantaged minorities, including African Americans, American Indians, Pacific Islanders, Mexicans, and Other Hispanics? Recall the “culture of poverty” interpretation was “invented” to address the plight of Puerto Ricans and African Americans in inner cities in the 1960s. Lewis and Moynihan did not argue that minorities were poor solely because of their cultural orientations, but because ambition, effort, and persistence were mired in social conditions of long-term discrimination and joblessness. The lack of incentives was hypothesized to have a negative impact on attitudes and effort, which led to a weak commitment to families and childrearing. According to this interpretation, the combination of structural conditions and a culture lacking in ambition “trapped” minorities in an intergenerational cycle of poverty. The results in this chapter do not provide any support for the culture of poverty hypothesis for disadvantaged race and ethnic minorities. Cultural context, orientations, and expressions do matter—especially as individual attitudes—they have strong direct effects on academic outcomes. However, net of social origins, these cultural variables do not explain (mediate) the lower educational outcomes of disadvantaged minorities. Simple correlations (table 6.2) show that disadvantaged minorities do not have appreciably less cultural resources than other groups. Recall that white students had above-average levels of family support and engagement and a higher internal locus of control. Asian students scored low on these cultural variables but had higher values on other dimensions (encouragement and high expectations, higher proschool attitudes). White and Asian students also shared some common cultural resources

Explaining Educational Transitions   237

(friends planning to attend college and above-average GPA). African Americans and the other disadvantaged groups tended to have intermediate values—closer to the population averages. The only consistent minority disadvantages were lower proportions of friends planning to attend college and slightly below-average GPA. The non-Asian minorities had more supportive and engaged parents than Asian students and experienced more encouragement than white students. There are no “red flags” of major cultural deficits for the non-Asian minorities—and their proschool attitudes were well above average. Our “test” of the cultural hypothesis is centered on the comparison of the minority coefficients in models 3, 4, and 5 (when the cultural variables were added), relative to model 2 (net of SES). For college aspirations, ambitions, and preparation, the non-Asian minority coefficients were generally not significant net of social origins. The addition of cultural variables showed little change in ascriptive odds ratios. If anything, the cultural context of disadvantaged minorities was a positive resource in the outcomes measured in high school (aspirations, expectations, and preparation). For college enrollment and completion, non-Asian minority students experience net disadvantages that are unrelated to any of the cultural variables measured here. Our tentative conclusion is that the problems experienced by nonAsian minorities in high school—lower college aspirations, lower expectations, and lack of college preparation—are primarily due to lower family socioeconomic resources, indexed here by parental education and occupation, home ownership, and family structure. One of the mechanisms by which family SES affects educational outcomes is through intergenerational socialization, including parental communication, encouragement, and other support. If minorities receive less intergenerational social support, it is because of the socioeconomic composition of their families, not because of any distinctive cultural deficit of socialization that is associated with ethnic origins. Non-Asian and some Asian minorities face major obstacles in college entry and college completion relative to white students, net of social origins. This also seemed to be true for college expectations. However, the full battery of cultural variables—context, orientations, and expressions— does not really “explain” (mediate) these persistent disadvantages. We suspect that some of the unmeasured financial costs of enrollment in a four-year college and the lack of social support (peer or institutional) for college students may be more significant than the cultural variables measured in this chapter. The direct effects of cultural background on the college-pathway transitions are consistent with this standard model—cultural resources do have direct effects on educational outcomes. These cultural variables, however, do not play a simple role in explaining why different groups

