Western Michigan University ScholarWorks at WMU Dissertations Graduate College 12-2009 Persistence and Success: A Study of Cognitive, Social, and Institutional Factors Related to Retention of Kalamazoo Promise Recipients at Western Michigan University Michelle Ann Bakerson Western Michigan University Follow this and additional works at: http://scholarworks.wmich.edu/dissertations Part of the Educational Leadership Commons, Education Policy Commons, and the Higher Education Commons Recommended Citation Bakerson, Michelle Ann, "Persistence and Success: A Study of Cognitive, Social, and Institutional Factors Related to Retention of Kalamazoo Promise Recipients at Western Michigan University" (2009). Dissertations. 647. http://scholarworks.wmich.edu/dissertations/647 This Dissertation-Open Access is brought to you for free and open access by the Graduate College at ScholarWorks at WMU. It has been accepted for inclusion in Dissertations by an authorized administrator of ScholarWorks at WMU. For more information, please contact [email protected]. PERSISTENCE AND SUCCESS: A STUDY OF COGNITIVE, SOCIAL, AND INSTITUTIONAL FACTORS RELATED TO RETENTION OF KALAMAZOO PROMISE RECIPIENTS AT WESTERN MICHIGAN UNIVERSITY by Michelle Ann Bakerson A Dissertation Submitted to the Faculty of the Graduate College in partial fulfillment of the requirements for the Degree of Doctor of Philosophy Department of Educational Leadership, Research and Technology Advisor: Gary Miron, Ph.D. Western Michigan University Kalamazoo, Michigan December 2009 PERSISTENCE AND SUCCESS: A STUDY OF COGNITIVE, SOCIAL, AND INSTITUTIONAL FACTORS RELATED TO RETENTION OF KALAMAZOO PROMISE RECIPIENTS AT WESTERN MICHIGAN UNIVERSITY Michelle Ann Bakerson, Ph.D. Western Michigan University, 2009 The Kalamazoo Promise, a universal scholarship program announced in November 2005 provides four years of tuition and fees at any of Michigan's two- or fouryear public colleges or universities for students who have attended Kalamazoo Public Schools. This investment in the community is being replicated elsewhere across the nation, including Denver and Pittsburgh. The scholarship program lowers the cost of postsecondary education, thereby increasing incentives for high school graduation, college enrollment, and college completion. Of the 307 Kalamazoo Promise Scholarship recipients who have attended Western Michigan University since its inception, 16% have been academically dismissed. The main objectives of this study were to: (1) examine persisters, those on probation, and non-persisters in terms of the Cognitive, Social and Institutional factors of retention, (2) examine persisters, those on probation, and non-persisters in terms of average courses taken per term and number of courses taken the first year and, (3) examine non-response bias in terms of respondents, late respondents, and nonrespondents. Following are highlighted some of the key findings from the dissertation: Persisters had higher high school GPAs and higher ACT composite scores and were more likely to be White. Similarly, persisters took, on average, more courses per term and more courses the first year than either those on probation or non-persisters. As a contribution to research and evaluation, a number of different approaches were used to study potential non-response bias among scholarship recipients. Depending on the approach, small or insignificant differences in non-response bias were identified. Because non-response bias was minimal, the overall findings and conclusions were viewed as valid and did not need to be adjusted. Various factors in the literature, such as parental income and living in a dorm, found to contribute to retention of students did not function as expected with this population. Also, the examination of non-response error and therefore possible nonresponse bias were extra steps taken to help ensure the quality of the generalizations being made. It is hoped that further research using these results as a benchmark will be conducted in order to more fully understand persistence and success of Kalamazoo Promise recipients. Copyright by Michelle Ann Bakerson 2009 UMI Number: 3392137 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. UMI Dissertation Publishing UMI 3392137 Copyright 2010 by ProQuest LLC. All rights reserved. This edition of the work is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 ACKNOWLEDGEMENTS A dissertation is a strange experience. In the beginning it feels like a distant, vague dream, one that you can barely see. Towards the middle you think, 'Maybe I will be able to do this, just maybe I can get this done.' While everyone encourages you, and those who have been through it know you can do it, you never really believe them, imagining that the end is just too far away. And when, after all the hours, days, late, late nights, and weekends, one day that last section is finally done, it is an awesome feeling to know that you reached high and achieved a goal that took so long to accomplish. Without Dr. Gary Miron, my chairperson, this goal of mine would not have been possible. Your tireless hours spent helping me, bouncing ideas, solving various problems, connecting me with stakeholders, helping with participant incentives or even helping me speak to participants when I had laryngitis, you were always supportive, encouraging and positive. I could not have asked for a better chair or a better experience. You have been an extremely knowledgeable mentor and I appreciate everything that you have done for me. My sincere thanks go to Dr. Jessaca Spybrook for your substantial feedback, attention to detail and extensive knowledge. Your time and guidance have been very much appreciated. Dr. Andrea Beach, thank you for agreeing to be on my committee; your feedback has also been very helpful and your expertise very much appreciated. I ii Acknowledgments—Continued could not have asked for a better committee. Dr. Brooks Applegate, thank you for always challenging me and encouraging me to reach farther. Your classes were some of the most challenging and worthwhile. I appreciate everything you have taught me over the years and thank you for always making me think outside of the norm. Special thanks go to Bob Jorth, administrator of the Kalamazoo Promise; Tracy Pattok, director of the Student Academic and Institutional Research office at Western Michigan University, and to Patricia Williams, facilitator of the Kalamazoo Promise at WMU. Without the three of you, I would not have been able to complete this research. Thank you for your time and willingness to help with this endeavor. A sincere thank you goes to my business partner, Nakia James, for pulling the weight at Momentum Consulting and Evaluation, LLC, and for being so supportive. Katya L. Gallegos Custode and Tammy DeRoo, thank you for your help with the interviews. Without your willingness to help this all would have taken so much longer. Thank you to all of my friends at Western, including June Gothberg, John Hoye, Michael Noakes, Maxine Gilling, Nadini Persaud, Fatma Ayyad, Julien Kouane, Manuel Brennes, Chris Coryn, Daniella Schroter, Brandon Youker, Christian Gugiu, Wesley Martz, Amy Gullickson, Otto Gustafson, Willis Thomas, and Thomaz Chianca it was always a pleasure working with you. I enjoyed our many conversations together. iii Acknowledgments—Continued Thank you to all of my friends and colleagues at Indiana University South Bend for your support and encouragement, including but not limited to Dr. Yvonne Lanier, Dr. Denise Skarbeck, Dr. Gwendolyn Mettetal and Erika Zynda. Thank you to Dr.Michael Horvath, Dr. Karen Clark and Dr. Bruce Spitzer for believing in me and giving me the opportunity to be part of the IUSB community. A very special thank you goes to my family for hanging in there with me. To my brothers, Charley and Andy, thank you for continually calling and checking on me, even when I didn't have time to call you back. Your belief in me and support have helped me get this far. I love you guys. Thank you to Tina, Brandan, Courtney, Alex and Tanner, too. Thank you to Lori, Michael, Mika, Taylor, Sandy, Vera and Judy you have been such a tremendous support to me. Your help with the kids and encouragement has meant so much to me. Michael and I could not do the things we do without you. Thank you to all of my friends for not giving up on me and still inviting me when you knew I wouldn't be able to go. You made me feel loved and that is exactly what I needed during this time. A special thank you also goes to Tori Davies, University of Notre Dame, for your wonderful editing, Erik Gunn with Great Lakes Editorial Services for your final editing, to Dr. Sandra Harris, Troy University and