Analysis of Baseline Program Data from the Ph.D. Completion Project

PH.D. COMPLETION AND ATTRITION: Analysis of Baseline Program Data from the Ph.D. Completion Project

PH.D. COMPLETION AND ATTRITION Analysis of Baseline Program Data from the Ph.D. Completion Project

Council of Graduate Schools One Dupont Circle, NW Suite 230 Washington, DC 20036-1173 Phone (202) 223-3791 Fax (202)331-7157 www.cgsnet.org

Ph.D. Completion and Attrition: Analysis of Baseline Program Data from the Ph.D. Completion Project

The Council of Graduate Schools

i PH.D. COMPLETION AND ATTRITION: Analysis of Baseline Program Data from the Ph.D. Completion Project

Robert Sowell Vice President, Programs and Operations

Ting Zhang Research Analyst

Kenneth Redd Director, Research and Policy Analysis

Guest Editor: Margaret F. King

ALL RIGHTS RESERVED. No part of this work covered by the copyright herein may be reproduced or used in any form by any means—graphic, electronic, or mechanical, including photocopying, recording, taping, Web distribution, or information storage and retrieval systems—without the prior written permission of the Council of Graduate Schools.

ISBN: 1-933042-13-3

Printed in the United States TABLE OF CONTENTS

List of Figures and Tables...... i Foreword...... viii Acknowledgments...... xi Chapter 1 Introduction...... 1 Chapter 2 The CGS Completion Project: Context and Overview...... 3 The National Context, Including Prior Studies...... 3 Project Overview...... 7 Chapter 3 Collection of Baseline Completion Data at the Program Level by Discipline and Broad Field...... 9 Templates Used for Data Collection...... 11 Ph.D. Completion Data Template...... 11 Ph.D. Attrition Data Template...... 12 Chapter 4 Completion and Attrition Trends: Analysis of Baseline Program Data...... 14 Ten-Year Completion Rates...... 14 Completion Rates by Broad Field...... 16 Completion Rates by Disciplines...... 18 Completion Rates by Institution Type...... 29 Completion Rates by Cohort Size...... 31 Seven-Year Completion Rates...... 35 Attrition Rates...... 37 Attrition Rates by SEM vs. SSH Fields...... 38 Attrition Rates by Broad Fields...... 39 Attrition Rates by Disciplines...... 40 Attrition Rates by Institution Types...... 46 Attrition Rates by Cohort Size...... 48 Early Attrition...... 50 Components of Early Attrition...... 52 Continuing Completion, and Attrition Rates...... 58 Conclusions...... 63 Appendix A Ph.D. Completion Project – Advisory Board Members...... 66 Appendix B Ph.D. Completion Project - Phase I Institutions...... 67 Appendix C Number of Entering Students and Doctoral Programs by Broad Field and Discipline...... 69 Appendix D Data Templates...... 70 Appendix E Investigation of Cohort Size Effects...... 72 References...... 75

iii List of Figures and Tables Table 1 Summary of Prior Major Studies of Ph.D. Completion and Attrition Patterns...... 4

Figure 4-1 Average Cumulative Overall Ten-Year Completion Rates for Students Entering Doctoral Programs from 1992-93 through 1994-95, by Year...... 15

Figure 4-2 Average Cumulative Ten-Year Completion Rates for SEM and SSH Cohorts Beginning Doctoral Study from 1992-93 through 1994-95, by Year...... 16

Figure 4-3 Average Cumulative Ten-Year Completion Rates for Cohorts Entering Doctoral Study from 1992-93 through 1994-95, by Broad Field and Year...... 17

Figure 4-4 Average Ten-Year Cumulative Completion Rates for Students Entering Five Engineering _ Doctoral Disciplines from 1992-93 through 1994-95, by Year...... 20

Figure 4-5 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts Beginning from 1992-93 through 1994-95, for Engineering Disciplines with Five or More Programs, at Year Ten...... 21

Figure 4-6 Average Ten-Year Cumulative Completion Rates for Students Entering Five Life Science _ Doctoral Disciplines from 1992-93 through 1994-95, by Year...... 22

Figure 4-7 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts from 1992-93, through 1994-95, for Life Science Disciplines with Five or More Programs, at Year Ten...... 23

Figure 4-8 Average Ten-Year Cumulative Completion Rates for Students Entering Mathematics & Physical Science Doctoral Disciplines from 1992-93 through 1994-95, by Year...... 24

Figure 4-9 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts from 1992-93 through 1994-95, for Mathematics & Physical Science Disciplines with Five or More Programs, at Year Ten...... 25

Figure 4-10 Average Ten-Year Cumulative Completion Rates for Students Entering Six Social Science Doctoral Disciplines from 1992-93 through 1994-95, by Year...... 26

Figure 4-11 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts from 1992-93 through 1994-95, for Social Science Disciplines with Five or More Programs, at Year Ten.....27

Figure 4-12 Average Ten-Year Cumulative Completion Rates for Students Entering Four Humanities _ Doctoral Disciplines from 1992-93 through 1994-95, by Year...... 28

Figure 4-13 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts from 1992-93 through 1994-95, for Humanities Fields with Five or More Programs, at Year Ten...... 29

Figure 4-14 Average Cumulative Ten-Year Completion Rates by Institution Type for Students Entering Doctoral Programs from 1992-93 through 1994-95, by Year...... 30

Figure 4-15 Average Ten-Year Cumulative Completion Rates by Institution Type for Doctoral Students Entering from 1992-93 through 1994-95, by Broad Fields and Year...... 31

Figure 4-16 Average Cumulative Ten-Year Completion Rates by Cohort Size for Doctoral Students ._ Entering from 1992-93 through 1994-95, by Year...... 32

iv Figure 4-17 Average Cumulative Ten-Year Completion Rates by Cohort Size for Doctoral Students Entering from 1992-93 through 1994-95, by Broad Field and Year...... 33

Figure 4-18 Average Cumulative Ten-Year Completion Rates for Chemistry by Cohort Size for Doctoral Students Entering from 1992-93 through 1994-95 and Year...... 34

Figure 4-19 Average Cumulative Ten-Year Completion Rates for Computer & Information Sciences _ by Cohort Size for Doctoral Students Entering from 1992-93 through 1994-95 and Year...... 35

Figure 4-20 Comparison of Seven-Year Cumulative Completion Rates between A- and B- Cohorts at Year Seven, by SEM vs. SSH Fields...... 36

Figure 4-21 Seven-Year Cumulative Doctoral Student Completion Rates for A- and B- Cohorts at ..._ Year Seven, by Broad Field...... 37

Figure 4-22 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts Who Began from 1992-93 through 1994-95, by Year...... 38

Figure 4-23 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts Entering from 1992-93 through 1994-95, for SEM vs. SSH Fields, by Year...... 39

Figure 4-24 Cumulative Ten-year Attrition Rates for Doctoral Student Cohorts Entering from 1992-93 through 1994-95, by Broad Field and Year...... 40

Figure 4-25 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts That Entered Engineering Disciplines from 1992-93 through 1994-95 at Five or More Programs, by Year.. 41

Figure 4-26 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts That Entered Life Science Disciplines from 1992-93 through 1994-95 at Five or More Programs, by Year...... 42

Figure 4-27 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts That Entered Mathematics & Physical Sciences Disciplines from 1992-93 through 1994-95 at Five or More Programs, by Year...... 43

Figure 4-28 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts That Entered Social Sciences Disciplines from 1992-93 through 1994-95 at Five or More Programs, by Year...... 44

Figure 4-29 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts That Entered Humanities Disciplines from 1992-93 through 1994-95, at Five or More Programs, by Year...... 45

Figure 4-30 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts from 1992-93 through 1994-95, by Institution Type and Year...... 46

Figure 4-31 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts from 1992-93 through 1994-95, by Institution Type and Broad Field and Year...... 47

Figure 4-32 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts from 1992-93 through 1994-95, by Cohort Size and Year...... 48

Figure 4-33 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts from 1992-93 through 1994-95, by Cohort Size, Broad Field, and Year...... 49

Figure 4-34 Overall Four-Year Cumulative Attrition Rates for A-, B- and C-Cohorts...... 50

v Figure 4-35 Four-Year Cumulative Attrition Rates for A-, B- and C-Cohorts by SEM/SSH Fields...... 51

Figure 4-36 Four-Year Cumulative Attrition Rates for A-, B- and C-Cohorts, by Broad Field...... 52

Figure 4-37 Components of Overall Early Student Attrition for Years One through Four...... 53

Figure 4-38 Components of Early Student Attrition in Engineering for Years One through Four....54

Figure 4-39 Components of Early Attrition in Life Sciences for Years One through Four...... 55

Figure 4-40 Components of Early Attrition in Mathematics & Physical Sciences for Years One through Four...... 56

Figure 4-41 Components of Early Attrition in Social Sciences for Years One through Four...... 57

Figure 4-42 Components of Early Attrition in Humanities for Years One through Four...... 58

Figure 4-43 Cumulative Continuing, Completion, and Attrition Rates for Doctoral Students Entering _ from 1992-93 through 1994-95, by Broad Field and Year...... 59

Figure 4-44 Overall Four-Year Cumulative Continuing, Completion, and Attrition Rates, for A-, B- and C-Cohorts...... 61

Figure 4-45 Four-Year Cumulative Continuing, Completion, and Attrition Rates, for A-, B- and C- Cohorts, by Broad Field...... 61

Figure 4-46 Overall Four-Year Cumulative Continuing, Completion, and Attrition Rates by Cohort ._ Year...... 62

Table E.1. Student Counts by Cohort Size in Engineering for Cohorts Entering from 1992-93 through 1994-95...... 72

Figure E.1 Cumulative Ten-Year Attrition Rates for A-Cohorts in Engineering (except Electrical & Electronics Engineering), by Cohort Size...... 73

Figure E.2 Cumulative Ten-Year Attrition Rates for A-Cohorts in Engineering (except Electrical & Electronics Engineering), by Discipline-specific Cohort Size...... 74

vi Foreword By Debra W. Stewart, CGS President

“Knowing what to measure and how to measure it makes a complicated world much less so. If you learn how to look at data in the right way, you can explain riddles that otherwise might have seemed impossible. Because there is nothing like the sheer power of numbers to scrub away layers of confusion and contradiction [sic].” (Levitt and Dubner, 2005, p. 14)

octoral education in the United States is the model for much of the rest of the world. Its strengths are reflected both in the excellent preparation Dof graduates and the capacity to produce research and researchers who are primary contributors to economic growth and civic engagement. However, these results come at a price. Doctoral education is very expensive, both in financial terms and in terms of the time and effort invested by doctoral students, faculty, and others who sustain the enterprise.

It was the stature and the costliness of the doctoral enterprise that motivated the graduate community in the mid-1990s to launch a self-examination directed at identifying areas of weakness and at generating strategies for addressing them. The result has been a proliferation of studies and reports on doctoral education in the United States. These reports focused on different disciplines, different sub-sets of graduate students, different time frames, and the efficacy of different interventions (Ph.D. Completion and Attrition: Policy, Numbers, Leadership, and Next Steps. Washington, DC: Council of Graduate Schools, 2004). By 2003, it was clear that all of this work provided an enormously rich stew for creative speculation about how doctoral education might be further strengthened, a context that was probably essential for motivating the “jewel in the crown of higher education” to believe that even it could benefit from polishing. It was also clear that the time had come for CGS to launch a national initiative that would result in firming up a foundation for specific best-practice recommendations to U.S. graduate schools, programs, funders, and policy makers.

In order to reach this point, two things had to happen. First, we needed to identify a common empirical measure for assessing positive change. And second, in selecting that mode of measurement, we needed to think about the critical leverage points that could help unpack the mélange of issues that

vii had emerged in the “rich stew” of discussion and scholarship cited above. We settled on student completion and attrition rates from Ph.D. programs as the key point of leverage to ultimately generate best-practice recommendations to improve the effectiveness of America’s Ph.D. programs.

Completion was the key because we believed that of all of the issues raised in nearly a decade of our self-criticism of doctoral education, the most urgent was that too few students admitted into U.S. doctoral programs actually graduated. We also believed, as Levitt and Dubner note in the quotation introducing this Foreword, that “there is nothing like the sheer power of numbers to scrub away layers of confusion and contradiction.” We hoped that if we could launch a project that would empower all stakeholders, especially the deans of U.S. graduate schools, to lead conversations with faculty and students about what the completion and attrition rates actually were, and about what kinds of interventions might most successfully be implemented to improve completion, that alone would move the conversation forward. But if we could actually study a carefully selected set of interventions, specifically designed to address attrition in clearly defined disciplinary, programmatic, and university settings, we could ultimately generate the information upon which solid best-practice recommendations could be provided by CGS to our membership community. The Ph.D. Completion Project is aiming to achieve that objective.

This book, Ph.D. Completion and Attrition: Baseline Program Data from the Ph.D. Completion Project, is the first in a series of monographs that will be forthcoming from the project. Most CGS publications advance best practices in defined fields or, at the very least, describe the current state of discussion regarding best practices in emerging fields. As the first book in our major national demonstration project on Ph.D. completion and attrition, this book is both similar to and different from the typical CGS publication. Most readers familiar with CGS publications will be struck by the fact that this book trades in the currency of numbers rather than the more typical broad policy-statement or curriculum-oriented best-practice document. The book gives more emphasis to presenting data than to interpreting its meaning, though of course we do some of both. It invites more attention to the granularity of charts, tables, and numbers rather than the 10,000-foot-view prose that is our normal trade. But we begin with this level of granularity precisely because we believe that now is the time for action to increase Ph.D. completion, and that this action needs to be based on both a solid empirical understanding of the current situation and a transparent approach to how completion and attrition are calculated. While there is a best-practice element to this monograph, it lies in the elaboration of a methodology for assessing Ph.D. completion and attrition in the fine grain viii essential to moving the discussion forward. The CGS completion team headed by Robert Sowell struggled over how far we should go on interpretation of the baseline data and generally came down on the side of less rather than more.

One of the risks of such a hard-hitting data presentation that characterizes trends across student cohorts and fields is that there is a temptation to generalize from these data to the enterprise as a whole. Thus at this very introductory stage we want to be clear as to what these data represent and what they do not. The data displayed here were provided by institutions selected in a national competition that invited graduate schools to record their own history of completion and attrition, craft strategies to address issues, implement those strategies, and measure their impact in part by continued tracking of student completion. Participants were selected for inclusion based on the belief that they were committed to carrying through with these tasks. As it turned out, the project also represents a set of institutions that are broadly representative of doctoral- granting institutions: public and private, large and small, geographically dispersed universities, with reasonably diverse missions regarding doctoral education. Nonetheless, we do not claim this data set to represent the universe of doctorate-granting universities or programs in the U.S. and Canada. The sample is limited in both fields covered and characteristics of institutions participating. But the field coverage does provide good insight into core disciplines as well as into most major broad fields of doctoral study. And the “judgmental sample” does give a window into performance at typical major public and private, geographically dispersed, and large and small institutions. The bias is clearly in the direction of universities and graduate schools tangibly committed to the mission of systematically understanding and acting upon the challenge of increasing completion rates.

Other important data-gathering activities will allow interested parties to consider the universe of research doctoral programs with respect to at least some of the aspects of completion and attrition documented here.1 But the Ph.D. Completion Project institutions as a whole provide a benchmark against which institutions who equally aspire to measure and then act on their completion and attrition challenges can assess their own performance. We are pleased to share this baseline data as a first step in launching a national, even international, discussion about achieving success in doctoral education.

1 National Academy of Science, National Research Council, An Assessment of Research-Doctorate Programs, forthcoming.

ix Acknowledgments By Debra W. Stewart, CGS President

Many organizations and individuals deserve thanks for their commitment to this project, for supporting it financially and for bringing it to fruition through their labors and their leadership. Let me begin with our funders. First thanks go to Pfizer, Inc., for both its very generous financial support of the infrastructure for the Completion Project and for the specific funding for the entire fields of the sciences, engineering and mathematics. The deep commitment of Pfizer to developing the domestic talent pool in America and to building a global talent pool emerged in the very first conversation we had about increasing completion as a key strategy for success. Their commitment has been sustained throughout. There would be no Completion Project without the leap of faith that the leadership of Pfizer Global Research and Development and Pfizer Corporate Human Resources took with CGS in the early days. Deep thanks also go to the Ford Foundation, whose strong support of the humanities and social sciences allowed us to expand the project to those important fields.

The case that we made to Pfizer and the Ford Foundation was based in part on the dialogue that occurred at an invitational workshop that CGS held in spring 2003 with funding from the Alfred P. Sloan Foundation and the National Science Foundation. Thanks also to both the Sloan Foundation and NSF for this critical early support.

Then in the spring of 2004, as a result of the generous programmatic grants provided by Pfizer and Ford, CGS embarked upon one of the largest and most diverse efforts ever undertaken to address the underlying issues behind student completion of and attrition from Ph.D. programs across the sciences, engineering, mathematics, the social sciences, and the humanities. After reviewing a strong pool of 45 proposals, an external advisory committee selected 21 universities for funding as Research Partners to spearhead this effort. The remaining 24 universities were invited to participate as Project Partners. Shortly before this publication went to press, a strengthened commitment to the project by both Pfizer and Ford has enabled the pool of partnering institutions to grow, with the addition of eight new Research or Data Partners and continued funding to 14 Research Partners from the first phase of the project. It is the great partnership between Pfizer and the Ford Foundation that will continue to sustain the Ph.D. Completion Project through 2010.

x Of course, in addition to the funders, we offer our deep thanks to the member universities that are the core partners in this project. We are grateful to all those who have informed and improved this project: the student respondents, the committed faculty, and the graduate deans and senior leaders in graduate education who have served as principal investigators or otherwise supported this important effort. We also acknowledge the good work of the project Advisory Board. To the Advisory Board members, listed in Appendix A, we say thank you for the sound counsel, the thoughtful selection of participants, the willingness to provide periodic advice, and the sustained commitment to reading drafts and offering comment as we begin to publish findings from the initial phase of the project.

Final thanks go to the professional efforts of the CGS staff (and friends) in all phases of this project. At the formative stages of the project I thank Joan Lorden, Les Sims, Carol Lynch, Robert Sowell, Jennifer Slimowitz, and Daniel Denecke for their good efforts and advice as we convened the discussions and developed the proposal to the funders. In the first phase of the project thanks go to Daniel Denecke for his very important program direction and Helen Frasier and Matthew Loveless for their tireless efforts in data collection. And in the current phase of the project, including data analysis, authoring publications, and continuing project leadership, I offer very special thanks to Robert Sowell, the director of the Ph.D. Completion Project, as well as Kenneth Redd, Ting Zhang, and Emily Neubig who have provided their expertise in data collection and analysis and to Lewis Siegel, for his especially helpful ongoing advice and counsel. Like everything at CGS, the volumes emerging from this project are the result of a team effort. But it is only fair to recognize the unique leadership of Robert Sowell and the special and determined labors of the Sowell, Redd, and Zhang team that brought this current publication to fruition.

xi xii Chapter 1 Introduction

Doctoral education is the “jewel in the crown” of American higher education. Judged by standards of research productivity, quality of education, and alumni success, U.S. research universities have merited highest honors for their doctoral programs. In the area of Ph.D. productivity, however, many graduate students who begin these programs do not complete them. In response to growing national concern about high levels of attrition from doctoral programs, the Council of Graduate Schools (CGS) has developed the Ph.D. Completion Project, a research and demonstration program that examines and documents attrition and completion patterns at a variety of major institutions, develops and models intervention projects, and studies and validates the impact of these interventions upon Ph.D. completion.