238   From High School to College

are more successful than others. The female edge is explained by their educational performance and a supportive cultural milieu. Part of the Asian edge, especially in high school, is due to higher levels of cultural support and encouragement. But the underachievement of non-Asian minorities (and some Asian minorities) is due to socioeconomic origins, not cultural deficits. An important theme in the child development literature stresses the importance of authoritative childrearing practices in contrast the passive or laissez-faire parenting and authoritarian parenting. Authoritative parenting provides a model of behavior for children to emulate and relies on frequent discussion, interaction, and boundary-setting to guide and shape the child’s goals and behaviors. The aim of authoritative parenting is to create independent and self-regulating children through reason and understanding. According to this model, positive parents are nurturing, interested, engaged, and encouraging in interactions with their children. If successful, authoritative parenting will lead to confident adolescents with positive self-images, a willingness to defer gratification in order to reach goals, and high educational ambitions.38 If the cultural-context variables can be considered as indicators of authoritative parenting, the results of this chapter provide support for this popular theory of successful child rearing. In particular, the parental communication and support index and the encouragement index are highly predictive of positive outcomes on most (but not all) pathways to college graduation. But the impact of many other cultural variables, including friends’ college plans, proschool attitudes, and an internal locus of control, have direct positive effects on pathways to college graduation.

Chapter 7 Work and Extracurricular Activities in the Lives of High School Seniors

I

n his classic study, The Adolescent Society: The Social Life of the Teenager

and Its Impact on Education, James Coleman described the lives of American high school students in the 1950s as a quasi-independent “adolescent world” in which academic pursuits and preparation for adult roles were secondary concerns.1 Although students lived with their families and attended schools run by adults, the priorities of students were centered on their social ties to other adolescents. Students lived in their own “bubble”—a very competitive world of peer evaluations based on athletic prowess, participation in student activities and clubs, personal attractiveness and dress, personality, and popularity. In addition to athletics and extracurricular activities, high school students spent most of their leisure time listening to music, attending movies, working on cars, or just being with friends. Most students spent more time watching television than doing homework. Being in the right group was very important, and academic success was valued much less than athletics and popularity. There were significant variations in the status system across the ten high schools in Coleman’s sample, which ranged from small high schools in rural communities to larger schools in suburbs and mediumsized cities. However, the ostensible objectives of schooling—preparing students for careers and higher education—seemed to be a secondary priority in almost every context. One theme of Coleman’s study, although not explicitly expressed, was that the considerable time spent on non­ academic pursuits by high school students had a detrimental effect on their schooling and preparation for adult careers. In this chapter, we address whether and how two dimensions of the lives of high school seniors—paid employment and participation in school activities—affect college aspirations, expectations, preparedness, 239

240   From High School to College

enrollment, and completion (the stages of the College Pathways Model). This question is important because many people, though they may not be familiar with Coleman’s book, believe that all the time spent on nonacademic pursuits interferes with the primary objectives of schooling. Although the normal school day (including homework assignments after school) is generally considered equivalent to a full-time job, many students are also part-time workers in the labor force. A generation ago, over one-third of all students aged sixteen to seventeen held jobs. This figure dropped dramatically during the recessions of the 2000s, and only one in seven teenage high school students was working in 2009.2 In our sample of UW-BHS students, almost half of high school seniors were working in a paying job. The higher level of employment in the UW-BHS sample, relative to national data, was most likely due to age composition. High school seniors in our sample were generally seventeen and eighteen years old, and teen employment is higher at older ages. There was a modest decline in paid employment of UW-BHS high school seniors from 2000 to 2005, which is consistent with the national trend. Other than the level of employment, none of the findings reported in this chapter varied significantly from year to year. Most of the UW-BHS student workers, much like their counterparts around the country, were concentrated in the secondary labor market of part-time jobs in fast-food enterprises, retail sales, child care, and similar pursuits.3 Most of these jobs are in the private sector and are not coordinated with the formal schooling—unlike the European model of apprenticeship training that is considered part of educational experience of vocational-orientated high schools.4 In addition to classes and employment, many high school students participate in a variety of school programs including athletics, clubs, music, theater, journalism (school newspapers and yearbooks), and social service organizations. For some students, the commitment of time and energy to extracurricular activities exceeds that of their academic pursuits. This is particularly true of student athletes in major interscholastic sports (football and basketball) and others involved in artistic and musical performances. For the average student, participation in after-school organizations and clubs represents a relatively modest commitment of less than ten hours per week—less than the time spent in many other social and leisure-time activities. About 38 percent of the UW-BHS high school seniors reported zero hours spent in any extracurricular activities, a figure that is in line with other studies of American high schools.5 Although Coleman’s study might be read as a critique of adolescent society that is overly focused on popularity, cars, and athletics to the detriment of academic pursuits, a reasonable case can be made for the American high school’s being organized to socialize and prepare youth for adult roles beyond vocational and scholastic objectives. Employment