to Maureen Hogue, University of Notre Dame, and Dr. Martin Klubek, University of Notre Dame. My deepest thank you goes to my husband, who has relentlessly stood by my side, iv Acknowledgments—Continued encouraging me all the way through this process. Thank you for picking up the pieces I have dropped along the way and keeping our family together. Without you I would not have been able to accomplish this. I love you with all of my heart. Equally important a huge thank you goes to my children, Bailey, Aliea and Audrey, you have been extremely understanding with my lack of time. Without your help around the house, your understanding and your patience I am sure I would not have been able to finish this. Thank you for always being happy for me when I was one step closer. You three are my life and I cannot wait to finally do all the things we have been planning. Bailey, Aliea and Audrey, this dissertation is dedicated to you: may you reach high and dream big. Michelle Ann Bakerson v TABLE OF CONTENTS ACKNOWLEDGEMENTS ii LIST OF TABLES x LIST OF FIGURES xiv CHAPTER I. INTRODUCTION 1 Background of the Kalamazoo Promise 2 Development of the Kalamazoo Promise 3 Problem Statement and Research Questions 6 Research Question One 6 Research Question Two 8 Research Question Three 8 Methodological Overview 10 Rationale for the Dissertation 10 Structure and Overview of the Dissertation 11 II. REVIEW OF RELATED LITERATURE 12 Kalamazoo Promise 12 College Retention 18 Non-response Bias 23 Conclusion 31 vi Table of Contents—Continued CHAPTER III. METHODOLOGY 32 Purpose 32 Research Design 33 Sample 34 Procedures for Data Collection 36 Informed Consent Process 38 Research Procedure 39 Data Analysis 43 Research Questions 1 and 2 47 Research Question 3 48 Ethical Considerations 52 Limitations 55 Summary 56 IV. RESULTS 57 Summary Academic Data 57 Survey Summary 64 Research Question One Results 73 Research Question 1.1 Results 74 Research Question 1.2 Results 84 vii Table of Contents—Continued CHAPTER Research Question 1.3 Results 85 Research Question Two Results 87 Research Question 2.1 Results 87 Research Question 2.2 Results 91 Research Question ThreeResults 93 Data Analysis for Research Question 3 94 Research Question 3.1 Results 98 Research Question 3.2 Results 102 Research Question 3.3 Results 105 V. CONCLUSIONS Ill Central Findings Ill Research Question One Conclusion 112 Research Question Two Conclusion 118 Research Question Three Conclusion 120 Research Question 3.1 120 Research Question 3.2 123 Research Question 3.3 125 Future Research 129 Potential Implications 130 vin Table of Contents—Continued 135 REFERENCES APPENDICES A. Participant Paperwork 142 B. Cognitive, Social and Institutional Factor of Retention and Corresponding Survey Items, Academic Data Variable Names with Measurement Type and Cognitive, Social and Institutional Factors of Retention and Corresponding Survey Items with Subscales Identified 171 C. Summary Demographic Data on Interval Level Data Across 178 D. Tests of Normality and Persistence, those on Probation and Non-persistence, Subscale Scores of the Survey of Promise Recipients by Response Category.... 183 E. Tests of Homogeneity of Variance and Subscales of the Subscale Scores of the Survey of Promise Recipients by Response Category 185 F. Summary Results of Item Analysis, Institutional Support, Social Engagement, Social Demands and Cognitive Engagement 187 G. Survey Summary Tables 198 IX LIST OF TABLES 1. Kalamazoo Promise Summary Data 13 2. College or University Attendance for Current Promise Users as of Fall 2009 14 3. Bias and Percentage Bias in Respondent Mean Relative to Total Sample Mean 30 4. Data Obtained from the Facilitator of the WMU Kalamazoo Promise and the Office of Student Academic and Institutional Research 35 5. Breakdown by Type of Data Collected and Persistence 35 6. Summary Research Questions 1 and 2 with Independent and Dependent Variables, Data Source and Method of Analysis 7. Summary Research Question 3 with Independent and Dependent Variables, 50 Data Source and Method of Analysis 51 9. Average High School GPA by Probation Status at WMU 59 10. Most Recent WMU GPA by Probation Status at WMU 60 11. First Promise Semester 61 12. FTIAC Cohort by Persistence and Non-Persistence 62 13. High School Attended by WMU Kalamazoo Promise Recipients 63 14. Distribution of Promise Students by Race, Gender and High School 63 15. High School by Persistence and Non-Persistence 64 16. Distribution of Promise Students who Answered the Survey by Probation Status .... 65 17. Did You Begin College at WMU or Elsewhere? 66 18. Do You Expect to Enroll for an Advanced Degree When, or if, You Complete Your Undergraduate Degree? 19. Where Do You Live During the School Year? 67 67 x List of Tables—Continued 20. What is the Highest Level of Education Obtained by Your Father or Mother? 68 21. About How Many Hours Do You Spend in a Typical 7-day Week Doing Each of the Following? 69 22. If You Have a Job, How Does it Affect Your School Work? 69 23. How Likely is it That the Following Issues Would Cause You to Withdraw From Class or From WMU? 70 24. Are You a Member of a Social Fraternity or Sorority? 70 25. How Supportive Are Your Friends of Your Attending WMU? 71 26. How Supportive is Your Immediate Family of Your Attending WMU? 71 27. Which Best Represents the Quality of Your Relationship With Students at WMU? 28. Which Best Represents the Quality of your Relationships With Instructors at WMU? 29. Which Best Represents the Quality of Your Relationship With Administrative Personnel & Office Staff at WMU? 72 72 72 30. If You Could Start Over Again, Would You Still Attend WMU? 73 31. Would You Recommend WMU to a Friend or Family Member? 73 32. Summary of Omnibus MANOVA Test of Group Differences across Dependent Variables 76 33. Summary of Test for Differences by Race across the Dependent Variables 77 34. Summary of Test for Differences by Gender across the Dependent Variables 79 35. Summary of MANOVA Results for Persistence across the Dependent Variables 80 36. Summary of MANOVA Results for Persistence and Race Interaction across the Dependent Variables 81 xi List of Tables—Continued 37. Crosstab for Tests of Differences Among Persisters, Those on Probation, and Non-persisters Across the Variables of Persistence and Taking Remedial Courses at WMU, AP Credit, Gender, and Race 83 38. Crosstab for Tests of Differences Among Persisters, Those on Probation, and Non-persisters Across the Variables of Persistence and Taking Remedial Courses at WMU, AP Credit, Gender, and Race 86 39. ANCOVA Results of Persisters and Non-persisters, Controlling for Race and Gender for Average Number of Courses Taken Per Term 88 40. Descriptive Statistics for Gender, Race and Persistence for Average Number of Courses Taken Per Term 89 41. ANCOVA Results of Persisters, Those on Probation, and Non-persisters, Controlling for Race and Gender for Number of Courses Taken the First Year 90 42. Descriptive Statistics for Gender, Race and Persistence for Number of Courses Taken the First Year 90 43. ANCOVA Results of Respondents, Late Respondents and Non-respondents, Controlling Across Race and Gender for Average Number of Classes Taken Per Term 91 44. Descriptive Statistics for Gender, Race and Response for Average Number of Courses Taken Per Term 92 45. ANCOVA Results of Respondents, Late Respondents and Non-respondents, Controlling across Race and Gender for the Number of Courses Taken the First Year 93 46. Descriptive Statistics for Gender, Race and Response for Number of Courses Taken the First Year 93 47. Summary of Results from the Reliability Analysis and Descriptive Statistics for Subscales of the Survey of Promise Scholarship Recipients 96 48. Distribution of Promise Students who Answered the Survey by Probation 99 49. Summary of MANOVA Comparison of Survey of Promise Scholarship Recipients at WMU Spring 2009 across the Demographic Variables 101 50. Test Groups Differences across the Categorical Variables 102 xii List of Tables—Continued 51. Summary of MANOVA Results for Early and Late Respondents of the Survey of Promise Scholarship Recipients on the Four Summated Subscale Scores 103 52. Summary of Comparison of Early Respondents and Late Respondents across the Categorical Variables of the Survey of Promise Scholarship Recipients 104 53. Descriptive Statistics for the Subscale Scores of the Survey of Promise Scholarship Recipients 106 54. Response Bias Estimates for the .Subscale Scores of the Survey of Promise Scholarship Recipients Based on Mean Scores of the Participants Using the 53% Response Rate 107 55. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Median Scores of the Participants Using the 53% Response Rate 108 56. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Mean Scores of the Participants Using the 33% Rate 109 57. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Median Scores of the Participants Using the 33% Response Rate 110 58. Results of Mean and Median used in Bias Ratio Formula at the 53% Response Rate 125 59. Results of Mean and Median used in Bias Ratio Formula at the 53% and 33% Response Rate 126 xni LIST OF FIGURES 1. Kalamazoo District Enrollment Trend 16 2. District Enrollment Trends for Kalamazoo, Battle Creek and Flint City Public School Districts 17 xiv CHAPTER I INTRODUCTION Across the nation, states and school districts struggle with high dropout rates and fairly low college enrollment rates (Cataldi, E., Laird, J., & KewalRamani, A. [NCES], 2009). These issues can leave them at a competitive disadvantage as state and federal education authorities increasingly tie aid to measures of performance such as test scores and graduation rates. In addition, the cost of higher education has increased 7% more than the rate of inflation over three years (Cunningham, A. [NCES], 2005). The disparity between the cost of education and the meager rise in family incomes leaves some families wondering whether or not post-secondary education will be attainable. Recognizing these issues, many states, cities and citizens have acknowledged the need to improve the education systems, hoping to increase their communities' human and social capital. In Kalamazoo, Michigan, an anonymous group of concerned citizens created a novel and innovative post-secondary education scholarship program solely for Kalamazoo district school children. The idea of this universal scholarship is very simple: students who live in the district and have gone to school in the district for at least four years qualify to receive scholarship money; all they have to do is be accepted into a Michigan higher education institution. The one-page application is straightforward and simple for families to understand and fill out. In short, this program provides extraordinary access to higher education. Yet since the first scholarships were awarded in 2006, 49 Western Michigan University (WMU) Kalamazoo Promise recipients have attended WMU but have not been 1 retained—nearly 1 out of every 6. This is of major concern to WMU, the Kalamazoo Public School district, and to administrators of the Kalamazoo Promise scholarship itself. Cognitive, social or institutional factors may all contribute to retention or non-retention under this program. Before these factors can be explored in depth, however, it is necessary to understand the backdrop of the Kalamazoo Promise program. Background of the Kalamazoo Promise Since its announcement in November 2005, the Kalamazoo Promise universal scholarship program has already garnered attention far beyond its home community; places as far-flung as Denver, Colorado, Pittsburgh, Pennsylvania, and El Dorado, Arkansas, are attempting to replicate the program in their own communities. For students who already planned to attend college, the Promise may ease the financial burden of a post-secondary education. For students who are unsure if they will attend college, the Promise gives a tangible opportunity to do so, and may enable them to attend full-time, work less, or both. For students who have written off college as financially out of reach, the Promise helps lessen the financial burden and offers low-income students new hope and possibilities. Younger children, especially, who may not have role models who have gone to college, will benefit from a K-12 experience that expects them to continue their education into post-secondary school. The Kalamazoo Promise Scholarship, unlike other scholarship programs, does not look at need or merit; the only stipulation is having attended Kalamazoo Public Schools for at least four years and getting accepted into a Michigan higher education institution. Students who meet these two criteria will have all four years of tuition and fees paid for, 2 regardless of their financial need or other scholarships they may have received. In essence, all Kalamazoo Public School children now have the opportunity for a free K-16 education if they are able to take advantage of it. Despite this opportunity, currently 49 students at Western Michigan University (WMU) who met the criteria of the Kalamazoo Promise Scholarship and attended WMU did not finish their degree. Understanding why is imperative to determine whether WMU can do something to retain these students or whether, instead, keeping them in college is more appropriately the responsibility of the Kalamazoo Public Schools or the Promise program itself. Development of the Kalamazoo Promise Initially, the Kalamazoo Promise was run out of the Kalamazoo Public School district superintendent's office with no official staff. Alex Lee, Kalamazoo Public Schools' (KPS) executive director for communications, fielded communication from the press while the KPS website was the primary source for information regarding the scholarship. The first official staff member of the program, Robert Jorth, was hired in March 2006 as the Kalamazoo Promise administrator. The official duties of this position included determining eligibility, maintaining a database of eligible students, and disbursing Promise funds to colleges and universities. Jorth was solely responsible for setting up and managing the scholarship program and has managed to create a system in which data is easily usable and accommodates the terms of the scholarship. The system he has established works, and it's simple: for example, the Kalamazoo Promise 3 scholarship application form is a one-page document. Initially, there was confusion about KPS and the Kalamazoo Promise being one entity because they were located in the same building, but in the fall of 2006, the Promise administrator moved out of the school district building and into an office in the Kalamazoo Communities in Schools (KCIS) downtown facility, which was donated by a real estate developer. KCIS organizes community partners in their efforts to help children learn and stay in school. The Kalamazoo Promise incorporated as a 501(c)(3) organization and established a web site with its own domain name, https://www.kalamazoopromise.com. These changes helped more concretely distinguish the Promise from KPS. Robert Jorth continues to run the program and has disbursed $10.45 million in scholarship money to 26 schools (Jorth, 2009). Jorth was the sole employee dedicated to the program until September 2008, when Janice Brown joined as executive director. Brown, who had been superintendent of the Kalamazoo school district until August of 2007, has had a key role in implementing the Kalamazoo Promise, arranging for anonymous donors to fund the scholarship. Kalamazoo's Vice Mayor, Hannah McKinney, stated, "I don't think we would have the Promise were it not for Janice. She's a little Energizer bunny with a lot of integrity" (Yung, 2007, p.l). Brown credits the donors themselves with the concept: giving every Kalamazoo student the opportunity to attend post-secondary school. "It was really the idea of the donors after a long, long series of discussions about a question. What can we do to make a turnaround, to make an impact on an urban city in which we really wish to invest? And over and over again, the answer seemed to be invest in education, invest in out youth" (Yung, 2007, p.l). Brown continues to be the bridge between donors and the 4 community and is the only person who has publicly acknowledged knowing the identity of the donors. Backdrop. Kalamazoo, Michigan, is located in the Southwest corner of the state. According to the U.S. Census Bureau, Kalamazoo County's high school graduation rate is 88.8%, 31.2% of its population holds a bachelor degree or higher, and the median household income is $43,450, with 13.8% of its population falling below poverty (2009). Even during the current economic downturn, college costs continue to rise. In-state costs are up 37% at public post-secondary institutions, with charges for tuition, fees, room and board of $14,333 for in-state students and $25,200 for out-of-state students. Private colleges in the state charge an average of $34,132 (Jayson, 2009). Such expenses are increasingly burdensome for families, leaving many unable to even consider the option of higher education. Thus, for people especially hard-hit by the economy, the Kalamazoo Promise Scholarship is seen by some as "a miracle" (Miron, Spybrook & Evergreen, 2008, p. 18). This universal scholarship, the first of its kind, represents an enormous commitment on the part of its funders. The anonymous donors have promised