In conjunction with this project, CGS published Ph.D. Completion and Attrition: Policy, Numbers, Leadership and Next Steps (Council of Graduate Schools) in 2004. This report helped set the agenda for the project. It summarized the current state of knowledge about completion of and attrition from doctoral programs in the U.S. and Canada and described measures that research universities were taking to increase Ph.D. completion rates in North America. This publication developed out of a white paper produced by the Council of Graduate Schools in the summer of 2003.

The current report, the first in a new series, provides an overview of the CGS Ph.D. Completion Project and focuses on the baseline program completion and attrition data from the 30 universities that participated in the first phase (2004-2007) of the project. The data are analyzed by discipline, broad field, cohort size, and institutional type (public or private, non-profit). This is the first of three reports that will present and analyze project baseline data. The two subsequent baseline data reports will focus on: (1) completion and attrition by demographic characteristics and (2) exit surveys collected both from students who complete their programs and from those who do not. Further publications will report on self-assessments and interventions being implemented by the participating institutions and will provide more in-depth analysis of completion and attrition data, including the impact of intervention projects on completion and attrition.

1 These publications are not intended to serve as best-practice monographs, which CGS generally issues at the conclusion of major grants initiatives.2 In 2010, CGS will issue such a publication, which will include a comprehensive analysis of the quantitative and qualitative data submitted by the partnering universities in Phases I and II of the Ph.D. Completion Project, as well as a comprehensive description of those policies and practices that appear to have had a demonstrated effect on completion rates and attrition patterns over time. The wide range of institutions participating in the project has been designed to ensure that the findings and practices that emerge will be representative of and applicable to the majority of North American graduate institutions engaged in the doctoral enterprise.

2 Examples of such best-practice monographs include: Preparing Future Faculty in the Sciences and Mathematics: A Guide for Change by Anne S. Pruitt-Logan, Jerry G. Gaff, and Joyce E. Jentoft. (Washington, DC: Association of American Colleges and Universities, 2002, retrieved at World Wide Web http://www.preparing-faculty.org/PFFWeb.PFF3Manual.htm) and Preparing Future Faculty in the Humanities and Social Sciences: A Guide for Change by Jerry G. Gaff, Anne S. Pruitt-Logan, Leslie B. Sims, and Daniel D. Denecke (Washington, DC: Associa- tion of American Colleges and Universities, 2003, retrieved from World Wide Web at http://www.preparing-faculty.org/PFFWeb.PFF4Manual.htm); and Professional Master’s Education: A CGS Guide to Establishing Programs by Leslie B. Sims (Washington, DC: Council of Graduate Schools, 2005)

2 Chapter 2 The CGS Completion Project: Context and Overview

The National Context, Including Prior Studies hile there is not complete agreement on the appropriate measurement of doctoral completion or attrition, commentators regularly decry the Wsignificant waste of student talent and university resources implied by high doctoral student attrition. Several prior studies (e.g., Bowen & Rudenstine, 1992; Espenshade & Rodriguez, 1997; and Lovitts, 2001) have suggested that close to half of all U.S. doctoral students fail to complete their degrees. It is particularly difficult to accept or even understand such a rate, given that law schools and medical schools typically report completion rates at 95% and above, while some highly selective colleges also typically report undergraduate completion rates of 95%, even for members of underrepresented racial/ethnic minority groups.

Although there have been no nationwide longitudinal studies that cover a large number of universities or disciplines and that calculate completion in a uniform way, smaller studies have reported Ph.D. completion rates ranging from a low of 33.4% (in the humanities and social sciences [Bowen & Rudenstine, 1992]) to a high of 76% (for students in biomedical and behavioral sciences supported on NIH National Research Service Award training grants [Pion, 2001]). Summarized in Table 1 below, these studies suggest that even under highly favorable conditions, no more than three-quarters of the students who enter doctoral programs complete their degrees. They also suggest that completion rates are higher in the Physical and Life Sciences than in the Social Sciences and Humanities, higher for men than for women, higher for majority than for minority students, and higher in smaller than larger doctoral programs. In addition, demographic studies show that completion rates are higher for international students than for students who are U.S. citizens and permanent residents.

3 Other Variables Significant White students higher completion than minorities; foreign students higher completion than U.S. degree led to higher persistence White & international students higher completion than African Americans in STEM & social sciences than U.S. students GRE, master’s Reduced cohort size & increased financial aid and student quality have positive impact Smaller cohorts higher completion than larger Foreign students higher completion White students higher completion than minorities Public higher attrition than private; no cohort size effect Integrated students more likely to complete Effects Gender No effect Greater $, higher higher higher higher higher Effects * Disciplinary Humanities and related social sciences considered Rate (Avg. 43%) 39% (attr.); 50% (comp.) 13-82% H < S N Men Type of Study Completion Completion Years Addressed 1991-2001 Attrition/ 1995-2001 Completion 50-75% H < S P, N No effect 1970-791989-93 Completion 4-yr persistence 74% 30-66% 19501950-531962- 76 Attrition Attrition Completion1962-76 50-65% 29-50% 20-40% Completion1979-88 H < S N H > S P L 50-54% Completion1975-2000 Men 1978-79 Completion 71-74%1976-78 Completion S < P L1978 76% Completion 37-72% Men H < P L 33-75%1980-89 H < S N Men Attrition1984-891982-84 11-61% Attrition Attrition1992 18-37% 33-68% Completion H < P L 46-74% H < S P L 10 31 24 10 11 3 many 1990-91 2-yr persistence 80% Only math 2 Author Institutions Groen, Jakubson, Ehrenberg, Condie, Liu Nettles & Millett 21 Medical Research Council of Canada Attiyeh Berelson Tucker Bowen & Rudenstine Espenshade & Rodriguez NSF GRF NIH NRSA Nerad & CernyMiselis et al. 1 Zwick 1 Golde [1998] 2 Golde [2005] 1 Case & Blackwelder Lovitts Canadian cohort Table 1 Summary of Prior Major Studies Ph.D. Completion and Attrition Patterns Table

4 Other Variables Significant White students higher completion than minorities; foreign students higher completion than US degree led to higher persistence White & international students higher completion than African Americans in STEM & social sciences Smaller cohorts higher completion than larger Foreign students higher completion than US students White students higher completion than minorities Public higher attrition than private; no cohort size effect Integrated students more likely to complete GRE, master’s Reduced cohort size & increased financial aid and student quality have positive impact Effects Gender higher higher higher No effect Greater $, higher higher L = Life Sciences. Effects * Disciplinary considered Humanities and related social sciences Rate (Avg. 43%) 39% (attr.); 50% (comp.) 13-82% H < S N Men N = Natural Sciences, Type of Study Completion Completion P = Physical Sciences, Years Addressed 1978 1976-78 Completion 33-75% H < S N 1978-79 Completion 37-72% H < P L Men 1962- 76 Completion1962-76 50-65% Completion1979-88 H < S N 50-54% Completion1975-2000 Men Completion 71-74% S < P L 76% Men 1950-53 Attrition 29-50% H > S P L 1950 Attrition 20-40% 1980-89 Attrition1984-891982-84 11-61% Attrition Attrition19921970-79 18-37%1989-93 33-68% Completion Completion H < P L 4-yr persistence 74% 46-74% 30-66% H < S P L 1991-2001 Attrition/ 1995-2001 Completion 50-75% H < S P, N No effect 3 10 11 24 2 31 many10 1990-91 2-yr persistence 80% Only math Author Institutions = Humanities, S Social Sciences, H Miselis et al.Zwick 1 Nerad & Cerny 1 Bowen & Rudenstine Espenshade & Rodriguez NSF GRF NIH NRSA Tucker Berelson Golde [1998] 2 Golde [2005]Lovitts 1 Canadian cohort Medical Research Council of Canada Attiyeh Case & Blackwelder Groen, Jakubson, Ehrenberg, Condie, Liu Nettles & Millett 21 •

5 The first 16 of the studies listed in Table 1 were discussed in more detail in the 2004 Ph.D. Completion and Attrition (Council of Graduate Schools) report. The first of the two additional studies cited in this table, the study by Groen et. al (2005), is a report from the Cornell Higher Education Research Institute on the Mellon Graduate Education Initiative (GEI), which documented the effects of an $80-million initiative to improve Ph.D. programs in the humanities and related social sciences through grants to 51 top-ranked graduate departments from 1991 to 2001. This study showed decreases in average attrition rates of 3 percentage points for students who entered their programs between 1982 and 1990, which were made possible through increases in financial aid, improved student quality, and reductions in cohort size (Groen et. al 2005; Ehrenberg et. al 2006).

The final study listed in Table 1 is by Nettles and Millet (2006), who conducted the largest survey to date of American doctoral students. Their analysis was based on a sample of over 9,000 students from 21 institutions and 11 fields of study. They found an overall six-year completion rate of 62% among the sample of students who had completed their first year of doctoral study when surveyed. They also found that African Americans completed at lower rates than white and international students in engineering, the sciences, mathematics, and the social sciences—and that Hispanics completed at lower rates than whites only in engineering. In all fields, research productivity appeared to be an important predictor of Ph.D. completion. In education, engineering, and the social sciences, having a mentor was positively related to degree completion. And students with children were more likely to interrupt their doctoral enrollment.

Studies such as these have generated growing public interest in Ph.D. completion rates in the U.S. for a number of reasons: demographic changes both within universities and in their surrounding populations, heightened public discussion of the need for greater accountability in higher education, and a widely recognized imperative to deepen the domestic pool of high-end talent in order to meet the employment needs of the twenty-first century. In addition, as the focus of graduate program assessment shifts from inputs to outcomes, Ph.D. completion rates are increasingly recognized as measures of graduate program quality. This link to program quality is reflected in the National Research Council’s decision to include completion rate data in its most recent assessment of U.S. research doctorate programs (forthcoming).

6 Project Overview In response to rising concerns about the costs and implications of Ph.D. attrition, the Council of Graduate Schools initiated its Ph.D. Completion Project. In 2004, CGS engaged senior administrators, faculty, graduate students, and the graduate education research community from across the spectrum of research universities in the U.S. and Canada in an effort to enhance our understanding of Ph.D. completion rates, attrition patterns, and the institutional factors that influence them. The broader goal of the project is to empower universities with proven strategies for positive change. The Ph.D. Completion Project is now a six-year, grant-funded initiative that addresses the issues surrounding Ph.D. completion and attrition. During the first phase of this project (2004- 2007), the Council of Graduate Schools, with generous support from Pfizer, Inc. and the Ford Foundation, provided funding to 21 major U.S. and Canadian research universities to create and pilot intervention projects and to evaluate the impact of these projects on doctoral completion rates and attrition patterns. An additional 24 project partner universities participated in various aspects of this project.3 This pool of universities is expanding with the second phase (2007-2010), in which 22 research partners and several project partners are included. The Ph.D. Completion Project aims to produce a cadre of graduate deans who, as leaders on their campuses and in the national community, can speak to the importance of reducing attrition in doctoral programs and can point to specific strategies for increasing doctoral completion.

One of the main goals of this project is to identify proven strategies to increase Ph.D. completion rates of underrepresented racial/ethnic minorities (African Americans, Latinos, and Native Americans) in all fields, as well as the completion rates of women, especially in engineering, mathematics, and the sciences—fields in which their overall completion rates are lower than those of men. CGS hopes that this project will produce documented evidence about interventions that have proven effective for men and women and for minority and majority students, overall and by field.

As the project develops and additional data are submitted and analyzed, one long-term goal will be to document the net impact of clusters of interventions. Some of these may be documented to be most effective within specific fields and programs, whereas other interventions may work better in some institutional contexts than in others. While we recognize that the project will probably be unable to isolate one strategy from all others as having a decisive

3 For a complete listing of Research Partners and Project Partners in Phase I of the project, see Appendix B.

7 effect on completion, correlations of cluster-effects should be demonstrable, and case studies will supplement quantitative analysis. Finally, because CGS encourages its member universities to engage each other in a constant exchange of ideas about those practices that appear to be having an early positive impact and to make mid-course corrections where appropriate, we have not attempted at this stage of the project to identify “controls” (though it may be possible to do so in retrospect at a later stage of the project).

8 Chapter 3 Collection of Baseline Completion Data at the Program Level by Discipline and Broad Field

or Phase I of the Ph.D. Completion Project, program-level completion and attrition data were submitted by 30 institutions in 2004 and 2005. FCovering twelve academic years starting in 1992-93 and ending in 2003- 04, the data represent 330 programs and 49,113 students in 62 disciplines. (See Appendix C for more details.) The primary motive for collecting these data has been to give participating universities and programs a baseline from which to measure the impact of their proposed intervention strategies. According to many of the participating universities, a secondary benefit has been the increased communication among the various units that collect these data on each campus (e.g., the participating departments, the institutional research offices, and the graduate schools), which has led to more consistent data definitions within their institutions and a better understanding of the data’s meaning and value.

Beyond its usefulness to participants in this project, we believe that the following analysis of program-level baseline data will be valuable to the larger graduate and research community, as well as to the other stakeholders in the enterprise of doctoral education. As noted in the first study in the current CGS series, Ph.D. Completion Project: Preliminary Results from Baseline Data (Denecke and Frasier, CGS Communicator, 2005).

A system for providing more complete and accessible information about graduate completion and attrition is in the interest of federal agencies (which must decide how to fund doctoral education and where to direct those funds), of research universities (with responsibility for recruiting students, providing the education, and granting the degrees), and doctoral students. A more transparent system of data-sharing on graduate completion and attrition is particularly important for prospective students who must decide whether and where to spend a significant portion of their lives in pursuit of a Ph.D. degree. (p.5)

9 The growing need for accurate data on Ph.D. attrition and completion rates by program and institution also reflects two national trends: the demand for greater accountability at all levels of higher education and the increased use of attrition and completion rates among the outcomes upon which program quality is judged. But collecting accurate–and meaningful–data on Ph.D. completion and attrition is fraught with challenges: e.g., differences among universities and even individual programs as to when doctoral study begins, the difficulty of tracking students who “stop out” but do not officially terminate their programs, and the difficulty of tracking and classifying students who transfer from one institution to another or from one doctoral program to another within the same university. Moreover, as Denecke and Frasier (2005) point out in an article in the November 2005 edition of the CGS Communicator, “Meaningful attrition rates cannot be calculated simply by subtracting the number of students who complete from the number of enrolled students [within a given time frame],” because some of those students enrolled at the end of the time frame studied will go on to complete their degrees in subsequent years (“Ph.D. Completion Project: Preliminary Results from Baseline Data,” p. 2). It is also important to understand variations in attrition and completion patterns by discipline and broad field, including whether, within a given field, attrition is most likely to occur early or late in students’ programs.

Finally, the data generated by this study should be useful to universities and programs other than the project participants as they scrutinize their own attrition and completion patterns, as well as to professional societies seeking to increase completion rates within the disciplines they represent. As the following data analysis will show, these patterns differ significantly both within and across broad fields, so that a “target” completion rate for Ph.D. programs in engineering may not be appropriate for Ph.D. programs in the humanities. On the other hand, the average completion and attrition rates presented in this study by discipline and broad field may serve as benchmarks for other programs in the same disciplines as they are assessed or as they assess themselves.

Four features of the baseline data from the Ph.D. Completion Project may significantly advance our understanding of Ph.D. completion and attrition patterns nationally: the templates and methodology that CGS has developed for the collection of baseline attrition and completion data in order to ensure consistency, the longitudinal and nuanced analyses of those data, the size of the database, and the diversity of programs and institutions contributing to it.

10 Templates Used for Data Collection CGS developed three templates to collect the quantitative baseline data on completion and attrition patterns for the Ph.D. Completion Project: the completion data template, the attrition data template, and the aggregate demographic completion data template. CGS also designed a survey instrument to get feedback from students, both completers and non-completers, on the factors they believe contributed to their success, or lack thereof, in completing the Ph.D. To protect the identity of individual students, demographic data and completion/attrition data are collected and initially presented separately. However, for the later analysis of policy factors and interventions that affect completion and attrition, data will be combined at the institutional or disciplinary level.

In the following paragraphs, the two templates that were used to collect completion and attrition data at the program level are described. In subsequent publications, the template used to collect the demographic data and the student survey instrument will be described. Each of the templates provides standard definitions for the collection and reporting of completion and attrition data.

Ph.D. Completion Data Template The completion template (Appendix D) was designed to collect twelve years of program-level completion data, aggregated by cohort, for selected doctoral programs at each participating institution. Data were collected on the size of the entering cohort of doctoral students, the number who withdrew with a master’s degree, the number admitted to doctoral candidacy, and the number of doctoral degrees awarded annually (not cumulative) in each of years three to ten beyond initial enrollment. In addition to the aggregate numbers, data are collected in response to three questions:

1) Does your doctoral program require a master’s degree prior to admission? 2) Does your doctoral program have a continuous enrollment policy? If so, when is this policy effective? For all entering students? For students admitted to candidacy? 3) Does your program or institution distinguish between those seeking a master’s and those seeking a Ph.D.?

To ensure consistency of data submitted to the project, in the completion template CGS also established “default” definitions of both “cohort” and

11 “candidacy,” which are frequently defined in different ways across universities and programs. The template also provides an option with which universities can diverge from the default definition of candidacy, if institutional and/or program data cannot be adjusted to match the default. But by the establishment of default definitions, comparability between data from different institutions is significantly enhanced.

Completion Data Definitions

Cohort: students entering a doctoral program during a given academic year (summer, fall/winter, or spring)

Candidacy: the successful completion of coursework and qualifying examinations

Alternative definitions of Candidacy:

A. Successful completion of preliminary exams and/or defense of the dissertation prospectus B. Award of the master’s degree signifies admission to doctoral candidacy C. Candidacy is not defined or granted by the institution D. Other

Ph.D. Attrition Data Template The attrition template (Appendix D) captures year-by-year data, aggregated by cohort, on the attrition, enrollment, or completion status of students from the point of enrollment through ten years, as well as data that provide information on those enrolled for more than ten years. The attrition data effectively complement the completion data for further investigation of policy factors related to attrition and completion. The identified categories of attrition are as follows: withdrawal without a master’s degree; withdrawal with a master’s degree (before and after candidacy); transfer to another Ph.D. program; temporary leave (“stopping out” from the Ph.D. program, for personal, family, financial, or other reasons, with intention to return); and “information unknown.”