Work and Extracurricular Activities   241

and participation in sports and school activities can develop a sense of responsibility as well as opportunities to learn leadership skills, personal initiative, teamwork, and sportsmanship.6 The ability to communicate and work effectively with others is an important skill in many professions and for civic participation in a democratic society. Another increasingly important skill, for both vocational and civic pursuits, is the ability to understand people who come from different social, economic, and cultural backgrounds. Immigration, along with geographic and social mobility, has increased exposure to diversity (ethnic, socioeconomic, and other dimensions) in most American companies, universities, organizations, cities, and neighborhoods. Informal interactions and friendships forged through employment, athletics, and extracurricular activities may be as important as classroom lessons in teaching students how to understand and work effectively with students from different backgrounds. There is, of course, a potential downside to student employment and extracurricular activities, namely that they compete for the limited attention and time of high school students. The length of the school day in high school varies from state to state, but it is generally between six and seven hours. Although this is a bit less than the standard adult workday, homework (outside of school hours) and participation in school activities can add up to a full-time job for most students. But for some students—those who spend evenings or weekends working a part-time job or participating in a time-consuming sport or extracurricular activity—it is a different story. These additional obligations may come at the expense of doing homework, keeping up with family responsibilities and obligations, participating in other leisure-time pursuits, and getting enough sleep. Some observers have argued that the roles of worker and student are incompatible, or at least incompatible with educational success.7 The underlying assumptions are that teenage employment reduces time that should be spent studying and teenage jobs inevitably erode academic commitment. Most empirical studies have found, however, few negative effects of employment on academic performance among students who work a moderate number of hours per week. Indeed, students who work fewer than fifteen hours per week generally have better educational outcomes than students who do not work at all.8 Students who work longer hours, especially more than twenty or twenty-five hours per week, have lower grades and are more likely to drop out of school.9 However, it is unclear whether a time-intensive job is a cause of poorer educational outcomes or whether both are caused by a lack of interest in school. A parallel argument, known as the overscheduling hypothesis, has been presented on the impact of participation in extracurricular activities.10 The prototypical image of this perspective is the suburban mother shuffling multiple children between music and ballet classes, soccer practice, and Girl and Boy Scout meetings. Although these practices begin

242   From High School to College

with younger children, the overscheduling hypothesis posits that adolescents continue on their own initiative to overparticipate to fulfill the expectations of their parents and teachers with the belief that a high school resume filled with activities will impress college admissions officers and prospective employers. Although the overscheduling of children and adolescents may be a burden for harried parents, Annette Lareau argues that many of the verbal and interpersonal skills developed through “concerted cultivation” during childhood prepare middle-class children for participation in complex organizations when they leave home.11 Empirical research has failed to support the claims of the overscheduling hypothesis. While, these studies vary considerably in the measurement of activities, the ages of children or adolescents, and data sources and periods, they have common findings. Children and youth who participate in organized activities, including athletics and clubs, have more positive developmental outcomes as measured by school grades, attitudes, and mental-health status.12 There are some modest qualifications for students at the highest end of the participation scale and those with very demanding parents, but the overall findings suggest that participation in organized activities generally leads to positive outcomes for children and youth. The finding that student employment and participation in extracurricular activities may have a positive impact on educational outcomes is compatible with the hypothesis that student engagement is a critical contributor to academic success and college graduation in particular.13 The literature suggests that engagement—namely, close contact with teachers, involvement with peers, and identification with their educational institution— helps motivate students and enables them to persist through obstacles and discouragement. In her doctoral dissertation based on UW-BHS data, Irina Voloshin found that moderate hours of student employment had a positive impact on college ambitions and enrollment.14 Building on the student engagement literature, the prior chapters in this volume, and Voloshin’s research, we propose the model in figure 7.1 to estimate the impact of student employment and school participation on the College Pathways Model. The questions we seek to answer are: (1) Do student employment and school participation have positive or negative effects on the pathways from college aspirations to college graduation? (2) Do student employment and school participation mediate any of the observed effects of gender, race-ethnicity, and immigrant generation on educational outcomes? It is particularly important to test whether the net of effects of student employment and school participation on educational outcomes are due simply to the selection of high-SES students into employment and participation. We also seek to test if the net effects of student employment and school participation on educational outcomes are a result of student engagement, as measured by cultural characteristics.