to carry the financial burden for all students who live within the district boundaries and attend Kalamazoo Public Schools for at least four years or 130 credits, whichever comes first. This program is intended to last in perpetuity and involves no public funds. Despite this opportunity, however, the office of Institutional Research at Western Michigan University reports that approximately 49 Kalamazoo Promise Scholarship recipients who attended Western Michigan University (WMU) are no longer at WMU and have not finished their degrees. Understanding the influences that led these students to leave 5 college would be valuable information for the Kalamazoo Promise and Western Michigan University, and may aid in decisions regarding retention of students who are Kalamazoo Promise Scholarship recipients. Problem Statement and Research Questions The purpose of this dissertation was to determine to what extent persisters, those on probation, and non-persisters differ on demographic characteristics, and on each of the following in Swail's (2003) Geometric Model of Student Persistence and Achievement: Cognitive, Social and Institutional retention factors. This study also examined the average number of courses take per term and the number of courses taken the first year by persisters, those on probation, and non-persisters; and respondents, late respondents and non-respondents. Additionally, the extent to which respondent, late-respondent and non-respondent Kalamazoo Promise recipients differ on each of Swail's (2003) Geometric Model of Student Persistence and Achievement was examined using known characteristics from the academic data. Similarly, the difference between early respondents and late respondents was examined using the same model, but this time using the actual survey data. Lastly, non-response error was examined using Groves and Couper's bias ratio formula (1998); the formula was modified, using the median to compare differences in indications of non-response bias in the survey depending on whether the mean or median estimator is used. The research questions are as follows: Research Question One To what extent do persister, those on probation, and non-persister Kalamazoo 6 Promise recipients differ by demographic characteristics on each of the following selected factors in Swail's (2003) Geometric Model of Student Persistence and Achievement: (a) Cognitive Factors, (b) Social Factors, and (c) Institutional Factors? More specifically: 1.1 Among the three groups of students, persister, those on probation, and nonpersister Kalamazoo Promise recipients by gender and race, are there any differences in the cognitive factors from Swail's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) high school GPA, (b) most recent WMU GPA, (c) ACT composite score, (d) taking a remedial math course at WMU, (e) taking a remedial reading course at WMU, (f) taking a remedial writing course at WMU, or (g) taking AP credit? 1.2 Among the three groups of students, persister, those on probation, and nonpersister Kalamazoo Promise recipients by gender and race, are there any differences in the social factors from Swail's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) living in a dorm, (b) being an athlete or (c) parental income? 1.3Among the three groups of students, persister, those on probation, and nonpersister Kalamazoo Promise recipients by gender and race, are there any differences in the institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) first year experience (FYE), and (b) 7 which high school Promise students came from? Research Question Two Research question two is broken into two sections, both of which examine the numbers of courses taken by students. Research question 2.1 examines persisters, those on probation, and non-persisters, while research question 2.2 examines respondents, late respondents, and non-respondents, controlling for race and gender. 2.1 Is there a difference in the average number of courses taken per term and number of courses taken the first year among the three groups of students: Persister, those on probation, and non-persister WMU Kalamazoo Promise recipients, controlling for gender and race, using the course summary data? 2.2 Is there a difference in the average number of courses taken per term and number of courses taken the first year among the three groups of students: Respondents, late respondents, and non-respondents of the Survey of Promise Scholarship Recipients at WMU Spring 2009, controlling for gender and race, using the course summary data? Research Question Three To what extent do respondent, late respondent, and non-respondent Kalamazoo Promise recipients differ on each of the following selected factors in Swail's (2003) Geometric Model of Student Persistence and Achievement: (a) Cognitive Factors, (b) Social Factors, and (c) Institutional Factors using known characteristics from the academic data? In addition, to what extent do early respondents differ from late 8 respondents on variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009? Lastly, using Groves and Couper's bias ratio formula, is there an indication of non-response bias? More specifically: 3.1 Among the three groups of students, respondent, late respondent, and nonrespondent Kalamazoo Promise recipients, are there any differences in the following known dependent variables from the academic data: (a) high school GPA, (b) most recent WMU GPA, (c) ACT composite score, (d) taking a remedial math course at WMU, (e) taking a remedial reading course at WMU, (f) taking a remedial writing course at WMU, (g) taking AP credit, (h) living in a dorm, (i) being an athlete, (j) parental income, (k) first year experience (FYE) or (I) high school that could indicate possible non-response bias? 3.2 Between the two groups of students, early respondent and late respondent Kalamazoo Promise recipients, are there any differences in the cognitive, social or institutional factors from Swail 's (2003) Geometric Model of Student Persistence and Achievement using the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009 indicating possible non-response bias? 3.3 Using and modifying Groves and Couper's bias ratio formula (1998), is there an indication of non-response bias, and is there a difference between using the mean and using the median, a more robust statistic, in determining a bias estimate on the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009? 9 Methodological Overview The mixed methods analysis is a non-experimental relational design of three samples of Kalamazoo Promise recipients, those Western Michigan University (WMU) retained (persisters), those on probation and those WMU did not retain (non-persisters). These samples were compared using the variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009 survey and variables obtained from academic records. In addition, non-persisters were examined in depth to identify what factors influenced their decision to leave WMU. Lastly, non-response bias was examined to account for those who responded to the Survey of Promise Recipients at WMU Spring 2009 and those who did not respond. Rationale for the Dissertation The results of this dissertation may have far-reaching implications for not only Western Michigan University and the Kalamazoo Promise Scholarship program, but the Kalamazoo community as a whole. The findings could offer insight into where retention efforts should be focused and by whom. The factors determined to affect retention might be associated with the efforts of WMU, the Kalamazoo Promise, the Kalamazoo Public School system, or individual circumstances that other community agencies might need to address. This process could be helpful to various stakeholders in establishing a baseline of information and determining a direction for future dialogue and interactions in higher education retention. One contribution of this dissertation is to evaluation technique, in that it illustrates 10 how non-response error and possible non-response bias can be detected even in smallscale research projects or program evaluations. This study offers insights about social science research sampling bias and its effects on this population. Non-response bias has not been brought to the field of evaluation; this dissertation is an opportunity to help bridge the fields of evaluation and research. This is one sampling error that evaluators need to take into account when collecting and interpreting data. Structure and Overview of the Dissertation This study is organized into five chapters. Chapter I, the introduction, provides the background and development of the Kalamazoo Promise scholarship, the problem statement, research questions, methodological