12 One challenge in analyzing the meaning of early attrition is the difficulty in determining the reliability of a student’s stated degree objective. A significant portion of early Ph.D. attrition may involve students who declare their intention to pursue a doctoral degree when they actually intend to leave with only a master’s. They may do so in order to gain admission to and/or funding from programs that favor doctoral students either partially or exclusively over master’s students. Although it is difficult to determine students’ intentions exactly, fields distinguishing between withdrawal without or with a master’s degree (before and after candidacy) are included for years one to four of the data template. These categories may help distinguish between students whose degree objective was actually a master’s degree from those whose degree objective was the Ph.D. Beginning in year five, data on withdrawal were collapsed into a single data column, “Withdrew.” The collapse of data reflects our belief that the Ph.D., not the master’s, is the likely degree objective of students who continue to enroll in a doctoral program beyond the fourth year. Data were also collected on the number of students continuing and the number of Ph.D. degrees awarded annually. As in the completion template, data in the attrition template reflect annual, not cumulative, values for the categories of attrition, completion, and enrollment.

The next chapter of this report presents selected analyses of the baseline program completion and attrition data.

13 Chapter 4 Completion and Attrition Trends: Analysis of Baseline Program Data

s previous chapters have indicated, increasing attention is being paid to students’ completion of their doctoral programs, and several Astudies of completion and attrition rates have been conducted over the past thirty years. Data collected by the Ph.D. Completion Project can be expected to provide a different picture of completion rates and patterns than those presented in studies conducted previously. However, the results of the Completion Project are not directly comparable to those from other studies, due to differences in the studies’ designs, numbers of participating institutions, and other factors. This chapter analyzes the Completion Project’s baseline data on doctoral student completion and attrition rates.

The analysis will be presented for three groups of entering cohorts. Most of the analysis will focus on ten-year completion rates for the cohorts that started their doctoral programs from 1992-93 through 1994-95. These cohorts will be referred to as A-cohorts. First, the overall cumulative average ten-year completion rates will be described and compared by broad field of study, discipline, institutional type, and cohort size. Then, to investigate the changes, if any, in average completion rates over time, comparisons will be made of the seven-year completion rates between two cohort groups: the A-cohorts, as described above, and B-cohorts, which include students who started between 1995-96 and 1997-98. Attrition rates will then be analyzed across disciplines, institutional types, and cohort sizes for the A-cohorts, and early attrition analyses will be presented for A- and B-cohorts, as well as for students who started their programs from 1998-99 through 2000-01, the C-cohorts.

Ten-Year Completion Rates The data (from the A-cohorts) used in the analysis of ten-year completion rates at the aggregate and broad-field and discipline levels consisted of 12,135 students across 58 disciplines. The students and disciplines were distributed across five broad fields as follows: Engineering (17.8%), Life Sciences (12.6%), Mathematics & Physical Sciences (30.1%), Social Sciences (22.0%), and Humanities (17.6%) (see Appendix C).

14 Analysis of data from the Ph.D. Completion Project suggests that almost half the students who enroll in doctoral programs within this broad set of fields complete their doctoral degrees within seven years. As Figure 4-1 shows, 45.5% of the doctoral students who started their programs in 1992-93 through 1994-95 had completed within seven years. By the end of the tenth year, however, the completion rate grows to 56.6%. The 11.1 percentage point growth in completion rates between the seventh and tenth years also indicates that there is a great deal of progress toward degree completion during these three years. As later figures in this chapter will show, Ph.D. completion after the seventh year of study varies by broad field of study, academic program, type of institution, student cohort size, and other factors.

Figure 4-1 Average Cumulative Overall Ten-Year Completion Rates for FigureStudents 4-1 Average Entering Cumulative Doctoral Overall Programs Ten-Year from Completion1992-93 through Rates 1994-95,for Students by EnteringYears. Doctoral Programs from 1992-93 through 1994-95, by Year

60 56.6 54.6 50.9

45.5 40 36.1

22.5 20

Cumulative Completion Rate (%) 10.5

4.5 0

345678910 Source: Council of Graduate Schools Completion and Attrition Program Data

Completion Rates by Broad Field

Completion rates vary a great deal between Science, Engineering and Mathematics (SEM) fields and Social Sciences and Humanities (SSH) fields. The average completion rate in SEM fields is noticeably higher than that for SSH fields. However, as Figure 4-2 shows, the average SSH combined rate of completion appears to increase steadily even at the ten-year mark, which suggests that a significant number of SSH students continue their studies after a decade. 15

In the SEM fields, more than half the students complete their doctorates by year seven, and by the tenth year, almost 60% of SEM students have received their Ph.D.s. While between years seven and ten, the overall completion rate for SEM fields grows by 7.2 percentage points (from 51.9% at the seventh year to 59.1% at the tenth year), the rate for SSH students rises by 17.2 percentage points (from 35.8% at the seventh year to 53.0% at the tenth year). At year ten, the overall completion rate in SSH fields is still increasing at a steady pace, but in SEM fields growth in the completion rate has become minimal. Completion Rates by Broad Field Completion rates vary a great deal between Science, Engineering and Mathematics (SEM) fields and Social Sciences and Humanities (SSH) fields. The average completion rate in SEM fields is noticeably higher than that for SSH fields. However, as Figure 4-2 shows, the average SSH rate of completion appears to increase steadily even at the ten-year mark, which suggests that a significant number of SSH students continue their studies after a decade. In the SEM fields, more than half the students complete their doctorates by year seven, and by the tenth year, almost 60% of SEM students have received their Ph.D.s. While between years seven and ten, the overall completion rate for SEM fields grows by 7.2 percentage points (from 51.9% at the seventh year to 59.1% at the tenth year), the rate for SSH students rises by 17.2 percentage points (from 35.8% at the seventh year to 53.0% at the tenth year). At year ten, the overall completion rate in SSH fields is still increasing at a steady pace, but in SEM fields growth in the completion rate has become minimal.

Figure 4-2 Average Cumulative Ten-Year Completion Rates for SEM and SSH Cohorts Beginning Doctoral Study from 1992-93 through 1994-95, Figureby Year4-2 Average Cumulative Ten-Year Completion Rates for SEM and SSH Cohorts Beginning Doctoral Study from 1992-93 through 1994-95, by Year

60 50 40 30 Completion Rate (%) Rate Completion

20 10 Cumulative 0

3 4 5 6 7 8 9 10 Year

SEM SSH

Source: Council of Graduate Schools Completion and Attrition Program Data

Of course,16 there is a great deal of variation in completion rates among the five individual broad fields, as Figure 4-3 illustrates. Engineering has the highest Ph.D. completion rate throughout almost all the ten-year time period, while Humanities has the lowest. From years four to seven, Engineering has a substantially higher completion rate than the other fields. Starting in year eight, however, the completion rate for Life Sciences approaches that of Engineering, and by the tenth year Engineering has a only slightly higher completion rate (63.6%) than Life Sciences (62.9%). The field of Mathematics & Physical Sciences has a higher overall completion rate than Social Sciences at year seven, but by year ten this pattern is reversed, and the Ph.D. completion rate for Social Sciences (55.9%) has slightly surpassed that of Mathematics & Physical Sciences (54.7%).

Of course, there is a great deal of variation in completion rates among the five individual broad fields, as Figure 4-3 illustrates. Engineering has the highest Ph.D. completion rate throughout almost all the ten-year time period, while Humanities has the lowest. From years four to seven, Engineering has a substantially higher completion rate than the other fields. Starting in year eight, however, the completion rate for Life Sciences approaches that of Engineering, and by the tenth year Engineering has a only slightly higher completion rate (63.6%) than Life Sciences (62.9%). The field of Mathematics & Physical Sciences has a higher overall completion rate than Social Sciences at year seven, but by year ten this pattern is reversed, and the Ph.D. completion rate for Social Sciences (55.9%) has slightly surpassed that of Mathematics & Physical Sciences (54.7%).

Figure 4-3 Average Cumulative Ten-Year Completion Rates for Cohorts FigureEntering 4-3 Average Doctoral Cumulative Study from Ten-Year 1992-93 Completion through Rates 1994-95, for Cohorts by Broad Entering Field Doctoraland Year Study from 1992-93 through 1994-95, by Broad Fields and Year

70 63.6 56.8 62.9 60 55.9 53.7 54.7 50 48.2 49.3 40.9 40

30 29.3 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Engineering Life Sciences Mathematics & Physical Sciences Social Sciences Humanities

Source: Council of Graduate Schools Completion and Attrition Program Data

Figure 4-3 also shows that the timing for reaching the 50% completion rate in Figurethe 4-3five also fields shows varies that considerably. the timing for Inreaching Engineering, the 50% the completion 50% completion rates in the rate five is fieldsreached varies shortlyconsiderably. after the In six-yearEngineering, mark; the in 50% Life completion Sciences, it rate occurs is reached between shortly years after6 theand six-year 7; in Mathematics mark; in Life & Sciences, Physical it Sciences,occurs between at about years year 6 and 7.5; 7; andin Mathematics in Social & Physical Sciences, at about year 7.5; and in Social Scienes, at about year 8.5. The average completion rate for Humanities does not exceed 50% during the ten-year period; however, it jumps considerably between years seven and ten. Conversely, the average completion rate for Engineering starts to plateau after the eighth year, which suggests17 that only a small number of Engineering students completed the Ph.D. after their eighth year of study. The plateau occurs in Life Sciences and Mathematics & Physical Sciences after the ninth year, indicating that limited degree completion would occur after that time. Completion rates in Social Sciences and Humanities appear to continue to rise after year ten; it is therefore possible that a significant number of students in these two fields will continue their studies and complete their degrees after their tenth year.

The differences in completion rates across the five broad fields may be due to various reasons, such as variations in the amount and duration of financial support, the quality of academic advising and mentoring, dissertation and degree requirements, and future job prospects. While some of these causes may be beyond the control of graduate deans and other program administrators, institutional and program policies may influence future changes in completion rates. Investigating the factors that affect timely completion is one Scienes, at about year 8.5. The average completion rate for Humanities does not exceed 50% during the ten-year period; however, it jumps considerably between years seven and ten. Conversely, the average completion rate for Engineering starts to plateau after the eighth year, which suggests that only a small number of Engineering students completed the Ph.D. after their eighth year of study. The plateau occurs in Life Sciences and Mathematics & Physical Sciences after the ninth year, indicating that limited degree completion would occur after that time. Completion rates in Social Sciences and Humanities appear to continue to rise after year ten; it is therefore possible that a significant number of students in these two fields will continue their studies and complete their degrees after their tenth year.

The differences in completion rates across the five broad fields may be due to various reasons, such as variations in the availability, amount and duration of financial support, the quality of academic advising and mentoring, dissertation and degree requirements, and future job prospects. While some of these causes may be beyond the control of graduate deans and other program administrators, institutional and program policies may influence future changes in completion rates. Investigating the factors that affect timely completion is one of the important goals of the Ph.D. Completion Project. Later publications in the Ph.D. Completion and Attrition series will examine these policies more closely.

Completion Rates by Discipline This section examines the differences in Ph.D. completion rates at the disciplinary level. The disciplines used here are based on definitions developed by the National Research Council. As previously stated, the Ph.D. Completion Project includes data from 330 doctoral programs at 30 institutions. However, in order to have a sufficient number of observations for the disciplinary analysis, only disciplines with five or more programs in the project were included. Again, only the A-cohorts, students who entered their doctoral programs between 1992-93 and 1994-95, are included in the analysis. Therefore, data from 258 programs are represented in the disciplinary analysis in this section. (See Appendix C for the number of programs and students in each discipline.) This section first presents an analysis of the cumulative completion rates over ten years for students in each of the disciplines represented by five or more programs within the five broad fields. The cumulative completion rates are shown in the line graphs.

18 Additionally, box-and-whisker plots are used to show the distribution of completion rates at year ten for each discipline that has five or more participating programs. The box plots are used to illustrate a graphical summary of the sample distribution indicators across institutions within each discipline: a measure of central location (the median4, shown as the line inside the box in Figure 4-5, for example), two measures of dispersion (the range, shown as the whisker, and the inter-quartile range5, shown as the box), the skewness (from the orientation of the median relative to the quartiles) and potential outliers (marked as individual dots). Each observation in the box plot represents a program. With the above set of indicators, the sample distributions of disciplines within a broad field can be compared. The range normally labels the minimum and maximum values that are less than or equal to 1.5 times the inter-quartile range below the first quartile or beyond the last quartile. The outliers are the observations outside the range.

The reasons for including the box charts are threefold. First, the box charts offer complementary information to the line graphs because these two types of graphs have different units of analysis. The line graphs present a completion rate computed across all sampled students, while the box charts present completion data across programs. Second, the two types of charts present completion rates over different time spans. The line graphs present data across ten years, while the box charts present data for only the tenth year. Third, these two types of charts exhibit different statistical information. The line graphs provide information on overall completion rates by year, while the box charts display the medians, variance and inter-quartile ranges across the completion rates for each program.

4 The median value is the half-way point in a distribution. Half the programs have completion rates that are above the median, and half have rates that are below. 5 The inter-quartile range is equal to the length of the box in a box plot. It includes the middle 50% distribution of the observations. Within the distribution of the completion rates, the lowest quarter of the observations are less than or equal to the first quartile, and the highest quarter are greater than or equal to the highest quartile.

19 displayEngineering the medians and inter-quartile ranges across the completion rates for each program.Civil Engineering has substantially higher completion rates than the other Engineering disciplines. At year seven, Civil Engineering has already reached aEngineering completion rate of 69%, as shown in Figure 4-4. Chemical Engineering and Mechanical Engineering have very similar completion rates from years Civilthree Engineering through has ten. substantially For those higher two disciplines, completion Ph.D.rates than completion the other Engineering rates reach disciplines. At year seven, Civil Engineering has already reached a completion rate of approximately 60% at year eight, but from year eight to year ten, the completion 69%, as shown in Figure 4-4. Chemical Engineering and Mechanical Engineering have veryrates similar increase completion by only rates 2 or from 3 percentage years three points. through While ten. For Biomedical those two Engineeringdisciplines, Ph.D.starts completion with one rates of the reach highest approximately early completion 60% at year rates eight, in this but group from yearat year eight three, to yearits ten, completion the completion rate doesrates increasenot increase by 2 oras 3fast percentage as other points. Engineering While Biomedical disciplines Engineeringfrom year starts three with throughone of the year highest six. early Electrical completion & Electronicsrates in this group Engineering, at year three,however, its completion has the rate lowest does ratenot increaseof completion as fast as after other year Engineering five. Completion disciplines rates from yearin three all fourthrough engineering year six. Electrical disciplines & Electronicsbegin to plateau Engineering, after yearhowever, seven, has though the lowestBiomedical rate of completion Engineering after and year Civil five. EngineeringCompletion rates still inhave all fourslight engineering growth. disciplines begin to plateau after year seven, though Biomedical Engineering and Civil Engineering still have slight growth. Figure 4-4 Average Ten-Year Cumulative Completion Rates for Students FigureEntering 4-4 Average Five EngineeringTen-Year Cumulative Doctoral Completion Disciplines Rates from for 1992-93 Students through Entering Four1994-95, Engineering by Year Doctoral Disciplines from 1992-93 through 1994-95, by Year

70 60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Biomedical Engineering Chemical Engineering Civil Engineering Electrical & Electronics Engineering Mechanical Engineering

Source: Council of Graduate Schools Completion and Attrition Program Data

Figure 4-5 shows the median and variations in completion rates for the five engineering disciplines at the tenth year. Figure 4-4 shows that Mechanical Engineering has a higher

20 Figure 4-5 shows the median and variations in completion rates for the five engineering disciplines at the tenth year. Figure 4-4 shows that Mechanical Engineering has a higher mean completion rate than Chemical Engineering at year ten; however, the median completion rates for the seven6 Chemical Engineering programs and the eleven Mechanical Engineering programs are the same (66.67%) at year ten (see Figure 4-5).

Also, completion rates for the Chemical Engineering programs have a smaller amount of variance and a smaller inter-quartile range. About half of the Chemical Engineering programs have completion rates in the narrow range between 66% and 70%. Mechanical Engineering programs have the largest variance, the largest inter-quartile range, and an outlier7. Electrical & Electronics Engineering programs have the lowest median completion rate at year ten and have the smallest range and smallest inter-quartile range, but have three outliers. Civil Engineering programs have the highest completion rate at year ten and have a small variance, which indicates that Civil Engineering programs have the most concentrated high completion rates among the Engineering disciplines. Civil Engineering programs also have a small range like Electrical and Electronics Engineering, but they have an inter-quartile range similar to Biomedical Engineering. The smaller range and smaller inter- quartile range normally indicate more homogenous completion rates.

Figure 4-5 Inter-Quartile Ranges of Cumulative Completion Rates for Cohorts Beginning from 1992-93 through 1994-95, for Engineering Disciplines with Five or More Programs, at Year Ten

Biomedical Engineering

Chemical Engineering

Civil Engineering

Electrical & Electronics Engineering

Mechanical Engineering

0 10 20 30 40 50 60 70 80 90 100 Cumulative Completion Rate (%) Source: Council of Graduate Schools Completion and Attrition Program Data

6 Appendix C lists the number of programs in each major discipline. 7 The outlier in Mechanical Engineering is for a program in which there were 11 entering students and none completed.

21 Life Sciences Among the Life Science disciplines, completion rates generally rise quickly from year four to year seven, as shown in Figure 4-6. After year eight, completion Liferates Sciences for most of these disciplines begin to plateau; by this time, more than 50% of the students have completed. The completion rate for Genetics & Molecular Among the Life Science disciplines, completion rates generally rise quickly from year fourGenetics to year seven, surpasses as shown other in Figuredisciplines 4-6. Afterat year year six eight, after completion sharp growth rates forbetween all of theseyears disciplines five and begin six. to Microbiology plateau; by this & time, Immunology more than starts 50% ofwith the the students highest have early completed.completion The completionrate and has rate the for secondGenetic shighest & Molecular completion Genetics rate surpasses among otherall Life disciplinesScience at disciplinesyear six after at sharp year growthseven, betweenand almost years shares five and the six. highest Microbiology completion & Immunologyrate at year starts ten with with the Genetics highest early& Molecular completion Genetics. rate and Biology,has the second Neuroscience, highest completionand Molecular rate among & all Cellular Life Science Biology disciplines start with at year barely seven, any and Ph.D. almost completion shares the highestat yearcompletion three andrate thenat year diverge. ten with At Genetics year eight, & Molecular Neuroscience Genetics. and Biology, Molecular Neuroscience,& Cellular and Biology Molecular have & Cellular similar Biology completion start with rates, barely while any Biology Ph.D. completion has the at year three and then diverge. At year eight, Neuroscience and Molecular & Cellular lowest completion rate among these disciplines. By the tenth year, the Ph.D. Biology have similar completion rates, while Biology has the lowest completion rate amongcompletion these disciplines. rates for By all the the tenth five year,disciplines the Ph.D. range completion between rates 59.4% for and all the69.3%. five disciplines range between 59.4% and 69.3%. Figure 4-6 Average Ten-Year Cumulative Completion Rates for Students FigureEntering 4-6 Average Five Life Ten-Year Science Cumulative Doctoral DisciplinesCompletion fromRates 1992-93 for Students through Entering Five1994-95, Life Science by Year Doctoral Disciplines from 1992-93 through 1994-95, by Year

70 60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Biology Genetics, Molecular Genetics Microbiology & Immunology Molecular & Cellular Biology Neuroscience

Source: Council of Graduate Schools Completion and Attrition Program Data

The distribution of the ten-year program completion rates within the Life Science disciplines (see Figure 4-7) reveals that programs within the discipline of Genetics & Molecular22 Genetics have the highest median completion rate, followed in turn by The distribution of the ten-year program completion rates within the Life Science disciplines (see Figure 4-7) reveals that programs within the discipline of Genetics & Molecular Genetics have the highest median completion rate, followed in turn by Molecular & Cellular Biology, Microbiology and Immunology, Biology, and Neuroscience. The values of the median completion Molecularrates at & the Cellular program Biology, level Microbiologyand their orders and differImmunology, from the Biology, overall and completion Neuroscience.rates across The all valuesstudents of thefor medianeach discipline completion (see rates Figure at the 4-6). program Neuroscience level and theirhas ordersthe differlowest from median the overall completion completion rate at rate thes programacross all level, students compared for each to discipline other Life (see FigureScience 4-6). disciplines.Neuroscience Its has median the lowest completion median ratecompletion is very rateclose at to the its program first quartile, level, compared to other Life Science disciplines. Its median completion rate is very close to its which indicates that half of the Neuroscience programs have completion rates first quartile and percentile rate, which indicates that half of the Neuroscience programs havebetween completion 55% rates and between60%. 55% and 60%.