Work and Extracurricular Activities   243 Figure 7.1    A Revised College Pathways Model of Ascription, Social Origins, Employment and Participation in Extracurricular Activities, Cultural Variables, and College-Pathway Transitions College-pathway transitions Ascription Gender Race-ethnicity Immigrant generation Cultural context Cultural orientations Cultural expressions

Social origins

Employment and participation in extracurricular activities

Source: Author’s compilation.

Measuring Student Employment and Participation in Extracurricular Activities The UW-BHS questionnaires included several items that measured multiple dimensions of student employment and participation in extracurricular activities. Here, we focus on one dimension of work and extracurricular activities—namely, the average number of hours spent working or participating per week. The UW-BHS senior survey asked students if they were currently working, had previously worked, or had never worked. Work was defined as paid employment, thus excluding household chores and volunteer activities. For example, babysitting at home is excluded from this definition of employment, but paid babysitting (generally outside the home) counts as work. The measure of hours worked per week was in response to a follow-up question for students who were currently working (at the time of interview): “How many hours do or did you usually work per week on this job during the school year?” Seniors were also asked about how much time they spent on extra­ curricular activities (“How many hours do or did you spend on each activity during a typical week when the activity is or was going on?”). This question was a follow-up to questions that asked students to write in up to three sports activities and up to three nonsports activities. The numbers

244   From High School to College

of hours for each activity were summed to yield total weekly hours spent in sports and other (nonsports) extracurricular activities. Some activities, especially sports, are seasonal and may not extend for the duration of the academic year. The wording of the question emphasized average time spent while the activities were occurring. The focus on time-intensity in this chapter does not mean that other dimensions of work and student participation are insignificant.15 However, as noted previously, hours of work and participation are generally the primary focus in the research literature. The questions to be addressed are whether work and student participation are positively or negatively associated with educational pathways and if the relationships are sensitive to the time spent on the activity.

The Prevalence of Paid Employment Before examining the consequences of student activities, we present a description of the prevalence and time-intensity of student employment and participation by the ascriptive dimensions of gender, race-ethnicity, and immigrant generation. Table 7.1 shows that slightly more than half of high school seniors in the UW-BHS sample were working in the spring of their senior year. The number of students with past or present work experience—those who have “ever worked,” including during summers or at a prior time during the current year or in previous years—is even higher. About 20 percent of all UW-BHS high school seniors worked a modest number of hours—fewer than fifteen hours per week. Another 20 percent worked fifteen to twenty-five hours, and 10 percent worked more than twenty-five hours per week. For those working more than twenty-five hours per week, the time spent working was sometimes as much as that spent in school. Even students in the middle category of fifteen to twentyfive hours per week were probably working several hours every day after school or full time on the weekends. In looking at the data, we see significant differences in the prevalence and time-intensity (hours per work week) in paid employment by gender, race-ethnicity, and immigrant generation, but the differences are more a matter of degree than a defining attribute of different groups. For example, the range of group differences in nonemployment is only around fifteen percentage points (from 45 percent to less than 60 percent). The overall patterns suggest that advantaged groups are more likely to be working, especially in the “good” (low-time-intensity) jobs of less than fifteen hours per week. By most indicators, females are moderately more successful in high school than males. They receive higher grades, have higher ambitions, and are more likely to enroll in (and graduate from) college after completing high school. Female seniors are also slightly more likely to be working than

21.7 14.1

12.2

18.2 15.4 27.0 13.1 16.3

54.7

58.1 50.4 48.6 58.8 53.6

21.0 17.1

19.3