overview, and rationale for this dissertation. Chapter II reviews the literature and is broken into three sections: 1) an overview of the Kalamazoo Promise program, its participants and community impacts; 2) a discussion of college retention, including definitions, a historical overview, rationale and factors for successful retention; and 3) a discussion of non-response bias, including definitions, an explanation of non-response and non-response bias, identification of nonresponse categories, and information on detecting non-response bias. Chapter III elucidates the methodology of this study in detail, including participant selection, types of and sources of data, primary and secondary data collection, data analysis, data verification and ethical considerations. The results are presented in Chapter IV, and the conclusion and implications are summarized in Chapter V. 11 CHAPTER II REVIEW OF RELATED LITERATURE The review of relevant literature and research is divided into the following sections: 1) an overview of the Kalamazoo Promise Scholarship and its impact on the Kalamazoo community; 2) a framework on higher education retention, exploring what factors assist successful retention of students; and 3) a discussion of relevant issues regarding non-response bias in social science research, its definitions and implications. Kalamazoo Promise Overview. The Kalamazoo Promise Scholarship, announced in November 2005, is funded by anonymous contributors. It is unique in that it is a universal scholarship awarded to all qualifying students regardless of income or other scholarships, and is the first program of this kind. It is estimated that current in-state tuition per semester ranges from $2,000 at a community college to more than $9,000 at the University of Michigan. This means that a family could receive as much as $18,000 per year per child. Once there are four graduated classes, donors will be spending approximately $12 million per year to fund the scholarship. As of 2009, 1,521 graduates have used the Kalamazoo Promise Scholarship, or 82.7% of those eligible (see Table 1). 12 Table 1. Kalamazoo Promise Summary Data Number of KPS graduates Eligible for the Promise % of graduates eligible for the Promise Number of graduates using the Promise the first semester after graduation % of eligible students using the Promise the first semester after graduation Number of graduates who have used the Promise3 % of eligible students who have used the Promise3 2006 2007 2008 2009 Total 517 409 579 502 549 475 515 455 2I6U 79.1 86.7 86.5 88.3 1841 85.3 303 359 370 72.7 74.6 78.1 339 414 388 370 1521 82.9 82.5 81.7 81.3 82.7 "Students who have used at least some portion of their scholarships as of September 8, 2009. Note. Data provided by Kalamazoo Promise administrator According to data obtained from Mr. Robert Jorth, the Kalamazoo Promise administrator, of those 1,521 graduates who have used the Promise, 323, or 21.2%, currently attend Western Michigan University (see Table 2). This is second only to Kalamazoo Valley Community College (KVCC), which currently enrolls 367, or 24.1%, of the Promise students. Michigan State University and University of Michigan come in a distant third and fourth, with 140, or 9.2% percent, and 110 or 7.2%, respectively, of Promise students currently enrolled. Ferris State University / Kendall School of Art & Design is in fifth place, with only 25, or 1.6 percent, of the Promise students currently enrolled. 13 Table 2. College or University Attendance for Current Promise Users as of Fall 2009 College / University 2006 2007 2008 2009 Total Community Colleges Totals Glen Oaks Community College Grand Rapids Community College Jackson Community College Kalamazoo Valley Community College Kellogg Community College Lake Michigan College Lansing Community College Mott Community College Muskegon Community College Northwestern Michigan College Oakland Community College Washtenaw Community College 41 86 1 2 98 176 1 5 77 1 91 1 3 3 161 1 2 2 2 2 401 2 7 2 367 3 2 9 2 2 0 1 4 693 16 11 25 23 140 8 12 3 2 110 2 1 17 323 1076 Universities Totals Central Michigan University Eastern Michigan University Ferris State University/ KS of A&D Grand Valley State University Michigan State University Michigan Technological University Northern Michigan University Oakland University Saginaw Valley State University University of Michigan University of Michigan Dearborn University of Michigan Flint Wayne State University Western Michigan University Grand Total % Retained 2 38 1 159 4 1 6 3 33 2 2 171 3 4 7 6 36 2 3 1 1 2 179 9 3 5 6 29 1 7 184 3 7 8 42 3 2 2 2 17 34 43 4 87 3 72 4 72 16 2 1 6 92 200 257 277 360 59.0 62.1 71.4 97.3 Note. The data for 2009 is projected. Data provided by Kalamazoo Promise administrator The Kalamazoo Promise has had a significant impact on the surrounding colleges and universities, with 64.1% (690/1076) of currently enrolled recipients attending either 14 WMU or KVCC. Other communities are trying to replicate this scholarship program: the El Dorado Promise in El Dorado, Arkansas, the Denver Scholarship Foundation, and the Pittsburgh Promise are among the first to be established. More communities around the country also are considering the Kalamazoo Promise as a potential model. These scholarship programs were represented at the first annual PromiseNet conference held in Kalamazoo in June of 2008 (Eberts, 2008). Recipients. Eligible recipients of the Kalamazoo Promise Scholarship are students who have attended Kalamazoo Public Schools—Kalamazoo Central High School, Loy Norrix High School, or Phoenix High School—for at least four years and who have been accepted by a Michigan public higher education institution. Recipients who have attended Kalamazoo Public Schools since Kindergarten receive 100% of tuition and fees paid. Students who have attended Kalamazoo Public Schools from ninth grade forward received 65% of tuition and fees paid. For students who have attended KPS for longer than high school but less than the full K-12 period, the scholarship's total value is prorated accordingly. The only requirement to maintain eligibility is that students work toward a degree and maintain a 2.0 GPA in their college courses. Community Impacts. The community impacts of such a scholarship are immense. Kalamazoo public school enrollment began increasing dramatically after the Promise was announced (see Figure 1); reaching in the 2008-09 school year the highest enrollment in approximately 10 years. 15 12,000 11,500 -a 11,000 10,500 * 10,000 9,500 9,000 $Hi ,« & 0> <Vs « • V &b > T5 d* V ^ Figure 1. Kalamazoo District Enrollment Trend Note. Data provided by Michigan DOE website located at: http://www.michigan.gove/cepi.html This trend is even more explicit when examining similar school districts in the State of Michigan. Over the past seven years Battle Creek Public Schools' enrollment has decreased 19%, from 7,922 in 2002-'03 to 6,439 in 2008-'09. Flint City Public Schools' enrollment has decreased 34%, from 21,007 in the '02-'03 school year to 13,798 in the '08-'09 school year. Lastly, Lansing School district has dropped 19% from 17,376 in the '02-'03 school year to 14,160 in the '08-'09 school year. These downward trends are seen across the state. Kalamazoo Public School district was on this same downward trend. In the '02-'03 school year, the district enrolled 11,084 students; enrollment declined consistently through the '05-'06 school year to 10,238, an 8% decrease. Immediately after the Kalamazoo Promise was announced in 2006, enrollment started rising in Kalamazoo school district as other districts continued to decline. From the '06-'07 school year until the '08-'09 school year, student enrollment increased 13%, to 11,696. 