FigureFigure 4-7 Inter-Quartile 4-7 Inter-Quartile Ranges Ranges of Cumulative of Cumulative Completion Completion Rates for Cohorts Rates for BeginningCohorts in from 1992-93, 1992-93 1993-94, through and 1994-95, 1994-95, for for Life Life Science Science Disciplines Disciplines with with Five or MoreFive Programs, or More At Programs, Year Ten at Year Ten

Biology

Genetics, Molecular Genetics

Microbiology & Immunology

Molecular & Cellular Biology

Neuroscience

0 10 20 30 40 50 60 70 80 90 100 Cumulative Completion Rate (%) Source: Council of Graduate Schools Completion and Attrition Program Data

Mathematics & Physical Sciences

Among the Mathematics & Physical Science disciplines (Figure 4-8), Chemistry has the highest rate of student completion, while Computer & Information Sciences has the lowest. At year ten, the completion rates of the four disciplines within this broad field range from 41.5% to about 61.6%. Most growth in completion occurs from year four to year eight or nine and then reaches a plateau. 23 Mathematics & Physical Sciences Among the Mathematics & Physical Science disciplines (Figure 4-8), Chemistry has the highest rate of student completion, while Computer & Information Sciences has the lowest. At year ten, the completion rates of the four disciplines within this broad field range from 41.5% to about 61.6%. Most growth in completion occurs from year four to year eight or nine and then reaches a plateau.

FigureFigure 4-8 Average4-8 Average Ten-Year Ten-Year Cumulative Cumulative Completion Completion Rates forRates Students for Students Entering MathematicsEntering &Mathematics Physical Science & Physical Doctoral Science Disciplines Doctoral from Disciplines 1992-93 through from 1992- 1994- 95, by93 Yearthrough 1994-95, by Year

60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Chemistry Computer & Information Sciences Mathematics Physics & Astronomy

Source: Council of Graduate Schools Completion and Attrition Program Data

Computer & Information Science programs also have the lowest median Computercompletion & Information rate and Sciencethe smallest programs variance also haveand inter-quartilethe lowest median range completion (see Figure rate and 4-9)the smallest at the program variance level and amonginter-quartile the Mathematics range (see Figure& Physical 4-9) atScience the program disciplines level amongat the the tenth Mathematics year. Mathematics & Physical hasScience the second disciplines lowest at the median tenth completionyear. Mathematics rate at has theyear second ten, but lowest has themedian largest completion variance rate and at inter-quartile year ten, but hasrange, the largestwith the variance highest and inter-quartile range, with the highest completion rate for a single program of about 80%. Thiscompletion high completion rate for rate a singleis similar program to the ofrates about for some80%. ChemistryThis high completionand Physics rate& is Astronomysimilar programs.to the rates for some Chemistry and Physics & Astronomy programs.

24 FigureFigure 4-9 Inter-Quartile 4-9 Inter-Quartile Ranges Ranges of Cumulative of Cumulative Completion Completion Rates for Cohorts Rates for from 1992-93Cohorts through from 1994-95, 1992-93 for throughMathematics 1994-95, & Physical for Mathematics Science Disciplines & Physical with Five or MoreScience Programs, Disciplines At Year with Ten Five or More Programs, at Year Ten

Chemistry

Computer & Information Sciences

Mathematics

Physics & Astronomy

0 10 20 30 40 50 60 70 80 90 100

Source: Council of Graduate Schools Cumulative Completion Rate (%) Completion and Attrition Program Data

Social Sciences Sciences As discussed earlier, the Social Science disciplines tend to have strong Ph.D. As discussedcompletion earlier, rates theafter Social the sixth Science year disciplines of initial enrollment. tend to have Generally, strong Ph.D. completion completion ratesrates after in the most sixth Social year Scienceof initial disciplinesenrollment. appear Generally, to continue completion to increase rates in aftermost yearSocial Scienceten, exceptdisciplines for Economics, appear to continue where tothe increase completion after rateyear appears ten, except to reach for Economics, a plateau where the completion rate appears to reach a plateau at year ten (see Figure 4-10). At between years nine and ten (see Figure 4-10). At year ten, Communications year ten, Communications has the highest average completion rate (67%), followed by Psychologyhas the (65%), highest while average the other completion disciplines rate have (67%), rates that followed range from by 44% Psychology to 52%. The(65%), completion while rate the for other Psychology disciplines increases have quickly rates thatbetween range years from four 44% and to seven. 52%. At yearThe three, completion while three rate of thefor disciplinesPsychology have increases average quickly completion between rates yearsbelow four 10%, and the completionseven. At rate year for Communicationsthree, while three is 24%.of the Sociology disciplines and have Political average Science completion have very similarrates completion below 10%, patterns. the completion They start withrate fora very Communications low average completion is 24%. Sociologyrate (below 5%)and at year Political three andScience then have steadyvery similar growth. completion By year ten, patterns. although They they starthave withthe a lowestvery completion low average rates completion (less than rate45%) (below among 5%) this atgroup year ofthree disciplines, and then there have seems steady to be continuinggrowth. By growth year ten,in their although completion they haverates, the which lowest suggests completion that some rates students (less than may complete after their tenth year. Anthropology & Archaeology has the lowest completion 45%) among this group of disciplines, there seems to be continuing growth in rate through year eight, but between years nine and ten, the completion rate of this disciplinetheir completion accelerates, rates,and by which year ten suggests it surpasses that some Sociology students and Politicalmay complete Science after and

25 their tenth year. Anthropology & Archaeology has the lowest completion rate through year eight, but between years nine and ten, the completion rate of this discipline accelerates, and by year ten it surpasses Sociology and Political Science and reaches a completion rate of 46%. Further, the slope of this curve continues to grow, which suggests that some students may complete their studies after the tenth year.

Figure 4-10 Average Ten-Year Cumulative Completion Rates for Students FigureEntering 4-10 Average Six Social Ten-Year Science Cumulative Doctoral Completion Disciplines Rates from for 1992-93 Students through Entering Six 1994-95,Social Science by Year Doctoral Disciplines from 1992-93 through 1994-95, by Year

70 60 50 40 30 20 10 0 Cumulative Completion Rate (%) Rate Completion Cumulative

3 4 5 6 7 8 9 10 Year

Anthropology & Archaeology Communications Economics Political Science Psychology Sociology

Source: Council of Graduate Schools Completion and Attrition Program Data

TheThe rank rankorder orderof the median of the completion median completion rates for the ratesSocial forScience the disciplines Social Science is basicallydisciplines consistent, is basically with one consistent, minor exception with one (for minor Sociology exception and Anthropology (for Sociology & and Archaeology),Anthropology with & the Archaeology), rank order of theirwith mean the rank completion order of rates, their as mean shown completion in Figure 4- 11. rates,However, as shown there isin a Figure great d 4-11.eal of However,variation in there the medianis a great ten-year deal of completion variation inrates the at the medianprogram ten-year level across completion these disciplines. rates at the Programs program in Economicslevel across and these Communications disciplines. havePrograms the smallest in Economicsvariance and and smallest Communications inter-quartile range,have the while smallest programs variance in and Psychologysmallest haveinter-quartile the largest range, variance while and theprograms largest inter-in Psychologyquartile range. have the largest variance and the largest inter-quartile range.

26 Figure 4-11 Inter-Quartile Ranges of Cumulative Completion Rates for FigureCohorts 4-11 Inter-Quartile from 1992-93 Ranges through of Cumulative 1994-95, forCompletion Social ScienceRates for Disciplines Cohorts fromwith 1992-93 Five through or More 1994-95, Programs, for Social at Year Science Ten Disciplines with Five or More Programs, At Year Ten

Anthropology & Archaeology

Communications

Economics

Political Science

Psychology

Sociology

0 10 20 30 40 50 60 70 80 90 100 Cumulative Completion Rate (%) Source: Council of Graduate Schools Completion and Attrition Program Data

Humanities HumanitiesPh.D. completion rates for disciplines within the Humanities are more homogeneous than those in other broad fields. As Figure 4-12 shows, they Ph.D. completion rates for disciplines within the Humanities are more homogeneous than thosegenerally in other broad start fields.with low As Figureaverage 4-12 completion shows, they rates generally (11% orstart less) with at low year average three completionand then rates have (11% steady or less) growth at year through three and year then ten. have Ten stea yearsdy growth after initial through student year ten. Tenenrollment, years after these initial disciplines student enrollmen all approacht, these a 50%disciplines average all completionapproach a 50%rate, but averagethe completionupward trajectory rate, but theof theupward rates trajec suggesttory ofthat the degree rates suggest completion that degree continues completionto grow continues thereafter. to grow At year thereafter. seven, AtForeign year seven, Languages Foreign & Languages Literature & holds Literature the holdshighest the highest average average completion completion rate rate (34.4%), (34.4%), followed followed by EnglishEnglish Language Language & & LiteratureLiterature (33.0%), (33.0%), Philosophy Philosophy (30.8%), (30.8%), and History and History(24.7%). (24.7%).

27 Figure 4-12 Average Ten-Year Cumulative Completion Rates for Students FigureEntering 4-12 Average Four Humanities Ten-Year Cumulative Doctoral Completion Disciplines Rates from for 1992-93 Students through Entering Four1994-95, Humanities by Year Doctoral Disciplines from 1992-93 through 1994-95, by Year

70 60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year English Language & Literature Foreign Languages & Literature History Philosophy

Source: Council of Graduate Schools Completion and Attrition Program Data

Consistent with the mean completion rates at year ten, the median completion Consistentrates of with programs the mean within completion the Humanities rates at year broadten, the field median also completion concentrate rates at of a programssimilar within value the (around Humanities 50%). broad Programs field alsoin English concentrate Language at a similar & Literature value (around have 50%). Programs in English Language & Literature have the highest median completion the highest median completion rate (as shown in Figure 4-13). The variances rate (as shown in Figure 4-13). The variances and inter-quartile ranges of programs withinand these inter-quartile disciplines rangesdo not vary of programs dramatically, within although these Foreign disciplines Languages do not & vary Literaturedramatically, appears althoughto have the Foreign largest variance.Languages & Literature appears to have the largest variance.

28 FigureFigure 4-13 4-13 Inter-Quartile Inter-Quartile Ranges Ranges of Cumulative of Cumulative Completion Completion Rates for Rates Cohorts for fromCohorts 1992-93 from through 1992-93 1994-95, through for Humanities 1994-95, forFields Humanities with Five orFields More with Programs, Five at Yearor More Ten Programs, at Year Ten

English Language & Literature

Foreign Languages & Literatures

History

Philosophy

0 10 20 30 40 50 60 70 80 90 100

Source: Council of Graduate Schools Cumulative Completion Rate (%) Completion and Attrition Program Data

CompletionCompletion Rates Rates by Institution by Institution Type Type

CompletionCompletion rate by rate institution by institution type is typeanother is factor another to consider. factor to There consider. are some There are differencessome differences between public between and prpublicivate institutionsand private that institutions could influence that could completion influence rates, suchcompletion as sources rates,of funding such andas sources flexibility of funding in using andthose flexibility funds. This in usingsection those thus funds. examinesThis section completion thus examinesrates for public completion and privat ratese universities for public thatand areprivate participating universities in the Completionthat are participatingProject. Of the in 30 the research Completion and project Project. partner Of the universities 30 research within and the project Completionpartner universities Project, 20 are within public the institutions. Completion Project, 20 are public institutions.

As Figure 4-14 shows, the overall average ten-year Ph.D. completion rate at public As Figure 4-14 shows, the overall average ten-year Ph.D. completion rate universities is nearly identical to that at private institutions. Public institutions seem to haveat slightlypublic higheruniversities completion is nearly rates identicalbefore year to five,that butat privatethereafter institutions. private institutions Public haveinstitutions a slightly higherseem toaverage have slightlycompletion higher rate. completion rates before year five, but thereafter private institutions have a slightly higher average completion rate.

29 Figure 4-14 Average Cumulative Ten-Year Completion Rates by Institution FigureType 4-14 for StudentsAverage Cumulative Entering Doctoral Ten-Year ProgramsCompletion from Rates 1992-93 by Institution through Type for1994-95, Students by Entering Year Doctoral Programs from 1992-93 through 1994-95, by Year

60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Private Public

Source: Council of Graduate Schools Completion and Attrition Program Data

In all broad fields except Mathematics & Physical Sciences and Humanities, Inaverage all broad completion fields except rates Mathematics by institution & Physical type are Sciences similar (shown and Humanities, in Figure average 4-15). completionIn Mathematics rates by & institution Physical typeSciences, are similar the two(shown types in Figureof institutions 4-15). In startMathematics with & Physicalnearly the Sciences, same early the two completion types of institutions rate at year start three, with then nearly begin the tosame diverge early after completionyear four, ratewhen at yearprivate three, institutions then begin clearlyto diverge have after higher year four,completion when private rates. In institutions clearly have higher completion rates. In Humanities, private institutions alwaysHumanities, have slightly private lower institutions completion always rates, have while slightly in Engineering, lower completion Social Sciences, rates, and Lifewhile Sciences, in Engineering public and and private Social institutions Sciences, have public nearly and identi privatecal institutionsrates after year have five. In Lifenearly Sciences, identical private rates institutions after year havefive. slightlyIn Life lowerSciences, completion private rates institutions until year have six but surpassslightly public lower institutions completion thereafter. rates until year six but surpass public institutions thereafter.

30 FigureFigure 4-15 4-15 Average Average Ten-Year Ten-Year Cumulative Cumulative Completion Completion Rates Rates by byInstitution Institution Type forType Doctoral for Doctoral Students Students Entering Enteringfrom 1992-93 from through 1992-93 1994-95, through by Broad1994-95, Fields by and YearBroad Field and Year

Engineering Life Sciences Mathematics & Physical Sciences 60 60 60

40 40 40

20 20 20

0 0 0 2 4 6 8 10 2 4 6 8 10 2 4 6 8 10

Social Sciences Humanities 60 60

40 40

20 20 Cumulative Completion Rate (%) Rate Completion Cumulative

0 0 2 4 6 8 10 2 4 6 8 10 Year

Private Public

Graphs by BF Source: Council of Graduate Schools Completion and Attrition Program Data

CompletionCompletion Rates Rates by Cohort by Cohort Sizes Size Some research (e.g., Brown & Rudenstine, 1992) has suggested that cohort Somesize affectsresearch Ph.D. (e.g., completionBrown & Rudenstine, rates. These 1992) previous has suggested studies that have cohort reported size affectsthat Ph.D.students completion in smaller rates. classes These have previous higher studies completion have reported rates, that presumably students in because smaller classesstudents have receive higher more completion individual rates, attention presumably from because their instructors.students receive more individual attention from their instructors. In this section, which compares average completion rates by cohort size, cohort In this section, which compares average completion rates by cohort size, cohort size is basedsize ison based the sample on the distribution sample distribution for the size for of enteringthe size cohortsof entering within cohorts each program within at theeach institutions program participating at the institutions in the Completion participating Project. in the The Completion cohorts are Project.placed into The three groups:cohorts are placed into three groups:

“Small”--“Small”-- cohort size cohort ranging size rangingfrom 1 fromto 7 students 1 to 7 students “Medium”“Medium” --cohort size--cohort ranging size rangifromng 8 fromto 14 8 students to 14 students “Large”--“Large”-- cohort size cohort of 15 size students of 15 students or above or above

Overall, as Figure 4-16 illustrates, among the 874 entering cohorts across all 313 programs at 30 institutions from 1992-93 to 1994-95, Ph.D. completion rates vary only slightly by cohort size during the ten years from initial entry into doctoral study. Small,

31 Overall, as Figure 4-16 illustrates, among the 874 entering cohorts across all 313 programs at 30 institutions from 1992-93 to 1994-95, Ph.D. completion rates vary only slightly by cohort size during the ten years from initial entry into doctoral study. Small, medium, and large cohorts all display a very similar pattern, particularly for years one through six. After year seven, large cohorts appear to have the highest average completion rate, while small cohorts have the lowest, but the differences between highest and lowest are minimal.

FigureFigure 4-16 4-16Average Average Cumulative Cumulative Ten-Year Ten-Year Completion Completion Rates by Rates Cohort by Size Cohort for StudentsSize forEntering Students Doctoral Entering Programs Doctoral from 1992-93 Programs through from 1994-95, 1992-93 by throughYear 1994-95, by Year 60 50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

Figure 4-17 below shows the differences in average completion rates by cohort size and broad field. In all broad fields, patterns of completion rates by cohort size are very similar. However, there are some small differences. Large cohorts tend to have slightly higher completion rates than medium and small cohorts in Social Sciences and Mathematics & Physical Sciences. Small cohorts have slightly higher completion rates in Engineering, while medium cohorts have slightly higher rates in the Humanities. Completion rates in the Life Sciences are almost identical for large and medium cohorts, both of which are slightly higher than for small cohorts after year six.

32 Figure 4-17 below shows the differences in average completion rates by cohort size and broad field. In all broad fields, patterns of completion rates by cohort size are very similar. However, there are some small differences. Large cohorts tend to have slightly higher completion rates than medium and small cohorts in Social Sciences and Mathematics & Physical Sciences. Small cohorts have slightly higher completion rates in Engineering, while medium cohorts have slightly higher rates in the Humanities. Completion rates in the Life Sciences are almost identical for large and medium cohorts, both of which are slightly higher than for small cohorts after year six.

FigureFigure 4-17 4-17 Average Average Cumulative Cumulative Ten-Year Ten-Year Completion Completion Rates byRates Cohort by CohortSize for DoctoralSize for Students Doctoral Entering Students from Entering 1992-93 throughfrom 1992-93 1994-95, through by Broad 1994-95, Fields and by Year Broad Field and Year

Engineering Life Sciences Mathematics & Physical Sciences 60 60 60 40 40 40

20 20 20

0 0 0 2 4 6 8 10 2 4 6 8 10 2 4 6 8 10

Social Sciences Humanities 60 60

40 40

20 20 Cumulative Completion Rate (%) Rate Completion Cumulative 0 0 2 4 6 8 10 2 4 6 8 10 Year

Small Medium

Large

Graphs by BF Source: Council of Graduate Schools Completion and Attrition Program Data

However, the relatively small or even unclear cohort-size effects shown in However,Figures the 4-16 relatively and 4-17 small might or even be unclear masked cohort-size by the substantialeffects shown differences in Figures 4-16 in anddisciplinary 4-17 might configurationbe masked by thebetween substantial the “small”, differences “medium”, in discip linaryand “large” configuration cohort betweengroups the within “small”, a broad “medium”, field. For and example,“large” cohort in the groups broad within field ofa broad Mathematics field. For & example,Physical in Sciences,the broad fieldcohort of Mathematicsize has an sopposite & Physical effect Sciences, for Chemistry cohort size than has it an has oppositefor Computer effect for & Chemistry Information than Sciences. it has for ComputerChemistry & has Information the highest Sciences. completion Chemistryrate in this has broad the highest field, completionwhile Computer rate in & this Information broad field, Sciences while Computer has the &lowest Information Sciences has the lowest completion rates after year five (see Figure 4-8). completion rates after year five (see Figure 4-8). Chemistry has a total of 1,384 Chemistry has a total of 1,384 students, while Computer & Information Sciences has 615 students. The cohort-size distribution for those two disciplines also differs: for Chemistry programs, the smallest one-third of cohorts range from 2 students to 15 students, medium cohorts range from 16 to 26 students, and the large cohorts range from 31 to 78 students;33 for Computer & Information Sciences, the small cohorts range from 4 to 10 students, medium cohorts range from 12 to 23 students, and large cohorts range from 25 to 55 students. Based on this specific-to-discipline definition for cohort sizes, instead of an overall definition for the broad field, the cohort-size effect becomes more pronounced and differs across disciplines. Figure 4-18 displays a clear positive correlation between cohort size and completion rates in Chemistry: the large cohorts have the highest completion rates, and small cohorts have the lowest completion rates after year six. However, for Computer & Information Sciences (Figure 4-19), large cohorts have the students, while Computer & Information Sciences has 615 students. The cohort- size distribution for those two disciplines also differs: for Chemistry programs, the smallest one-third of cohorts range from 2 students to 15 students, medium cohorts range from 16 to 26 students, and the large cohorts range from 31 to 78 students; for Computer & Information Sciences, the small cohorts range from 4 to 10 students, medium cohorts range from 12 to 23 students, and large cohorts range from 25 to 55 students. Based on this specific-to-discipline definition for cohort sizes, instead of an overall definition for the broad fields, the cohort- size effect becomes more pronounced and differs across disciplines. Figure 4-18 displays a clear positive correlation between cohort size and completion rates in Chemistry: the large cohorts have the highest completion rates, and small cohorts have the lowest completion rates after year six. However, for Computer & Information Sciences (Figure 4-19), large cohorts have the lowest completion rates, and the small cohorts have the highest completion rates, particularly for years four through nine.

FigureFigure 4-18 4-18 Average Average Cumulative Cumulative Ten-Year Ten-Year Completion Completion Rates for Chemistry Rates by for ChemistryCohort Size by for Cohort Doctoral Size Students for Doctoral Entering Studentsfrom 1992-93 Entering through from 1994-95 1992-93 and by throughYear 1994-95 and Year

50 40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

Note: Note: “Small”--“Small”-- cohortcohort size size rangi rangingng from from 2 to 15 2 studentsto 15 students “Medium”“Medium” --cohort size size ranging ranging from from16 to 2616 to 26 “Large”--“Large”-- cohortcohort size size ranging ranging from from 31 to 7831 to 78

34 FigureFigure 4-19 4-19 Average Average Cumulative Cumulative Ten-Year Ten-Year Completion Completion Rates Rates for for Computer Computer & Information& Information Sciences Sciences by Cohort by Cohort Size for Size Doctoral for Doctoral Students StudentsEntering fromEntering 1992-93 throughfrom 1992-93 1994-95 through and by Year1994-95 and Year

40 30 20 10 Cumulative Completion Rate (%) Rate Completion Cumulative 0

3 4 5 6 7 8 9 10 Year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

Note: Note:“Small”-- cohort size ranging from 4 to 10 students “Medium” “Small”-- --cohort cohort size size ranging ranging from from 12 to 4 23 to 10 students “Large”--“Medium” cohort --cohort size ranging size ranging from 25 tofrom 55 12 to 23 “Large”-- cohort size ranging from 25 to 55

Seven-YearSeven-Year CompletionCompletion Rates Rates

TheThe aboveabove comparisons comparisons are are all all based based on on cohorts cohorts of of students students that that entered entered doctoral doctoral programsprograms from from 1992-93 1992-93 to 1994-95.to 1994-95. Using Using these these cohorts cohorts makes makes possible possible the examination the ofexamination trends in doctoral of trends completion in doctoral for completionstudents who for could students have beenwho enrolledcould have in their been programsenrolled infor theiras many programs as ten years. for as To many investigate as ten whether years. studentsTo investigate who entered whether Ph.D. programsstudents whoafter 1994-95entered Ph.D.had different programs completion after 1994-95 rates and had patterns, different this completionsection compares seven-yearrates and patterns, completion this rates section between compares cohorts seven-year entering from completion 1992-93 torates 1994-95 between and cohortscohorts entering entering from from 1995-96 1992-93 to1997-98. to 1994-95 Students and whocohorts entered entering Ph.D. programsfrom 1995- from academic year 1992-93 through 1994-95 are designated as A-cohorts, and students who 96 to 1997-98. Students who entered Ph.D. programs from academic year began doctoral study from 1995-96 through 1997-98 are labeled as B-cohorts. A comparison1992-93 through of these 1994-95two cohort are groups designated makes it aspossible A-cohorts, to determine and students if seven-year who completionbegan doctoral rates studychanged from between 1995-96 the timethrough that 1997-98the first cohort are labeled of students as B-cohorts. began their doctoralA comparison programs of inthese 1992-93 two andcohort the timegroups that makesthe second it possible cohort began to determine in 1995-96. if

35 seven-year completion rates changed between the time that the first cohort of students began their doctoral programs in 1992-93 and the time that the second cohort began in 1995-96.

As Figure 4-20 shows, there are only minor differences in average seven-year As Figure 4-20 shows, there are only minor differences in average seven-year completion ratescompletion between A- rates and B-cohortsbetween A- in SEMand B-cohorts and SSH fields. in SEM In SEMand SSH fields, fields. B-cohorts In SEM have a slightlyfields, lower B-cohorts average have completion a slightly rate lower(50.6%) average than A-cohorts completion (51.9%); rate (50.6%) but in SSH than fields,A-cohorts B-cohorts (51.9%); have essentially but in SSH the same fields, average B-cohorts completion have rates,essentially 35.8% theversus same 35.7%.average completion rates, 35.8% versus 35.7%.

FigureFigure 4-20 4-20Comparison Comparison of Seven-Year of Seven-Year Cumulative Cumulative Completion Completion Rates between Rates A- andbetween B- Cohorts A- andat Year B- CohortsSeven, by at SEM Year vs. Seven, SSH Fieldsby SEM vs. SSH Fields

51.9 SEM Fields 50.6

35.8 SSH Fields 35.7

0 10 20 30 40 50 60 Cumulative Completion Rate (%)

A-cohorts B-cohorts

NOTE: A-cohorts include all students entering 1992-93 through 1994-95 B-cohorts include all students entering 1995-96 through 1997-98 Source: Council of Graduate Schools Completion and Attrition Program Data

There is relatively little difference in average seven-year completion rates Therebetween is relatively the A- little and difference B- cohorts in in averag termse seven-yearof the five individualcompletion broadrates between fields (see the A- and FigureB- cohorts 4-21). in termsOnly of in the Engineering five individual do thebroad two fields cohorts (see Figureshow a4-21). difference Only in in Engineeringseven-year do completionthe two cohorts rate showgreater a differencethan 1.2 percentage in seven-year points, completion but the difference rate greater thanis 1.2 still percentage relatively points, small but(56.8% the difference for A-cohorts is still and relatively 53.9% smallfor B-cohorts). (56.8% for A-cohorts and 53.9% for B-cohorts).

36 FigureFigure 4-21 4-21 Seven-Year Seven-Year Cumulative Cumulative Doctoral Doctoral Student Student Completion Completion Rates for Rates A- and B- Cohortsfor A- and at Year B- Cohorts Seven, by at Broad Year Seven,Fields by Broad Field

56.8 Engineering 53.9

53.7 Life Sciences 53.2

48.2 Mathematics & Physical Sciences 47.5

40.9 Social Sciences 40.2

29.3 Humanities 30.5

0 10 20 30 40 50 60 Cumulative Completion Rate (%)

A-cohorts B-cohorts

Attrition Rates Attrition Rates So far, we have described and compared completion rates by broad field,

So far,discipline, we have described institutional and type,compared and completion cohort size, rates but by webroad have field, yet discipline, to discuss institutionalthe students type, and who cohort did not size, complete but we have doctoral yet to discuss study—the the students patterns who of did student not completeattrition. doctoral Completion study—the and patterns attrition of student tell usattrition. different Completion things and about attrition doctoral tell us differentprograms. things Studentsabout doctoral who programs.do not complete Students doctoral who do notstudy complete within doctoral a designated study withintime a designatedframe either time have frame left either the program have left or the are program continuing or are to continuing pursue the to degree.pursue the degree.Analyzing Analyzing completion completion rate delineates rate delineates the magnitude the magnitude and andtiming timing of completion; of completion;studying studying attrition attrition allows allows for an for analysis an analysis of of the the magnitude magnitude and reasons reasons for for non-completion. Investigating attrition patterns also helps to enhance our understanding non-completion. Investigating attrition patterns also helps to enhance our of completion rates. understanding of completion rates. Overall cumulative attrition rates increase each year after the initial year of entering doctoralOverall study cumulative (see Figure attrition 4-22) .rates Most increase students each who year leave after their the doctoral initial year programs of entering appear to dodoctoral so relatively study early,(see Figure as attrition 4-22). rates Most incr studentsease sharply who leaveduring their the firstdoctoral four programsyears. In the firstappear year, to 6.6%do so of relatively Ph.D. students early, asleft attrition their programs; rates increase at year sharply two, the during cumulative the first attritionfour rateyears. more In the than first doubled. year, 6.6% At year of Ph.D. four, thestudents attrition left rate their grew programs; to 23.6%. at yearThereafter, two, the attritionthe cumulative rate tends attrition to stabilize. rate more than doubled. At year four, the attrition rate grew

to 23.6%. Thereafter, the attrition rate tends to stabilize.

37 FigureFigure 4-22 Cumulative 4-22 Cumulative Ten-Year Ten-Year Attrition Attrition Rates for Rates Doctoral for Student Doctoral Cohorts Student Who Began from 1992-93 through 1994-95, by Year Cohorts Who Began from 1992-93 through 1994-95, by Year

30.2 30.6 29.5 30 28.4 27 25.4 23.6

19.9 20

13.8 10

6.6 Cumulative Attrition Rate (%) Rate Attrition Cumulative 0 12345678910 Source: Council of Graduate Schools Completion and Attrition Program Data

AttritionAttrition Rates Rates by SEM by vs. SEM SSH vs. Fields SSH Fields Across fields there are noticeable differences in attrition trends. In SEM fields, Acrossattrition fields ratesthere areare highernoticeable than differences those in SSH in attrition in each trends.of the tenIn SEMyears fields, after students attrition ratesentered are higher their than doctoral those in programs SSH in each (as shown of the tenin Figure years after 4-23). students Cumulative entered attrition their doctoralrates programs for the first (as shownfour years in Figure increase 4-23). very Cumulative quickly inattrition SEM fieldsrates for but the thereafter first four yearsslow increase through very quicklyyear ten. in SEMIn SSH fields fields, but thereafter the average slow cumulative through year attrition ten. In SSH rate fields,increases the average more cumulative gradually attrition during ratethe firstincreases four moreyears gradually and thereafter during increases the first four at years and thereafter increases at a faster rate than in the SEM fields. a faster rate than in the SEM fields.

38 FigureFigure 4-23 4-23 Cumulative Cumulative Ten-Year Ten-Year Attrition Rates forfor DoctoralDoctoral StudentStudent CohortsCohorts EnteringEntering from from 1992-93 1992-93 through through 1994-95, 1994-95, for for SEM SEM vs. vs. SSH SSH Fields, Fields, by by Year Year

35 30 25 20 15 Cumulative Attrition Rate (%) Rate Attrition Cumulative 10 5

1 2 3 4 5 6 7 8 9 10 Year

SEM SSH

NOTE: 10-Year attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

AttritionAttrition Rates Rates by Broadby Broad Fields Field Cumulative attrition rates by broad field are shown in Figure 4-24. While Cumulativecumulative attrition overall rates attrition by broad rates field follow are ashown pattern in similarFigure 4-24. to that While of completion cumulative overallrates, theyattrition do notrates vary follow as amuch pattern across similar broad to that fields, of completion and the rank rates, order they acrossdo not vary asthe much broad across fields broad differs fields, from and that the ofrank the order completion across the rates. broad Cumulative fields differs attrition from that of therates completion are highest rates. in CumulativeMathematics attrition & Physical rates are Sciences. highest in By Mathematics year ten, 36.9% & Physical of Sciences. By year ten, 36.9% of students in this broad field have left their programs students in this broad field have left their programs without a doctoral degree. without a doctoral degree. Early attrition in Humanities does not grow as fast as other fields,Early butattrition it grows in Humanitiessteadily each does year notthrough grow year as fastten; asby otheryear seven, fields, it but has it the grows second higheststeadily rate each of cumulative year through attrition. year Engineeringten; by year ranks seven, third it hasafter the year second seven, highestfollowed veryrate closelyof cumulative by Social attrition. Sciences Engineering and Life Scien ranksces. thirdAt year after ten, year the attritionseven, followed rate of Social Sciencesvery closely catches by up Social with thatSciences of Engineering. and Life Sciences. At year ten, the attrition rate of Social Sciences catches up with that of Engineering. After year four, cumulative attrition rates in the Humanities increase much faster than in theAfter other year broad four, fields. cumulative The slope attrition of the curve rates for in Humanities the Humanities after year increase seven keepsmuch a strong momentum of increase, suggesting that its attrition rate may continue to increase afterfaster year than ten. in Social the other Sciences broad may fields. also experience The slope growing of the attrcurveition for after Humanities year ten, thoughafter year apparently seven not keeps as strongly a strong as the momentum Humanities. of Attrition increase, for suggesting Engineering that and its the Lifeattrition Sciences, rate mayconversely, continue appears to increase to have after reached year a ten.plateau Social at year Sciences eight, whichmay also suggests aexperience stabilized late growing attrition attrition rate. after year ten, though apparently not as strongly

39 as the Humanities. Attrition for Engineering and Life Sciences, conversely, appears to have reached a plateau at year eight, which suggests a stabilized late attrition rate.

FigureFigure 4-24 4-24 Cumulative Cumulative Ten-year Ten-year Attrition Attrition Rates for Rates Doctoral for DoctoralStudent Cohorts Student EnteringCohorts from Entering 1992-93 from through 1992-93 1994-95, through by Broad 1994-95, Field by and Broad Year Field and Year

40 35 30 25 20 15 10 Cumulative Attrition Rate (%) Rate Attrition Cumulative 5

1 2 3 4 5 6 7 8 9 10 Year

Engineering Life Sciences Mathematics & Physical Sciences Social Sciences Humanities

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

Attrition Rates by Discipline Attrition Rates by Disciplines Within broad fields, attrition rates vary considerably at the disciplinary level. As was the case with the completion analysis, the analysis of attrition patterns Within broad fields, attrition rates vary considerably at the disciplinary level. As was the will be presented only for disciplines with five or more programs participating case with the completion analysis, the analysis of attrition patterns will be presented only forin disciplines this study. with five or more programs participating in this study.

Figure 4-25 shows the variation in ten-year cumulative attrition trends in Engineering. All five Engineering disciplines start with similar attrition rates (3.5%-7.6%) at year one and begin to plateau after year four or five. However, attrition rates in years one to four increase with different magnitudes: Electrical & Electronic Engineering has the fastest growth of cumulative attrition rates during this period, followed by Chemical Engineering and Mechanical Engineering. At year ten the accumulative attrition rates for these three disciplines are 43.4%, 28.8 %, and 26.3%, respectively. Civil Engineering has the lowest cumulative attrition rate of 14.4% over the ten-year period, followed by Biomedical Engineering at 20.3%.

40 Figure 4-25 shows the variation in ten-year cumulative attrition trends in Engineering. All five Engineering disciplines start with similar attrition rates (3.5%-7.6%) at year one and begin to plateau after year four or five. However, attrition rates in years one to four increase with different magnitudes: Electrical & Electronic Engineering has the fastest growth of cumulative attrition rates during this period, followed by Chemical Engineering and Mechanical Engineering. At year ten the accumulative attrition rates for these three disciplines are 43.4%, 28.8 %, and 26.3%, respectively. Civil Engineering has the lowest cumulative attrition rate of 14.4% over the ten-year period, followed by Biomedical Engineering at 20.3%.

FigureFigure 4-25 4-25 Cumulative Cumulative Ten-Year Ten-Year Attrition Attrition Rates for Rates Doctoral for Doctoral Student Cohorts Student ThatCohorts Entered That Engineering Entered Disciplines Engineering from Disciplines 1992-93 through from 1994-95 1992-93 at Five through or More1994-95 Programs, at Five by or Year More Programs, by Year

45 40 35 30 25 20 15 10 5 Cumulative Attrition Rate (%) Rate Attrition Cumulative 0

1 2 3 4 5 6 7 8 9 10 Year

Biomedical Engineering Chemical Engineering Civil Engineering Electrical & Electronics Engineering Mechanical Engineering

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

In Life Sciences (shown in Figure 4-26), Biology has a much higher cumulative

attrition rate than the other four Life Science disciplines. Biology has the highest In the Life Sciences (shown in Figure 4-26), Biology has a much higher cumulative attritionattrition rate growth than the rate other in fouryears Life one Science through disciplines. four and Biology does not ha sseem the highest to plateau attrition growthuntil rateyear in nine. years However, one through the four other and four does disciplines not seem to all plateau reach until plateaus year nine. in their However,cumulative the other attrition four ratesdisciplines at year all seven, reach indicatingplateaus in stabilizedtheir cumulative attrition attrition after ratesthat at yeartime. seven, By indicatingyear ten, stabilizeda very clear attrition hierarchy after that of cumulativetime. By year attrition ten, a very rates clear occurs. hierarchyBiology of hascumulative the highest attrition attrition rates occurs. rate at Biology 38.5%, has followed the highest by attrition Neuroscience rate at 38.5%, followed by Neuroscience (22.2%), Microbiology & Immunology (20.3%), Molecular Cellular Biology (19.1%), and Genetics & Molecular Genetics (14.2%).