16 25,000 20,000 c 1 15,000 (§ 10,000 5,000 0 ^ 7 acfc c^r ^b c&f & ^ ^ $ / / c$F «r , / Figure 2. District Enrollment Trends for Kalamazoo, Battle Creek and Flint City Public School Districts Note. Data provided by Michigan DOE website located at: http://www.michigan.gove/cepi.html Beyond the boundaries of the K-12 or even the K-16 school system, the Kalamazoo Promise is envisioned as a catalyst for change and economic growth (Miron, G. & Evergreen, S., 2007). For students who already planned to attend college, the Promise may ease the financial burden of a post-secondary education. For students unsure if they will attend college, the Promise offers a tangible opportunity not only to do so, but to do so full time and, possibly, work less as well. For children who want to go to college but see the financial burden as insurmountable, the Promise lifts that burden. Low-income students now have hope and possibilities they lacked before the Kalamazoo Promise (Miller-Adam, 2009). Younger children lacking role models who have gone to college especially may benefit from a K-12 experience that routinely expects even them to continue their education through post-secondary school. Besides easing or even removing some families' financial burdens, the scholarship now creates an incentive for families to stay in the community and for new 17 families with children to move into the district so they can take advantage of this scholarship (Miron & Cullen, 2008). Community leaders hope the scholarship program will stabilize the housing market and increase property values. In addition, the community looks much more attractive to businesses seeking to invest, expand or relocate. The Promise can help local companies attract employees drawn by the availability of this scholarship for their own children, and it can help foster a well-trained future workforce as well. The resulting economic growth can create a domino effect, with more families and businesses moving in, increasing the need for other businesses, thus further expanding the job market. More Kalamazoo students than ever before have the opportunity to obtain a postsecondary education. The issue now becomes the capacity of these students to succeed in this environment and their ability to persist to achieve the college degree that they have started to pursue. College Retention Definitions. College retention and attrition is complex and difficult to define. The simplest definitions are those of persisters and non-persisters. Persisters are those students who stay in college and finish their degree; with more persisters, the higher education institution's retention increases. Non-persisters are those who leave for whatever reason; when they do so, the college's retention decreases. (Astin, 1975; Astin, 1999; Bean & Eaton, 2000; Bean & Metzner, 1985; Braxton, 2004; Cabrera, Nora & Cabrera, 1993; Pascarella & Terenzini, 1980; Spady, 1970; Stage, 1989; Swail, 2003; Tinto, 1975). Persistence in the literature refers to the student, while retention refers to 18 the institution. The other term used for non-persistence is "dropout." Where retention means staying in school, dropout means leaving school before degree completion. This seems straightforward. Alexander Astin, however, sees the concept of dropout as problematic because it is imperfectly defined. The so-called dropouts may ultimately become non-dropouts and vice versa.. .But there seems to be no practical way out of the dilemma: A "perfect" classification of dropouts versus non-dropouts could be achieved only when all of the students had either died without ever finishing college or had finished college (1971, p.15). Vincent Tinto (1987) states that there are definite limits to the understanding of student departure; the controversy lies in the labeling of persisters and non-persisters, and how the quality of persisting or non-persisting is measured. College students take numerous avenues in pursuit of their educational goals. One student might enroll, attend full- or part-time for a couple of years, leave, and then return five years later to finish. Others might transfer to another college, enroll in more than one college at the same time, take a full load but then only complete one class, or be put on academic probation. The scenarios are limitless; the lack of a clear way to measure persisters or non-persisters complicates any definition of college retention. Agreeing with Tinto, Bean and Metzner (1985) acknowledge that many students leave college because they have met their goals. Perhaps this process of self-discovery resulted in individual growth and maturation; thus, Bean and Metzner argue, leaving college should not be considered a failure by the student or the institution; that any definition of retention should consider student educational goals; and that a "dropout," therefore, would be defined in light of the student's original intent and outcome (1985). 19 Western Michigan University defines its persisters as those first-time, full-time, degree-seeking beginners (FTIAC) who start in the Fall semester and who are still attending the following Fall semester. According to A Comprehensive Report of Retention Rates, written by WMU's Office of Student Academic and Institutional Research, the current retention rate of the Fall 2006 cohort to their second year in Fall of 2007 is 75.1%, which is 1.1 percentage points lower than the average retention rate of Michigan public universities, 76.2% (2009). Along with the difficulties found in the definitions come the complexities of various models and theories of retention. Retention Models and Theories. Research on college retention, student persistence, student departure and achievement of higher education degree attainment has been enormous (Astin, 1975; Astin, 1884; Braxton, 2004; Bean & Eaton, 2000; Bean & Metzner, 1985; Cabrera, Nora & Cabrera, 1993; Carera, Stampen & Hansen, 1990; Carroll, 1988; Pascarella & Terenzini, 1980; Spady, 1970; St. John, Cabrera, Castaneda, Nora & Asker, 2004; Stage, 1989; Stoecker, Pascarella & Wolfe, 1988; Swail, 2003; Tinto, 1975; Tinto, 1993). Because of the complexity and importance of the issue of retention, research, theories and models continue to grow, develop and be evaluated in the hopes of illuminating its dynamics. Factors for Successful Retention. Many factors contribute to successful retention and much research has investigated what factors contribute to that success. One heavily researched area is the characteristics of persisters and non-persisters. Persisters are more likely to attend college full-time, while non-persisters are more likely to attend part-time (Adelman, 1999; Chan, 2002; Feldman, 1993; Lanni, 1993; Moore, 1995; Naretto, 1995; NCES, 1998; Panos & Astin, 1968; Price, 1993; St. John, 1990; Windham, 1994). Part- 20 time students are also more likely first-generation students, which increases the chance for non-persistence (NCES, 1998). Typically, non-persisters work more hours than persisters (Naretto, 1995). The factor of age is contentious (Grosset, 1991). Much research shows that non-persisters are usually older, while persisters are typically younger (NCES, 1998; Price, 1993; Windham, 1994). Conversely, some research reports the exact opposite (Feldman, 1993). Further factors that have been found to contribute to a student's decision to drop out of college include financial concerns (GAO, 1995), full-time employment, family responsibilities, low grade-point average (NCES, 1998), being an ethnic minority other than Asian, and being of the female gender (Bonham & Luckie, 1993; Guloyan, 1986; Levin & Levin, 1991; Rendon, Jalomo & Nora, 2004). The literature factors and models are all based on Tinto's model, which Tinto and others have modified and improved over time. The model used in this study is the Geometric Model of Student Persistence and Achievement developed by Scott Swail (2003), which is an improvement over Tinto's model. Unlike other models, Swail's is student-centered; nevertheless, he also incorporates all of the other factors from other models. In Swail's Retaining Minority Students in Higher Education: A Frameworkfor Success (2003, p. 92), Figure 19 illustrates these factors and their components to illustrate how they relate to the student and persistence. Swail depicts the "Student Experience" as the center of an equilateral triangle. The triangle's base is "Institutional Factors," consisting of financial aid, student services, recruitment and admissions, academic services, and curriculum and instruction, the triangle's base. The left-hand side of the triangle is "Cognitive Factors," consisting of academic rigor, quality of learning, aptitude, 21 content knowledge, critical-thinking ability, technology ability, study skills, time management, and acedmic-related extracurricular activities. The right-hand side of the triangle is "Social Factors," consisting of financial issues, educational legacy, attitude toward learning, religious background, maturity, social coping skills, communication skills, attitude toward others, cultural values, expectations, goal commitment, family influence, peer influence, and social lifestyle. When all categories are in balance, persistence is most likely to happen. The Swail model has been used in several other research studies, either in its entirety or in modified form (EPI, 2007; Hayman, 2007; NCRA, 2006). For example, the Minority Engineering Recruitment and Retention Program at the University of Illinois at Chicago used Swail's model with modifications: Borrowing from Terenzini's (2006) model in EC 2000 and using variables in Swail's (2005) geometric model of student persistence and achievement, MERRP