41 (22.2%), Microbiology & Immunology (20.3%), Molecular Cellular Biology (19.1%), and Genetics & Molecular Genetics (14.2%).

Figure 4-26 Cumulative Ten-Year Attrition Rates for Doctoral Student FigureCohorts 4-26 ThatCumulative Entered Ten-Year Life Science Attrition Disciplines Rates for Doctoral from 1992-93 Student Cohorts through That1994-95 Entered at FiveLife Scienceor More Disciplines Programs, from by Year1992-93 through 1994-95 at Five or More Programs, by Year

40 35 30 25 20 15 10 Cumulative Attrition Rate (%) Rate Attrition Cumulative 5

1 2 3 4 5 6 7 8 9 10 Year

Biology Genetics, Molecular Genetics Microbiology & Immunology Molecular & Cellular Biology Neuroscience

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

For Mathematics & Physical Sciences disciplines (see Figure 4-27), a clear Forhierarchy Mathematics of cumulative & Physical Sciencesattrition disciplinesrates is easily (see identifiedFigure 4-27), throughout a clear hierarchy the ten of cumulativeyears after attrition initial rates enrollment. is easily identifiedAt year ten, throughout Computer the &ten Information years after initial Sciences enrollment.has the highestAt year ten, attrition Computer rate & (51.4%), Informa tion followed Sciences by has Mathematics the highest attrition (39.9%), rate (51.4%),Physics followed & Astronomy by Mathematics (33.1%), (39.9%), and Chemistry Physics & Astronomy (31.7%). Mathematics (33.1%), and then and ChemistryChemistry (31.7%). appear Mathematics to have reached and Chemistry plateaus appeararound to year have nine, reached while plateaus the other around yeartwo nine, disciplines while the do other not twoappear disciplines to have do reached not appear plateaus to have around reached year plateaus ten. around year ten.

42 Figure 4-27 Cumulative Ten-Year Attrition Rates for Doctoral Student FigureCohorts 4-27 ThatCumulative Entered Ten-Year Mathematics Attrition & PhysicalRates for SciencesDoctoral DisciplinesStudent Cohorts from That Entered Mathematics & Physical Sciences Disciplines from 1992-93 through 1992-93 through 1994-95 at Five or More Programs, by Year 1994-95 at Five or More Programs, by Year

50 45 40 35 30 25 20 15 10 Cumulative Attrition Rate (%) Rate Attrition Cumulative 5

1 2 3 4 5 6 7 8 9 10 Year

Chemistry Computer & Information Sciences Mathematics Physics & Astronomy

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

As Figure 4-28 below indicates, Social Sciences disciplines have diverging cumulative attrition trends over the ten years after initial student enrollment. Although all disciplines have similar attrition rates at year one (ranging from 3.2% to 10.8%), Political Science has the largest growth in attrition and reaches 37.6% at year ten. Sociology and Anthropology & Archaeology have similar tenth-year attrition rates, 32.7% and 31.5% respectively, followed by Economics (28.4%), Psychology (23.0%), and Communications (12.8%). The attrition rate in Anthropology & Archaeology increases at a steady rate over the ten-year period after enrollment and is still rising at year ten.

43 As Figure 4-28 below indicates, Social Sciences disciplines have diverging cumulative attrition trends over the ten years after initial student enrollment. Although all disciplines have similar attrition rates at year one (ranging from 3.2% to 10.8%), Political Science has the largest growth in attrition and reaches 37.6% at year ten. Sociology and Anthropology & Archaeology have similar tenth-year attrition rates, 32.7% and 31.5% respectively, followed by Economics (28.4%), Psychology (23.0%), and Communications (12.8%). The attrition rate in Anthropology & Archaeology increases at a steady rate over the ten-year period after enrollment and is still rising at year ten.

Figure 4-28 Cumulative Ten-Year Attrition Rates for Doctoral Student Figure 4-28 Cumulative Ten-Year Attrition Rates for Doctoral Student Cohorts ThatCohorts Entered That Social Entered Sciences Social Disciplines Sciences from Disciplines 1992-93 through from 1992-93 1994-95 throughat Five or More1994-95 Programs, at Five by or Year More Programs, by Year

40 35 30 25 20 15 10 5 Cumulative Attrition Rate (%) Rate Attrition Cumulative 0

1 2 3 4 5 6 7 8 9 10 Year

Anthropology & Archaeology Communications Economics Political Science Psychology Sociology

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

Figure 4-29 illustrates increasing attrition rates for Humanities disciplines throughout the ten-year period after initial enrollment. Starting with attrition rates ranging from 4.6% to 8.2% at year one, the four Humanities disciplines have similar growth patterns. English Language & Literature has the lowest growth rate after year two and thus end up with the lowest attrition rate (26.7%) at year ten. Foreign Languages & Literature (36.9%) and History (35.5%) have the highest attrition rates among Humanities disciplines at year ten, while Philosophy has a ten-year rate of 31.8%. None of the four disciplines appears to have reached a plateau in attrition at year ten.

44 Figure 4-29 illustrates increasing attrition rates for Humanities disciplines throughout the ten-year period after initial enrollment. Starting with attrition rates ranging from 4.6% to 8.2% at year one, the four Humanities disciplines have similar growth patterns. English Language & Literature has the lowest growth rate after year two and thus ends up with the lowest attrition rate (26.7%) at year ten. Foreign Languages & Literature (36.9%) and History (35.5%) have the highest attrition rates among Humanities disciplines at year ten, while Philosophy has a ten-year rate of 31.8%. None of the four disciplines appears to have reached a plateau in attrition at year ten.

FigureFigure 4-29 4-29 Cumulative Cumulative Ten-Year Ten-Year Attrition Attrition Rates Rates for Doctoral for Doctoral Student Student Cohorts ThatCohorts Entered That Humanities Entered Disciplines Humanities from Disciplines 1992-93 through from 1994-95, 1992-93 at through Five or More1994-95, Programs, at Five by or Year More Programs, by Year

40 35 30 25 20 15 10 5 Cumulative Attrition Rate (%) Rate Attrition Cumulative 0

1 2 3 4 5 6 7 8 9 10 Year

English Language & Literature Foreign Languages & Literature History Philosophy

NOTE: 10-Year Attrition Rates include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

Attrition Rates by Institution Types

As can be seen in Figure 4-30, from years one through seven, attrition rates at public institutions are slightly higher than those at private institutions. However, the rates are essentially the same, particularly from years eight through ten.

45 Attrition Rates by Institution Types As can be seen in Figure 4-30, from years one through seven, attrition rates at public institutions are slightly higher than those at private institutions. However, the rates are essentially the same, particularly from years eight through ten.

FigureFigure 4-30 4-30Cumulative Cumulative Ten-Year Ten-Year Attrition Attrition Rates for Rates Doctoral for DoctoralStudent Cohorts Student fromCohorts 1992-93 from through 1992-93 1994-95, through by Institution 1994-95, Type by Institution and Year Type and Year

35 30 25 20 15 10 Cumulative Attrition Rate (%) Rate Attrition Cumulative 5 0

1 2 3 4 5 6 7 8 9 10 Year

Private Public

Source: Council of Graduate Schools Completion and Attrition Program Data

46 Figure 4-31, which shows the trends in attrition rates by institutional type and broad field, displays a very different pattern from that at the aggregate level (see Figure 4-30). Differences in early attrition appear to be small between the two types of institutions in most broad fields. In the later years, attrition rates in Engineering, Life Sciences, Social Sciences and Humanities are higher in private institutions, while in Mathematical & Physical Sciences, public institutions have higher attrition rates, particularly in the later years.

Figure 4-31 Cumulative Ten-Year Attrition Rates for Doctoral Student FigureCohorts 4-31 Cumulativefrom 1992-93 Ten-Year through Attrition 1994-95, Rates by Institutionfor Doctoral Type, Student Broad Cohorts Field fromand 1992-93 Year through 1994-95, by Institution Type, Broad Field and Year

Engineering Life Sciences Mathematics & Physical Sciences 30 30 40 30 20 20 2 10 10 0 1 0 0 0 0 0 5 10 0 5 10 0 5 10

Social Sciences Humanities 30 30 20 20

Cumulative Attrition Rate (%) Rate Attrition Cumulative 10 10

0 0 0 5 10 0 5 10 Year

Private Public

Graphs by BF Source: Council of Graduate Schools Completion and Attrition Program Data

Attrition Rates by Cohort Sizes

As Figure 4-32 indicates, attrition rates among doctoral students do not show much variation as a function of cohort size. Large cohorts have the highest average attrition rates after year three, and medium cohorts have the lowest attrition rates after year seven.

47 Attrition Rates by Cohort Size As Figure 4-32 indicates, attrition rates among doctoral students do not show much variation as a function of cohort size. Large cohorts have the highest average attrition rates after year three, and medium cohorts have the lowest attrition rates after year seven.

FigureFigure 4-32 4-32 Cumulative Cumulative Ten-Year Ten-Year Attrition Attrition Rates for Rates Doctoral for DoctoralStudent Cohorts Student fromCohorts 1992-93 from through 1992-93 1994-95, through by Cohort 1994-95, Size by and Cohort Year Size and Year

30 (%) Rate 20 10 Cumulative Attrition 0

1 2 3 4 5 6 7 8 9 10 Year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

Across broad fields, the previously described relationship between cohort size Acrossand broadattrition fields, rates the does previously not always described apply relationship (see Figure between 4-33). cohort Attrition size andrates in attritionEngineering rates does clearly not always increase apply with (see cohort Figure size:4-33). large Attrition cohorts rates have in Engineering the highest clearlyattrition increase rates, with and cohort small size: cohorts large have cohorts the havelowest the across highest all attrition years. rates,However, and small for cohortsother have SEM the fields, lowest this across trend all does years. not However, hold. In forLife other Sciences, SEM fields, large thiscohorts trend have does not hold. In Life Sciences, large cohorts have the highest attrition rates after year three, the highest attrition rates after year three, while small and medium cohorts while small and medium cohorts have very similar rates through year seven, and medium cohortshave have very lower similar rates rates in years through eight year through seven, ten. andMathem mediumatics &cohorts Physical have Sciences’ lower largerates and in small years cohorts eight havethrough similar ten. attrition Mathematics trends, while& Physical medium Sciences’ cohorts have large slightly and lowersmall attrition cohorts rates have after similar year three. attrition Among trends, SSH while fields medium, the Social cohorts Sciences have field slightly displayslower opposite attrition trends rates fromafter Engineering, year three. Amongwith small SSH cohor fields,ts having the Socialthe highest Sciences attrition ratesfield and displays large cohorts opposite the lowest. trends Humanitiesfrom Engineering, do not show with as small clear cohorts a cohort-size having effect the as the Social Sciences. The three types of cohorts in Humanities have very similar attrition rates, though smaller cohorts tend to have minimally higher rates after year six.

48 highest attrition rates and large cohorts the lowest. Humanities do not show as clear a cohort-size effect as the Social Sciences. The three types of cohorts in Humanities have very similar attrition rates, though smaller cohorts tend to have minimally higher rates after year six.

FigureFigure 4-33 4-33 Cumulative Cumulative Ten-Year Ten-Year Attrition Attrition Rates for Rates Doctoral for DoctoralCohorts from Cohorts 1992- 93from through 1992-93 1994-95, through by Cohort 1994-95, Sizes, by Broad Cohort Field, Size, and BroadYear Field, and Year

Engineering Life Sciences Mathematics & Physical Sciences 30 30 40

20 20 30

10 10 2 0 1 0 0 0 0 5 10 0 5 10 0 5 10

Social Sciences Humanities 40 40 30 30 20 20

Cumulative Attrition Rate (%) Rate Attrition Cumulative 10 10 0 0 0 5 10 0 5 10 Year

Small Medium

Large

Graphs by BF Source: Council of Graduate Schools Completion and Attrition Program Data

However, the above-mentioned cohort-size effect could actually be a compound However,effect of the cohort above-mentioned size and disciplines. cohort-size For effect example, could actually although be Figurea compound 4-33 effectshows of cohortclear size hierarchical and disciplines. attrition For ratesexample, as aalthough function Figure of cohort 4-33 shows sizes clearin Engineering, hierarchical attritionthe high rates attrition as a function rates ofin cohort the large sizes Engineering in Engineering, cohorts the high could attrition be rateslargely in the the largeresult Engineering of high attritioncohorts could rates be in largely Electrical the result & Electronicsof high attrition Engineering, rates in Electrical which & Electronics Engineering, which has high student counts in its large cohorts. More details ofhas these high findings student are countspresented in inits Appendix large cohorts. E. More details of these findings are presented in Appendix E.

Early Attrition

As indicated previously, attrition rates increase sharply during the first four years after initial enrollment. The following two sections therefore focus on early attrition rates, i.e. attrition rates in the first four years. Data are presented for three cohorts: the A- and B- cohorts described earlier, and C-cohorts, which include students who entered doctoral programs from academic year 1998-99 through 2000-01.

From an examination of attrition rates across the three cohort groups, it is obvious that C- cohorts consistently have a lower cumulative attrition rate than A- and B-cohorts at year49 four. Figure 4-34 shows the overall fourth-year cumulative attrition rates for all three cohorts. A- and B-cohorts have very similar attrition rates (23.6% and 24.2% Early Attrition As indicated previously, attrition rates increase sharply during the first four years after initial enrollment. The following two sections therefore focus on early attrition rates, i.e. attrition rates in the first four years. Data are presented for three cohorts: the A- and B-cohorts described earlier, and C-cohorts, which include students who entered doctoral programs from academic year 1998-99 through 2000-01.

From an examination of attrition rates across the three cohort groups, it is obvious that C-cohorts consistently have a lower cumulative attrition rate than A- and B-cohorts at year four. Figure 4-34 shows the overall fourth- year cumulative attrition rates for all three cohorts. A- and B-cohorts have very similar attrition rates (23.6% and 24.2% respectively) at year four, but respectively)C-cohorts have at year a substantially four, but C-cohorts lower haveattrition a substantially rate (20.2%). lower In attritionboth SEM rate and(20.2%). InSSH both fields, SEM C-cohortsand SSH fields, also haveC-cohorts lower also four-year have a lower attrition four-year rates attritionthan the rate other than the other two cohorts, which have similar attrition rates (see Figure 4-35). two cohorts, which have similar attrition rates (see Figure 4-35).

Figure 4-344-34 Overall Overall Four-Year Four-Year Cumulative Cumulative Attrition Attrition Rates Rates for A-, for B- A-, and B- C-and CohortsC-Cohorts

23.6

24.2

20.2

0 2 4 6 8 10 12 14 16 18 20 22 24 Attrition Rate (%)

A B C

NOTE: A-cohorts include all students entering 1992-93 through 1994-95 B-cohorts include all students entering 1995-96 through 1997-98 C-cohorts include all cohorts entering 1998-99 through 2000-01 Source: Council of Graduate Schools Completion and Attrition Program Data

50 FigureFigure 4-35 4-35 Four-Year Four-Year Cumulative Cumulative Attrition Attrition Rates for Rates A-, B- forand C-Cohorts A-, B- and by SEM/SSHC-Cohorts, by SEM/SSH Fields

26.5

SEM Fields 26.6

22.0

19.3

SSH Fields 20.2

16.6

0 2 4 6 8 10 12 14 16 18 20 22 24 Attrition Rate (%)

A B C

AtAt the the broad broad field field level level (see (see Figure Figure 4-36) 4-36) there thereis an evenis an greater even greaterdifference difference in four-year attritionin four-year rates betweenattrition A- rates and B-between cohorts A- for andindividual B- cohorts fields whenfor individual compared withfields overall, SEM, and SSH attrition rates. However, in all broad fields the four-year attrition rateswhen are compared higher for with A- and overall, B-cohorts SEM, than and for SSHC-cohorts. attrition Except rates. for However, the Humanities, in all C- cohortsbroad fieldshave substantiallythe four-year lowe attritionr attrition rates rates are than higher A- a ndfor B-cohorts. A- and B-cohorts These differences than arefor moreC-cohorts. pronounced Except in Lifefor theSciences, Humanities, Mathematics C-cohorts & Physical have substantiallySciences, and lowerSocial Sciencesattrition thanrates in than Engineering A- and B-cohorts.and Humanities. These differences are more pronounced in Life Sciences, Mathematics & Physical Sciences, and Social Sciences than in Engineering and Humanities.

51 FigureFigure 4-36 4-36 Four-Year Four-Year Cumulative Cumulative Attrition Attrition Rates for Rates A-, B- forand A-,C-Cohorts B- and by BroadC-Cohorts, Fields by Broad Fields

23.2 Engineering 23.0 21.4 20.9 Life Sciences 22.0 17.3 30.7 Mathematics & Physical Sciences 30.8 24.7 20.0 Social Sciences 19.5 15.3 18.5 Humanities 21.1 18.2

0 4 8 12 16 20 24 28 Attrition Rate (%)

A B C

TheThe lower lower four-year four-year at attritiontrition rate rate in inC-cohorts C-cohorts indicates indicates that that a lower a lower percentage percentage of studentsof students who whobegan began doctoral doctoral study instudy 1998-99 in 1998-99 through 2000-01through left2000-01 their programs left their in the first four years than of students who began study in 1992-93 through 1997-98. Future editionsprograms of thein the Ph.D. first Completion four years and than Attrition of students series whowill examinebegan study possible in 1992-93 reasons for thethrough lower 1997-98. level of attrition Future amongeditions students of the inPh.D. this laterCompletion cohort group, and Attrition C-cohorts. series will examine possible reasons for the lower level of attrition among students in the C-cohorts.

Components of Early Attrition Components of Early Attrition This section further analyzes the categories of students who contribute to early Thisattrition section during further the analyzes first four the years categories of enrollment. of students Earlywho contribute attrition includesto early attrition the duringfollowing the firstcategories four years of students:of enrollment. Early attrition includes the following categories of students: x• StudentsStudents who who left left their their institution institution with with no master’s no master’s (No Master’s (No Master’s) ) x• StudentsStudents who who left leftwith witha master’s a master’s degree degreebut had not but reached had not the reached candidacy the stage (Master’s,candidacy No stage Candidacy (Master’s,) No Candidacy) x• StudentsStudents who who left leftafter afterreceiving receiving a master’s a master’s degree and degree achieving and achievingcandidacy (Master’s,candidacy With (Master’s, Candidacy With) Candidacy) x Students who transferred either to another doctoral program at their original • Students who transferred either to another doctoral program at their institution or to another institution (Transferred) original institution or to another institution (Transferred)

52 Figure 4-37 shows the student components of overall early attrition. The majority of early attrition involves students who left their doctoral programs without achieving candidacy (including No Master’s and Master’s, No Candidacy), while the Transferred category accounts for only a very small proportion (less than 5%) of total attrition in each of the first four years after enrollment. At year one, approximately 6% of all entering Ph.D. students and over 90% of all students leaving their programs left without a master’s degree. At year two, Figurealmost 4-37 half shows of the those student leaving components their programs of overall left early without attrition. a The master’s majority degree. of early attritionThereafter, involves students students leaving who left with their a doctoralmaster’s programs (either with without or without achieving candidacy) candidacy (including No Master’s and Master’s, No Candidacy), while the Transferred category account for a growing share of attrition. Attrition due to Transferred students is accounts for only a very small proportion (less than 5%) of total attrition in each of the firsthighest four years in years after enrollment.two and three. At year The one,portion 6% of allearly Ph.D. attrition students by andstudents over 90%with of a all studentsmaster’s leaving degree their (Master’s, programs No left Candidacy without a andmaster’s Master’s, degree. With At yearCandidacy two, almost) and halfwith of those candidacy leaving is their highest programs in the left third without and fourtha master’s years. degree. Thereafter, students leaving with a master’s (either with or without candidacy) account for a growing share of attrition.Figure Attrition 4-37 Components due to Transferred of Overall students Early is highest Student in years Attrition two and for three. Years The One portionthrough of early Four attrition by students with Master’s degree (Master’s, No Candidacy and Master’s, With Candidacy) and with Candidacy is highest in the third and fourth years.

The next several figures illustrate the components of early attrition for each of

Figurethe five4-37 broadComponents fields. Inof OverallEngineering Early (see Student Figure Attrition 4-38), foronly Years a very One minimal through Fourshare of attrition occurs among students who have advanced to candidacy. There is a slightly higher portion of attrition by Engineering students who left 8 6 4 2 Percent Attrition (%) Attrition Percent 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4 No Master's Master's, No Candidacy Master's, With Candidacy Transferred

Source: Council of Graduate Schools Completion and Attrition Program Data NOTE: A-cohorts include all cohorts entering 1992-93 through 1994-95 B-cohorts include all cohorts entering 1995-96 through 1997-98 C-cohorts include all cohorts entering 1998-99 through 2000-01

The next several figures illustrate the components of early attrition for each of the five broad fields. In Engineering (see Figure 4-38), only a very minimal share of attrition53 occurs among students who have advanced to candidacy. There is a slightly higher portion of attrition by Engineering students who left their programs with master’s degrees when compared with overall attrition pattern across all fields (compared to Figure 4-37). A possible explanation for this finding is that Engineering students had more their programs with master’s degrees when compared with overall attrition pattern across all fields (compared to Figure 4-37). A possible explanation for this finding is that Engineering students had more opportunities for employment in highly paid positions upon completion of the master’s degree than students in other broad fields.

FigureFigure 4-38 4-38 Components Components of Early of EarlyStudent Student Attrition Attrition in Engineering in Engineering for Years One for throughYears FourOne through Four

10 8 6 4 Percent Attrition (%) Attrition Percent 2 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4

No Master's Master's, No Candidacy Master's, With Candidacy Transferred

In Life Sciences (Figure 4-39), a dominant share of attrition occurs among In doctoralLife Sciences students (Figure without 4-39), master’sa dominant degrees share ofat attrition year one occurs and amongto a much doctoral lesser studentsextent without thereafter; master’s attrition degrees of studentsat year one with and master’s to a much degrees lesser extent (with thereafter; or without attritioncandidacy) of students accounts with formaster’s the largest degrees portion (with orof without attrition candidacy) in years threeaccounts and for four. the largestAttrition portion of studentsof attrition with thereafter. candidacy Attri occurstion of studentsmore often with in candidacy later years occurs and moremore oftenfrequently in later yearsin Life and Sciencesmore frequently than in in mostthe Life other Sciences fields. than Transfer in most accounts other fields. for Transfera small accounts proportion for a ofsmall early proportion attrition, of fromearly attrition,year two but through more transfer year four appears but toa occur in year one. majority transfer appears to occur in year one.

54 FigureFigure 4-39 4-39 Components Components of Ea ofrly Early Attrition Attrition in Life in SciencesLife Sciences for Years for Years One through One Fourthrough Four

8 6 4 2 Percent Attrition (%) Attrition Percent 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4

No Master's Master's, No Candidacy Master's, With Candidacy Transferred

Attrition of students with candidacy seems to be more pronounced in Mathematics & Physical Sciences (see Figure 4-40) and to occur earlier (starting even in year one) than in Engineering and Life Sciences. Attrition due to transfer occurs most often in years two and three.

55 Attrition of students with candidacy seems to be more pronounced in Mathematics & Physical Sciences (see Figure 4-40) and to occur earlier (starting even in year one) than in Engineering and Life Sciences. Attrition due to transfer occurs most often, particularly in year two.

FigureFigure 4-40 4-40 Components Components of Ea rlyof EarlyAttrition Attrition in Mathematics in Mathematics & Physical & Sciences Physical for YearsSciences One through for Years Four One through Four

10 8 6 4 2 Percent Attrition (%) Attrition Percent 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4

No Master's Master's, No Candidacy Master's, With Candidacy Transferred

In the Social Sciences (Figure 4-41), except for year four, early attrition of students with master’s degree tends to be similar for all three cohorts. Some attrition due to transfer occurs in all four years, but to a lesser extent at year four.

56 In the Social Sciences (Figure 4-41), except for year four, early attrition of students with master’s degree tends to be similar for all three cohorts. Some attrition due to transfer occurs in all four years, but to a lesser extent at year four.

FigureFigure 4-41 4-41Components Components of Early of AttritionEarly Attrition in Social in Scie Socialnces forSciences Years Onefor Yearsthrough FourOne through Four

6 4 2 Percent Attrition (%) Attrition Percent 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4

No Master's Master's, No Candidacy Master's, With Candidacy Transferred

As Figure 4-42 shows, attrition of doctoral students who have advanced to candidacy increases each year in Humanities. Attrition due to transfer appears to occur less frequently in Humanities than in other broad fields.

57 As Figure 4-42 shows, attrition of doctoral students who have advanced to candidacy increases each year in Humanities. Attrition due to transfer appears to occur less frequently in Humanities than in other broad fields.

FigureFigure 4-42 4-42Components Components of Early of EarlyAttrition Attrition in Humanities in Humanities for Years for One Years through One Fourthrough Four

8 6 4 2 Percent Attrition (%) Attrition Percent 0 ABC ABC ABC A BC Year 1 Year 2 Year 3 Year 4

No Master's Master's, No Candidacy Master's, With Candidacy Transferred

It can be observed in Figures 4-37 to 4-42 that, with a few exceptions, early It canattrition be observed rates infor Figures C-cohorts 4-37 toare 4-42 lower that, than with for a few the exceptions, other two earlycohort attrition groups. rates for C-cohortsThis finding are loweris consistent than for thewith other the two findings cohort fromgroups. the This previous finding sectionis consistent (see withFigure the findings 4-34 tofrom 4-36). the previous section (see Figure 4-34 to 4-36).

Continuing,Continuing, Completion, Completion, and and Attrition Attrition RatesRates In the previous sections of this chapter, cumulative completion and attrition In thedata previous and analyses sections wereof this presented chapter, cumulative for the first completion ten, seven, and and attrition four yearsdata and (A-, analysesB-, and were C-cohorts, presented respectively) for the first ten, after seven, students and four entered years their (A-, doctoralB-, and C-cohorts, programs. respectively)However, after there students has been entered no mention their doctoral of the programs.students who However, were continuing there has been their no mentiondoctoral of the programs students whoat the were end continuingof the same their time doctoral periods. programs This section at the focusesend of the on samethe time rates periods. of continuing This section students. focuse s on the rates of continuing students.

Figure 4-43 shows the percentage of A-cohort students continuing each year from the third through the tenth year after entering doctoral programs, by broad field, along with the percentage who completed and the percentage who did not complete. As might be expected, Engineering, which has the highest completion rate, also has a low percentage of the58 entering cohort continuing at year ten (9.7%). Mathematics & Physical Sciences, which has the highest attrition rate and the fourth highest completion rate, has the lowest percentage of students continuing at year ten (8.3%). It is also not surprising that Figure 4-43 shows the percentage of A-cohort students continuing each year from the third through the tenth year after entering doctoral programs, by broad field, along with the percentage who completed and the percentage who did not complete. As might be expected, Engineering, which has the highest completion rate, also has a low percentage of the entering cohort continuing at year ten (9.7%). Mathematics & Physical Sciences, which has the highest attrition rate, has the lowest percentage of students continuing at year ten (8.3%). It is also not surprising that Humanities has the largest proportion of students continuing at year ten, because it has the lowest share of students who have completed their degrees at year ten. In general, the SSH broad fields have Humanitieslarger portions has the of enteringlargest proportion students ofwho students are continuing continuing at at year year ten ten, than because the SEM it has thebroad lowest fields. share The of studentsrate of decrease who have in completed continuing their students degrees atat yearyear ten.ten isThe greatest SSH broadfor Humanities fields have largeramong portions the other of entering broad fields.students At who year are ten, continuing the percentage at year ten of than thecontinuing SEM broad students fields. Thein the rate SEM of decrease fields instarts continuing to approach students plateaus, at year ten while is greater the for Humanities than any of the other broad fields. At year ten, the percentage of continuing studentspercentage in the of SEM continuing fields starts students to approach in the plateaus, SSH fields while is the still percentage decreasing. of continuing This studentspartially in explains the SSH thefields differences is still decreasing. in completion This partially rates betweenexplains theSEM differences fields and in completionSSH fields rates that betweenwas described SEM fields earlier. and SSH fields that was described earlier.

Figure 4-43 Cumulative Continuing, Completion, and Attrition Rates Figurefor Doctoral 4-43 Cumulative Students Continuing,Entering from Completion, 1992-93 and through Attrition 1994-95, Rates forby DoctoralBroad StudentsField and Entering Year from 1992-93 through 1994-95, by Broad Fields and Year

100 80 60 40 Cumulative Rate (%) Rate Cumulative 20 0 345678910 345678910 345678910 345678910 345678910 Engineering Life Sciences Math & Physical Sci. Social Sciences Humanities

Continuing Completion Attrition

Note: 10-Year Cohorts include all cohorts entering 1992-93 through 1994-95 Source: Council of Graduate Schools Completion and Attrition Program Data

Note that the proportion of continuing Ph.D. students at year ten in SSH broad fields almost doubles that in Engineering and Mathematics & Physical Sciences. If 50% of59 those continuing students eventually complete the Ph.D., the completion rates in SSH fields could be equal to or greater than those in Engineering and Mathematics & Physical Sciences.

The remainder of this section focuses on comparing the rates of completion, attrition, and continuing students over the time of enrollment, i.e., for A-, B-, and C-cohorts. As indicated earlier, C-cohorts have a lower rate of early attrition when compared with A- Note that the proportion of continuing Ph.D. students at year ten in SSH broad fields almost doubles that in Engineering and Mathematics & Physical Sciences. If 50% of those continuing students eventually complete the Ph.D., the completion rates in SSH fields could be equal to or greater than those in Engineering and Mathematics & Physical Sciences.

The remainder of this section focuses on comparing the rates of completion, attrition, and continuing students over the time of enrollment, i.e., for A-, B-, and C-cohorts. As indicated earlier, C-cohorts have a lower rate of early attrition when compared with A- and B-cohorts (see Figure 4-34). As shown in Figure 4-44, C-cohorts also have a lower four-year completion rate than A- and B-cohorts. Therefore, the percentage of students who are continuing their Ph.D. programs after four years is highest for C-cohorts (see Figure 4-44). As can be seen in Figure 4-45, in all five broad fields, except Humanities8, the C-cohorts have the lowest attrition rates, the lowest completion rates, and the highest percentages of continuing students after four years of enrollment in Ph.D. programs.

8 This trend also basically holds for Humanities, except that the completion rate of C-cohorts is slightly higher than B-cohorts for Humanities.

60 and B-cohorts (see Figure 4-34). As shown in Figure 4-44, C-cohorts also have a lower four-year completion rate than A- and B-cohorts. Therefore, the percentage of students who are continuing their Ph.D. programs after four years is highest for C-cohorts (see Figure 4-44). As can be seen in Figure 4-45, in all five broad fields, except Humanities1, the C-cohorts have the lowest attrition rates, the lowest completion rates, and the highest percentages of continuing students after four years of enrollment in Ph.D. programs. Figure 4-44 Overall Four-Year Cumulative Continuing, Completion, andFigure Attrition 4-44 Overall Rates Four-Year for A-, Cumulative B- and C-Cohorts Continuing, Completion, and Attrition Rates for A-, B- and C-Cohorts

A 65.9 10.5 23.6

B 67.5 8.3 24.2

C 72.7 7.1 20.2

0 20 40 60 80 100 Rate (%)

Continuing Completion Attrition

Source: Council of Graduate Schools Completion and Attrition Program Data

Figure 4-45 Four-Year Cumulative Continuing, Completion, and Attrition Rates, Figurefor A-, 4-45B- and Four-Year C- Cohorts, Cumulative by Broad Field Continuing, Completion, and Attrition Rates, for A-, B- and C- Cohorts, by Broad Field

A 59.9 17.1 23.1 Engineering B 61.9 15.1 23 C 67.5 11.3 21.3

A 69.6 9.4 21 Life Sciences B 71.3 6.7 22 C 77.1 5.6 17.3

A 60.3 8.9 30.8 Math & Physical Sci. B 60.3 8.9 30.8 C 68.1 7.3 24.6

1 This trend also basicallyA holds for Humanities, except68.5 that the completion rate11.5 of C-cohorts 20 is slightly higherSocial than B-cohortsSciences Bfor Humanities. 73.5 7 19.5 C 78.1 6.7 15.3

A 75.5 6.1 18.4 Humanities B 75.8 3.1 21.1 C 78.5 3.3 18.2 0 20 40 60 80 100 Rate (%)

Continuing Completion Attrition

Source: Council of Graduate Schools Completion and Attrition Program Data

61 As can be seen in Figure 4-46, this trend in declining completion and attrition rates and increasing portions of continuing students in the first four years of enrollment is basically prevalent over the nine entering cohort years, from 1992-93 through 2000-01. Of particular note is the fact that these trends are prevalent, without exception, during the last three years (i.e., C-cohort: years 1998-99, 1999-2000, and 2000-01).

Figure 4-46 Overall Four-Year Cumulative Continuing, Completion, and Attrition Rates by Cohort Year

As can be seen in Figure 4-46, this trend in declining completion and attrition rates and increasing portions of continuing students in the first four years of enrollment is basically prevalent over the nine entering cohort years, from 1992-93 through 2000-01. Of particular note is the fact that these trends are prevalent, without exception, during the last three years (i.e., C-cohort: years 1998-99, 1999-2000, and 2000-01).

Figure 4-46 Overall Four-Year Cumulative Continuing, Completion, and Attrition Rates by Cohort Year

1992-93 63.8 13.4 22.8

1993-94 67.5 9 23.5

1994-95 66.5 8.8 24.6

1995-96 67.3 8.2 24.4

1996-97 66.6 8.1 25.2

1997-98 68.5 8.6 22.9

1998-99 70.1 8.1 21.9

1999-2000 73 6.8 20.2

2000-01 75 6.5 18.5

0 20 40 60 80 100 Rate (%)

Continuing Completion Attrition

Source: Council of Graduate Schools Completion and Attrition Program Data

There are several possible explanations for the decreasing four-year completion Thereand attrition are several rates possible from theexplanations earlier to forlater the cohorts decreasing (i.e., four-year A-, B- and completion C-cohorts). and attritionLower ratescompletion from the rates earlier may to belater related cohorts to (i.e.,the relaxation A-, B- and of C-cohorts). requirements Lower for a completionmaster’s degree rates may or prerequisite be related to coursesthe relaxati for onPh.D. of requirements admission that for ahad master’s persisted degree or prerequisiteuntil the early courses 1990s. for Ph.D. Without admission the strict that requirementshad persisted until for athe master’s early 1990s. degree Without the strict requirements for a master’s degree or equivalent courses taken, it may take additionalor equivalent time forcourses students taken, to finish it may the take prerequisite additional courses time andfor thenstudents their toPh.D. finish degrees.the prerequisite This explanation courses canand be then particularly their Ph.D. supported degrees. by This the fact, explanation shown in can this bestudy, thatparticularly the first entering supported cohort by (studentsthe fact, whoshown entered in this their study, programs that thein 1992-93) first entering has a substantiallycohort (students higher who four-year entered completion their programs rate than in the 1992-93) later eight has entering a substantially cohorts (studentshigher four-year who were completion enrolled in yearsrate than 1993-94 the later through eight 2000-01). entering Ascohorts shown (students in Figure 4- 46,who the were cohort enrolled entering in in years1992-93 1993-94 has a completion through 2000-01). rate as high As as shown 13.4%, in while Figure students in the later eight cohort years have incrementally decreasing completion rates ranging from 9.0% to 6.5%. 62 The lower attrition rates for the C-cohorts might also be related to the improvements in financial support for students in the mid to late 1990s. Lower attrition rates could also be a result of more qualified students being admitted. However, these suggested causes are only hypotheses at this time. They will be explored in more detail later in the Ph.D. Completion Project. 4-46, the cohort entering in 1992-93 has a completion rate as high as 13.4%, while students in the later eight cohort years have incrementally decreasing completion rates ranging from 9.0% to 6.5%.

The lower attrition rates for the C-cohorts might also be related to the improvements in financial support for students in the mid to late 1990s. Lower attrition rates could also be a result of more qualified students being admitted. However, these suggested causes are only hypotheses at this time. They will be explored in more detail later in the Ph.D. Completion Project.

Conclusions The baseline completion and attrition data that were collected in this project allow for a thorough examination of completion and attrition patterns across broad fields, academic disciplines, student cohort sizes, and institutional types. The findings presented here suggest some potential improvements in Ph.D. completion patterns between the earliest and latest cohorts, given the lower early attrition rates in the latest cohort (C-cohorts). Even so, the attrition rates in a number of broad fields remain troublesome and deserve additional study.

At the aggregate level, the data show that nearly 57% of the doctoral candidates at the participating institutions completed their degree programs within a ten-year time span. However, Ph.D. completion rates do vary by broad field. Ten-year completion rates range from about 63% for Engineering and Life Sciences to approximately 49% for Humanities. The average completion rate in all SEM fields combined is noticeably higher than that of Social Sciences and Humanities fields, but the SSH combined completion rate appears to keep increasing even after the ten-year mark. This finding suggests that a number of students in these broad fields will continue to earn their degrees after ten years and that the differences in ultimate completion rates between the broad fields may diminish.

The Completion Project data also show that the cumulative Ph.D. completion trend in the five broad fields varies with time within ten years after enrollment. At the seven-year mark, more than half the students in Engineering and Life Sciences have earned doctorates, compared with just 29% of those in Humanities. But it should be kept in mind that 20% of doctoral students in Humanities complete their degrees between years seven and ten. Many of the students finishing after the seventh year are in SSH fields, which may indicate differences in program or field characteristics. Later publications in the Ph.D. Completion and Attrition series will examine these differences in more detail.