has developed a model of engineering student experience that identifies six background variables affecting an engineering student's experience in the academy, which in turn influence college matriculation outcomes (Hayman, D., 2007, p. 5). This study examines two types of retention: institutional-level and course-level. Two other types, major retention or system retention, will not be considered in this research. Institutional retention is a measure of students who remain at the same school year after year. System retention is a measure of students who remain in school, but without specifying that they continue in the same institution from one year to the next. Course retention is a measure of course completion. Major retention is a measure of specific major completion. Each measure has its own pitfalls: Institutional and system 22 level retention could be improved by including students who are part-time, transfer students, continuing education students, and all students regardless of their start date or fall cohort group. A shortcoming of Fall-to-Fall retention measures is that some institutions only admit students with high ACT or SAT scores in the fall semester, while admitting a second wave of students in the spring with lower ACT and SAT scores. Thus, institutions can admit students of better academic standing whose likelihood of retention is higher than for students admitted later, resulting in potentially inflated retention rates. Course retention is measured by a tool called Successful Course Completion Ratio (SCCR) (Hagedorn, 2004), the ratio of courses completed to courses taken. For example, a student who takes four courses and completes three of them has an SCCR of 75%. Institutions with high levels of student "stopouts," students enrolled in more than one institution, students who are not degree-seeking, or students with diverse academic goals all find this tool useful. It offers an alternative to the limitations of only examining degree-completion rates. Non-response Bias Reliable and valid techniques for measuring variables and constructs are the basis for all social science inquiries (Ary, Jacobs & Razavieh, 1996). Population parameters are estimated through various sampling procedures; the ability of researchers to generalize to broader populations hinges on these sampling procedures. Regardless of the quality of the sample and sampling technique, one major issue in survey research and evaluation is the declining response rate to surveys in the wealthier parts of the world (de 23 Leeuw and de Heer, 2002), which in some cases can produce non-response bias. Definition. Response rates measure all responses that are returned and usable as a proportion of all surveys distributed to the sample population. For example, 80 returned surveys out of 100 would give a response rate of 80%. If 20 of the returned surveys were incomplete or had other problems making them not usable, however, the response rate would fall to 60%. This might seem sufficient, but in fact might skew the sample. Non-response bias exists when there is a difference in the interpretation of results that would be made regarding those who respond and those who do not respond. "The bias created by non-response is a function of both the level of non-response and the extent to which non-respondents are different from respondents" (Kano, Franke, Afifi & Bourque, 2008, p.l). For example, in a study of gender issues with 50 men and 50 women in the population, 50 returned and usable surveys received might seem quite good if 25 were from men and 25 from women. If all 50 were from men and no women responded, however, the results are likely to be severely distorted: the resulting response bias would make the research or evaluation interpretations and conclusions invalid. A high response rate can be obtained, but may require costly follow-up procedures. Even with a high response rate, non-response bias may exist. Miller and Smith (1983) stated that even with a response rate as high as 90%, non-response bias may still be present. Dealing with Non-response and Non-response Bias. There are many ways to deal with non-response. These include ignoring non-respondents, following up with nonrespondents, comparing sample estimates of respondents to the population, comparing respondents to non-respondents, re-sampling non-respondents and comparing sample 24 estimates of respondents with other sources. Each method has advantages and disadvantages. Ignore non-respondents. Ignoring non-respondents means that the research can only be generalized to the sample responding and not to the population. This is not a good method of controlling for non-response error (Miller & Smith, 1983). Follow-up with non-respondents. A better alternative to ignoring them is to follow-up with non-respondents by sending out reminders, such as e-mails, postcards, or phone calls, or by redistributing the survey. "Two to three reminders (and even more) have proven effective" (Diem, 2004, p. 1). Compare Respondents with Non-respondents. Another method for controlling non-response error is to compare respondents with non-respondents (Miller & Smith, 1983). After a comparison on known characteristics shows no statistically significant difference, the results can be generalized both to the sample and the population (Diem, 2004). "While the level of non-response does not necessarily translate to bias, large differences in the response rates of subgroups serve as indicators that potential biases may exist" (Brick, Bose, W., and Bose, J., 2001, p. 2). A commonly used formula to calculate bias of the mean between respondents and non-respondents is: Ky r) = (l-r)( yr- ynr) where subscript r signifies the respondents while nr signifies the non-respondents. The notation 1- r is the non-response rate (Brick, Bose, W., and Bose, J., 2001) and y represents the mean of the chosen variable. This suggests that if the response rates for respondents and non-respondents are very different, the difference of the mean could 25 indicate bias. In some social science research this simple formula may not work due to the use of weighting, imputation of missing items, or any other non-response adjustments made to the data. In this study, however, no adjustments were made, and the formula functioned as intended. This formula can only be used when variables are known for both the respondents and non-respondents; academic records provided data so that variables for both groups were known, allowing the formula to function as intended. Compare Sample Estimates of Respondents to the Population. Another way to test for response bias is to compare sample estimates of respondents to that of the population "values computed from the sampling frame" (Brick, Bose, W., and Bose, J., 2001; Miller & Smith, 1983). With this method the problems of weighting and non-response adjustments are not an issue. Re-sampling ofNon-respondents. Diem (2004) describes another way to increase a low response rate: By taking another sample of "10 to 20 percent of the nonrespondents, and securing responses from this subsample, a statistical comparison can be made with subjects responding by the original deadlines and if they are similar, the data can be pooled and generalized to the sample/population" (Diem, 2004, p.2). This method is called "double-dip" by Miller and Smith (1983). This validation approach works, but two samples from the same population are needed. When a response rate of less than 80% is achieved Gall, Borg, and Gall (1996) suggest that a random sample of 20% nonrespondents be contacted and "double-dipped" and that responses from the nonrespondent sub-sample be compared with each item on the instrument to establish if nonresponse error is indicated. Compare Early to Late Respondents. Comparing early or on-time respondents 26 with late or reluctant respondents is commonly done in social science research to determine the effect, if any, of non-response on the statistics being considered (Miller & Smith, 1983; Smith, 1984). Extrapolation methods are used to compare early with late respondents. There are three types of extrapolation methods: successive waves, time trends and concurrent waves. Each method of extrapolation has its drawbacks, but all are based on the assumption that participants who respond later are more like nonrespondents. "Evidence has shown that late respondents are often similar to nonrespondents. If a statistical comparison of late respondents shows no difference from early respondents, then data from respondents can be generalized to the population" (Diem, 2004, p.2). Successive waves refers to the stimulus done over time, i.e. reminder emails, post cards, follow-up calls. It is