63 Another interesting finding of the Completion Project is that the Ph.D. completion rates at public universities do not differ much from those at private institutions. Public institutions seem to have slightly higher early completion rates (before year five), while private institutions have slightly higher rates thereafter. This trend, however, varies across fields. The data also show that cohort size generally does not affect completion rates, with large cohorts having slightly higher rates of completion than smaller ones in some fields but lower rates than smaller cohorts in other fields. Attrition rates of SEM fields are clearly higher than those of SSH for all ten years after initial student entry into doctoral programs. Cumulative attrition rates in SEM fields increase quickly during the first four years after initial cohort entry but then decline. In SSH fields, the cumulative attrition rates increase more gradually. By broad field, cumulative attrition rates are highest in Mathematics & Physical Sciences but increase more consistently in Humanities. Attrition rates vary more at the program level within each broad field.

Attrition rates do not show much variation across cohort size and institution type for the combined fields. In some broad fields, cohort size and institution type appear to have a relationship to attrition rates. However, the cohort-size effect may be insignificant when controlling for disciplinary effects.

Most attrition in years one through four (early attrition) is composed of those students who leave without achieving candidacy. About half of the students who left their programs early did so without a master’s degree. As expected, attrition after achieving candidacy tends to occur after the first year. Only a small portion of early attrition is by students who transferred to another university or to another program within the same university.

C-cohorts, compared to A- and B-cohorts, consistently have lower attrition rates at year four. The lower fourth-year attrition rates in C-cohorts indicate that a lower percentage of students who were enrolled in 1998-99 through 2000-01 left their programs in the first four years than of students who were enrolled in 1992-93 through 1997-98.

64 The ultimate goal of the Ph.D. Completion Project is to analyze the factors that affect doctoral degree completion and attrition patterns and to develop best-practice recommendations for graduate deans, program administrators, policy makers, and the general public. Factors included in this document are broad field, discipline, institution type and cohort size. The next publication in the Ph.D. Completion and Attrition series will include data on student demographics (gender, race, and citizenship status). These demographic data, when aligned with the completion and attrition patterns, will be used to better inform institutional policy on doctoral student completion.

Research universities have proactively agreed to collect these data for doctoral programs before any external federal requirement to do so might emerge. The National Research Council’s Assessment of Research Doctorate Programs, which closely follow the CGS Ph.D. Completion Project’s data templates, has done much to ensure that completion-rate data will be comparable across institutions. As a result of these initiatives, completion-rate data by institution will soon be available for nearly all U.S. research doctoral programs. The availability of these data will mark an important milestone for the U.S. doctoral enterprise. With this availability, however, comes increasing scrutiny of programs that may be perceived as underperformers with respect to Ph.D. completion. When benchmarks for degree completion are applied, it will be important to consider the characteristics of each institution, each disciplinary field, and the students who enroll in them, because the meaning of “completion” and “attrition” will vary by context. Because they understand variations in such contexts, as well as the impact of employment opportunities within a given field, graduate deans and other senior graduate administrators should be leading participants in public discussions of completion rates and inquiries into which students are completing and under what conditions.

65 Appendix A Ph.D. Completion Project – Advisory Board Members

CGS appointed an Advisory Board to guide the project. This group comprises individuals in leadership positions in academia, industry, disciplinary societies, funding agencies, and research programs on minority graduate education. Earl Lewis (Chair) Executive VP for Academic Affairs & Provost Emory University John Benbow Senior Principal Scientist Pfizer Global Research and Development James Duderstadt President Emeritus / Professor of Science & Engineering Director of the Millennium Project University of Michigan Gertrude Fraser Vice Provost for Faculty Advancement University of Virginia Charlotte Kuh Deputy Executive Director National Research Council Joan Lorden Provost University of North Carolina-Charlotte Michael Nettles Senior Vice President, Policy Evaluation and Research Center Educational Testing Service Suzanne Ortega Vice Provost and Dean, Graduate School University of Washington Richard Shavelson Professor of Education and Psychology Stanford University Barbara Williams Senior Director, PGRD Staffing, Diversity and HR Planning Pfizer Global Research and Development

66 Appendix B Ph.D. Completion Project – Phase I Institutions

Among the 46 proposals submitted by universities to participate in Phase I of the Ph.D. Completion Project (2004-2007), 21 universities were selected by an external advisory committee to receive grant funding as “Research Partners” based on the competitiveness of their proposals. The other 25 universities were included in the project as “Project Partners.” Many of these Project Partners voluntarily submitted data, and most of them actively participated in CGS sessions and events dedicated to the project and to issues of doctoral completion and attrition.

Research Partners: Arizona State University University of California, Los Angeles University of Cincinnati Cornell University Duke University University of Florida University of Georgia Howard University University of Illinois at Urbana-Champaign University of Louisville University of Maryland, Baltimore County University of Michigan University of Missouri–Columbia Université de Montréal University of North Carolina at Chapel Hill North Carolina State University University of Notre Dame Princeton University Purdue University Washington University in St. Louis Yale University

67 Project Partners9: University of California, Berkeley University of Colorado at Boulder Florida State University * Fordham University * George Washington University University of Iowa ** Jackson State University University of Kansas Louisiana State University Marquette University * McGill University (Canada) University of Melbourne (Australia) Michigan State University University of Minnesota New Mexico State University New York University ** North Dakota State University Pennsylvania State University * University of Puerto Rico University of Rhode Island Rutgers, the State University of New Jersey ** University of Southern California Southern Illinois University Carbondale ** Syracuse University Western Michigan University *

9 Nine project partners also submitted data to the baseline data analysis of our Ph.D. Completion Project, as indicated by asterisks above. * Program and Demographic Data ** Program Data Only

68 Appendix C Number of Entering Students and Doctoral Programs by Broad Field and Discipline

All Cohorts A-cohorts All Cohorts B-cohortsA-cohorts C-cohortsB-cohorts C-cohorts 1992/93-2003/04 1992/93-1994/951992/93-2003/04 1995/96-1997/98 1992/93-1994/95 1998/99-2000/011995/96-1997/98 1998/99-2000/01 Entering EnteringEntering EnteringEntering EnteringEntering Entering Broad Field/DisciplineBroad Programs Field/DisciplinePrograms Programs ProgramsPrograms Programs Programs Students StudentsStudents StudentsStudents Programs StudentsStudents Students Engineering Engineering Biomedical Engineering Biomedical 6Engineering 658 56 143 658 5 5 125 1436 5 125147 6 147 Chemical Engineering Chemical Engineering7 923 77 240 923 7 7 187 2407 187218 7 218 Civil Engineering Civil Engineering8 1299 88 340 1299 8 8 288 3408 288298 8 298 Electrical and Electronics Electrical and Electronics Engineering Engineering9 2888 99 652 2888 9 9 692 6529 692796 9 796 Mechanical Engineering Mechanical11 Engineering 2205 1111 486 2205 11 11 458 486 11 458 602 11 602 Other Other 13 1523 1213 294 1523 12 12 322 294 12 322 355 12 355 Engineering Total Engineering54 Total 9496 5254 2155 9496 52 52 2072 2155 5352 2072 2416 53 2416 Life Sciences Life Sciences Biology Biology 13 2130 1213 436 2130 13 12 566 436 13 566 648 13 648 Genetics, Molecular GeneticsGenetics, Molecular 5 Genetics 456 5 5 127 456 5 5 103 1275 103113 5 113 Microbiology and ImmunologyMicrobiology 8 and Immunology 679 8 8 217 679 8 8 164 2178 164145 8 145 Molecular and Cellular BiologyMolecular and 11 Cellular 1489Biology 9 11 262 1489 10 9 364 262 10 364 402 10 402 Neuroscience Neuroscience7 463 57 81 463 6 5 121 817 6 121126 7 126 Other Other 22 1708 2022 406 1708 20 20 397 406 20 397 382 20 382 Life Sciences Total Life Sciences66 Total 6925 5966 1529 6925 62 59 1715 1529 6362 1715 1816 63 1816 Mathematics & Physical SciencesMathematics & Physical Sciences Chemistry Chemistry 18 5490 1818 1384 5490 18 18 1284 1384 18 1284 1384 18 1384 Computer and Information Computer and Information Sciences Sciences 13 2976 1113 615 2976 13 11 646 615 13 646 737 13 737 Mathematics Mathematics19 3082 1919 800 3082 19 19 759 800 19 759 779 19 779 Physics and Astronomy Physics and19 Astronomy 3213 1819 765 3213 18 18 749 765 1918 749 814 19 814 Other Other 9 501 99 86 501 9 9 142 869 142137 9 137 Mathematics & Physical Mathematics & Physical Sciences Total Sciences Total78 15262 7578 3650 15262 77 75 3580 3650 7877 3580 3851 78 3851 Social Sciences Social Sciences Anthropology and ArchaeologyAnthropology 6 and Archaeology 792 6 6 184 792 6 6 207 1846 207203 6 203 Communications Communications6 820 66 343 820 6 6 162 3436 162162 6 162 Economics Economics 6 1001 66 275 1001 6 6 232 2756 232218 6 218 Political Science Political Science11 1666 1111 489 1666 11 11 384 489 11 384 379 11 379 Psychology Psychology19 3783 1919 1007 3783 19 19 942 1007 19 942 873 19 873 Sociology Sociology 12 1216 1212 297 1216 12 12 318 297 12 318 289 12 289 Other Other 5 393 35 75 393 4 382 754 10982 4 109 Social Sciences Total Social Sciences65 Total 9671 6365 2670 9671 64 63 2327 2670 64 2327 2233 64 2233 Humanities Humanities English Language and English Language and Literature Literature 19 2889 1819 799 2889 19 18 807 799 19 807 652 19 652 Foreign Languages and Foreign Languages and Literatures Literatures10 655 1010 244 655 10 10 130 244 109 130132 9 132 History History 17 2430 1717 653 2430 17 17 665 653 17 665 551 17 551 Philosophy Philosophy 8 774 88 195 774 8 8 187 1958 187193 8 193 Other Other 13 1011 1113 240 1011 11 11 214 240 1211 214 281 12 281 Humanities Total Humanities67 Total 7759 6467 2131 7759 65 64 2003 2131 65 2003 1809 65 1809

Grand Total for All Broad Grand Total for All Broad Fields Fields 330 49113 313330 12135 49113 320 313 11697 12135 323320 11697 12125 323 12125

69 6

Appendix D

Data Templates

Data Templates Templates Data

Appendix D D Appendix

71 Appendix E

Investigation of Cohort-Size Effects vs. Disciplinary Effects on Attrition Appendix E

Figure 4-25 shows that Electrical & Electronics Engineering has a very high E.1: Investigation of Cohort-Size Effects vs. Disciplinary attrition rate, 44.1% at year ten. Also, 41.2% of all large-cohort students comeEffects from on Electrical Attrition & Electronics Engineering, as shown in Table E.1. This phenomenon implies that the disciplinary effect from Electrical & Electronics Figure 4-25 shows that Electrical & Electronics Engineering has a very high Engineeringattrition rate, 44.1% may atpartially year ten. contributeAlso, 41.2% toof theall large-cohort hierarchical students distribution come of attrition ratesfrom Electricalacross cohort & Electronics sizes. Engineering, Figures E.1 as shown and E.2in Table show E.1. furtherThis evidence of the influencephenomenon from implies Electrical that the disciplinary & Electronics effect from Engineering Electrical & Electronics on the cohort effects. Engineering may partially contribute to the hierarchical distribution of attrition Whenrates across Electrical cohort sizes. & Electronics Figures E.1 and Engineering E.2 further evid is enceremoved, the influence large from cohorts do not haveElectrical the &highest Electronics attrition Engineering rates, on as the is cohort shown effects. in WhenFigure Electrical E.1 and & E.2. Instead, mediumElectronics cohorts Engineering have is higher removed, attrition large cohorts rates do than not havethe largethe highest cohorts attrition for the rest of rates, as is shown in Figure E.1 and E.2. Instead, medium cohorts have higher theattrition Engineering rates than thedisciplines. large cohorts for the rest of the Engineering disciplines.

Table E.1. Student Counts by Cohort Size in Engineering for Cohorts EnteringTable E.1. inStudent 1992-93 Counts though by Cohort 1994-95 Size in Engineering for Cohorts Entering in 1992-1993 though 1994-1995

Student Counts by Cohort Sizes Engineering Disciplines Small Medium Large # % # % # % Major Engineering disciplines Biomedical Engineering 130 6.99 790 14.03 510 3.63 Chemical Engineering 390 20.97 1230 21.85 780 5.55 Civil Engineering 320 17.20 430 7.64 2650 18.85 Electrical and Electronics Engineering 120 6.45 640 11.37 5760 40.97 Mechanical Engineering 150 8.06 1600 28.42 3110 22.12 Total 1110 59.68 4690 83.30 12810 91.11 Other Engineering disciplines Aerospace Engineering 170 9.14 Computer Engineering 70 3.76 220 3.91 630 4.48 Environmental Engineering 30 1.61 410 7.28 Industrial Engineering 180 9.68 Materials Engineering 60 3.23 90 1.60 Nuclear Engineering 90 4.84 220 3.91 Engineering, Other 150 8.06 620 4.41 Total 750 40.32 940 16.70 1250 8.89 Grand Total 1860 100.00 5630 100.00 14060 100.00

8 72

Figure E.1 Cumulative Ten-Year Attrition Rates for A-Cohorts in FigureEngineering E.1 Cumulative (except Electrical Ten-Year & Electronics Attrition Engineering),Rates for A-Cohorts by Cohort in Engineering (except Electrical & Electronics Engineering), by Cohort Sizes2 Sizes10 and Year and Year 25 20 15 10 5 Cumulative Attrition Rate Attrition Cumulative 0

1 2 3 4 5 6 7 8 9 10 year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

“Small”-- cohort size ranging from 1 to 7 students “Medium”“Small”-- --cohort cohort size size ranging ranging from from 8 to 1 14 to 7 students “Large”--“Medium” cohort --cohort size ranging size ranging from 15 fromto 51 8 to 14 (This“Large”-- is the common cohort sizeclassi rangingfication offrom cohort 15 size)to 51 (This is the common classification of cohort size)

10 Cohort-size definition in this graph uses the shared definition for all disciplines and fields.

2 Cohort-size definition in this graph uses the shared definition for all disciplines and fields. 73

Figure E.2 Cumulative Ten-Year Attrition Rates for A-Cohorts in FigureEngineering E.2 Cumulative (except Electrical Ten-Year & Attrition Electronics Rates Engineering), for A-Cohorts by in Discipline- Engineeringspecific Cohort (except Sizes Electrical11 and Year & Elect ronics Engineering), by Discipline- specific Cohort Sizes3 and Year 25 20 15 10 5 Cumulative Attrition Rate Attrition Cumulative 0

1 2 3 4 5 6 7 8 9 10 year

Small Medium Large

Source: Council of Graduate Schools Completion and Attrition Program Data

Note: A-cohorts are the cohorts entering in the academic year 1992-1993 through 1994-1995 “Small”--Note: A-cohorts cohort size are rang theing cohorts from 1 toentering 6 students in the academic year 1992-1993 through 1994-1995 “Medium”“Small”-- --cohort cohort size size ranging ranging from from 7 to 1 13 to 6 students “Large”--“Medium” cohort --cohort size ranging size ranging from 14 fromto 51 7 to 13

“Large”-- cohort size ranging from 14 to 51

11 Cohort size in this case is defined according to the sample distribution in this discipline.

3 74 Cohort size in this case is defined according to the sample distribution in this discipline.

10 References

Attiyeh, G. (1999). “Determinants of Persistence of Graduate Students in Ph.D. Programs.” Princeton, NJ: GRE Board Research Report 95-18R.

Berelson, B. (1960). Graduate Education in the U.S. New York: McGraw-Hill.

Bowen, William, and Neil Rudenstine. (1992) In Pursuit of the Ph.D. Princeton, N.J.: Princeton University Press.

Case, B. A. & Blackwelder, M. A. (1993). “Doctoral Retention, Departmental Expectations, and Teaching Preparation.” Notices of the AMS 40(7): 803-11.

Council of Graduate Schools. (2004). Ph.D. Completion and Attrition: Policy, Numbers, Leadership, and Next Steps. Washington, D.C.: Council of Graduate Schools.

Denecke, Daniel, and Helen Frasier. (2005), “Ph.D. Completion Project: Preliminary Results from Baseline Data.” CGS Communicator (November).

Ehrenberg, R.G. Zukerman, H., Groen, J. & Brucker, S. M. (2006). The Graduate Education Initiative: Description and Preliminary Findings. Ithaca, NY: Cornell Higher Education Research Institute.

Espenshade, T. J. & Rodriguez, G. (1997). “Completing the Ph.D.: Comparative Performances of U.S. and Foreign Students.” Social Science Quarterly 78(2): 593-605.

Golde. C. M. (1998). “Beginning Graduate School: Explaining First-Year Doctoral Attrition.” New Directions for Higher Education, 26(1): 55-64.

Golde. C. M. (2005). The Role of the Department and Discipline in Doctoral Student Attrition: Lessons from Four Departments. Columbus, OH: Ohio State University Press.

75 Groen, J.; Jakubson, G.; Ehrenberg, R.; Condie, S. & Liu, A. (2005). Program Design and Student Outcomes in Graduate Education. Cornell, NY: Cornell Higher Education Research Institute.

Levitt, S. D. and Dubner, S. J. (2005). Freakonomics: A Rogue Economist Explores the Hidden Side of Everything. New York: William Morrow

Lovitts, B. E. (2001). Leaving the Ivory Tower: The Causes and Consequences of Departure from Doctoral Study. Lanham, MD: Rowman & Littlefield Publishers, Inc

Nerad, M. & Ceny, J. (1991). “From Facts to Action: Expanding the Educational Role of the Graduate Division.” CGS Communicator. Special Edition (May).

Nettles, M.T and Millet, C. M. (2006). Three Magic Letters: Getting to Ph.D. Baltimore, MD: Johns Hopkins Press.

Miselis, K. L., McManus, W. & Kraus, E. (1991). “We Can Improve Our Graduate Programs: Analysis of Ph.D. Student Attrition and Time- to-Degree at the University of Pennsylvania.” Paper presented at the Annual Forum of the Association of Institutional Research, San Francisco, California, May 29.

Pion, G. M. (2001). The Early Career Progress of NRSA Predoctoral Trainees and Fellows. National Institutes of Health Publication No. 00-4900.

Tucker, A. (1964). Factors Related to Attrition Among Doctoral Students. Retrieved from Education Resource Information Center Database, Document No. ED 003049

Zwick, R. (1991). “Differences in Graduate School Attainment Patterns Across Academic Programs and Demographic Groups.” Princeton, NJ: Minority Graduate Education Project, Educational Testing Service and Graduate Record Examinations.

PH.D. COMPLETION AND ATTRITION: Analysis of Baseline Program Data from the Ph.D. Completion Project

PH.D. COMPLETION AND ATTRITION Analysis of Baseline Program Data from the Ph.D. Completion Project

Council of Graduate Schools One Dupont Circle, NW Suite 230 Washington, DC 20036-1173 Phone (202) 223-3791 Fax (202)331-7157 www.cgsnet.org