assumed that subjects who respond in later waves responded because of the increased contacts made; therefore, they are expected to be similar to nonrespondents (Armstrong & Overton, 1977). Time trends refers to looking at the time between when the subject received notice of the survey or interview and his or her completion time. The drawback to this method is that it is sometimes difficult to know when the subjects were aware of the survey. If it is possible to determine when subjects became aware, subjects who respond later are assumed to be similar to non-respondents. The advantage to using a time trend method over the use of waves is the elimination of bias being introduced by the stimulus itself (Armstrong & Overton, 1977). Finally, concurrent waves refers to the same survey or stimulus being sent out to several randomly selected subsamples at the same time. "Wide variations are used in the inducements to ensure a wide range in rate of return among these subsamples. This 27 procedure allows for an extrapolation across the various subsamples to estimate the response for 100% rate of return" (Armstrong & Overton, 1977, p.2). The advantage of this method is that only one wave is needed from each subsample; therefore, an early cutoff date can be used. Compare Sample Estimates of Respondents with other Sources. Lastly, the existence of response bias can be established by comparing sample estimates of respondents with other sources, such as surveys, that ask similar questions (Brick, Bose, W., and Bose, J., 2001). Large differences could indicate bias, or at least suggest that further consideration is necessary. This method comes with many limitations, however, as survey items "may not be comparable because of coverage disparities, time periods that are not the same, differences in question wording, context effects and a host of other non-sampling error sources" (Brick, Bose, W., and Bose, J., 2001, p.4). Non-response Categories. In order to give the most accurate interpretation of the data, non-response can be broken into three categories: non-contacts, refusals and other. Non-contacts for this research are those students for whom contact information was either not available or not correct. Refusals for this research are those students who opted out of taking the online survey, which they could do by selecting the opt-out link instead of the link to the survey. Once a student opts out no further contact (such as reminders) is made. The third category, other, includes late respondents and those who did not respond at all. Detecting Response Bias. In Groves and Couper (1998) are found four figures that illustrate potential frequency distributions for non-respondents and respondents based on a hypothetical variable, y, measured on all cases in a hypothetical population. 28 Figure 1-1 a depicts conditions in which respondents and non-respondents are similar and there is a high response rate. If the response rate is 95%, which is extremely high, the mean for respondents is $201.00 and the mean for non-respondents is $228.00, then the non-response error is .05($201.00-$228.00) = -$1.35. Figure 1-lb depicts conditions in which respondents and non-respondents are not similar and there is a high response rate. If the response rate is again 95%, but the mean for respondents is $201.00 and the mean for non-respondents is $501.00, then the non-response error is .05($201.00$501.00) = -$15.00. Figure 1-lc depicts conditions in which respondents and nonrespondents are similar and there is a low response rate. If the response rate this time is 60%, the mean for respondents is $201.00 and the mean for non-respondents is $228.00, then the non-response error is .40($201.00-$228.00) = -$10.80 Lastly, Figure 1-ld depicts conditions in which respondents and non-respondents are not similar and there is a low response rate. If the response rate is again 60%, but this time the mean for respondents is $201.00 and the mean for non-respondents is $501.00, then the nonresponse error is .40($201.00-$501.00) = -$120.00. This means that the bias is 37% with regards to the total sample mean ($321, see Table 3) (Groves & Couper, 1998). Table 3 summarizes data from from Groves and Couper (1998) to demonstrate sample sizes of non-respondents needed with each of the four scenarios in order to determine a stable bias ratio. 29 Table 3. Bias and Percentage Bias in Respondent Mean Relative to Total Sample Mean NonResponse Response Respondent respondent Mean Rate Difference Rate % Mean Small 95 $201 $228 High Large High 95 $201 $501 Low Small $228 60 $201 Low Large 60 $201 $501 Total Sample Size $202 $216 $212 $321 Bias Percentage Bias -0.7 $1.35 $15.00 -6.9 $10.80 -5.1 $120.00 -37.4 Note. From Non-response in Household interview surveys, by R.M. Groves & M.P. Couper, 1998, New York: Wiley and Sons, p. 19. As shown in the section Detecting Response Bias, the formula to calculate the Bias term is as follows: Bias = 1 - Response Rate % (Respondent Mean - Non-respondent Mean) Ky r) = (\-r)( yr- ynr) For example, line four in Table 3 represents a low response rate, with a large difference between the respondents and non-respondents. Therefore, 1-60 (201 - 501) = -$120.00, is approximately a -37% bias, which is a large (over 10%) bias percentage. If it is believed that non-respondents are different from respondents in ways critical to the research or evaluation questions being asked, non-response bias should be examined thoroughly to make accurate generalizations of the population being examined. Because the Kalamazoo Promise Scholarship is such a new program and carries enormous implications for Kalamazoo students, the community and for other cities replicating the universal scholarship program, it is imperative to know that generalizations made of this population are not skewed by non-response error or possible non-response bias within the population. The purpose of this research is to consider this issue in depth by looking at not only the factors of retention affecting WMU's Kalamazoo Promise Scholarship recipients but also looking at non-response error to 30 determine if bias exists. Conclusion This chapter reviewed the relevant literature and research focused on the Kalamazoo Promise Scholarship and what impacts this universal scholarship has on the Kalamazoo community. Second, it examined a framework for analyzing higher education retention, exploring what factors assist successful retention of students. The last section covered relevant issues regarding non-response bias in social science research, including its definitions and implications. This framework is intended to help in understanding the purpose and results of this dissertation, namely retention factors for Western Michigan University Kalamazoo Promise recipients. It also enables the discussion of non-response bias in the survey research conducted for this dissertation. Understanding, or at a minimum, describing, retention factors of WMU Kalamazoo Promise recipients will facilitate future efforts to examine retention and factors that enable students to succeed in higher education institutions. 31 CHAPTER III METHODOLOGY The previous chapter explained the Kalamazoo Promise Scholarship and the impacts of this universal scholarship on the Kalamazoo community; presented a framework on higher education retention, exploring what factors assist successful retention of students; and discussed relevant issues regarding non-response bias in social science research, its definitions and implications. This chapter discusses the methods and procedures to be used in conducting this research. It is outlined as follows: (a) purpose, (b) research design, (c) population and sample, (d) procedure for data collection, (e) informed consent process, (f) research procedure, (g) data analysis, (h) ethical considerations, (i) limitations and (j) summary. Purpose The purpose of this study was twofold, centering on retention factors and nonresponse bias in the population of WMU Kalamazoo Promise recipients. It sought to determine to what extent persisters, those on probation, and non-persisters differ on demographic characteristics and on Cognitive, Social and Institutional retention factors as identified in Swail's (2003) Geometric Model of Student Persistence and Achievement. . As part of that inquiry, this study also examined the average courses taken per term and number of courses taken in the first year for persisters, those on probation, and nonpersisters, and for respondents and non-respondents. Additionally, the extent to which respondent, late respondent, and non-respondent Kalamazoo Promise recipients differed 32 on each of Swail's (2003) Geometric Model of Student Persistence and Achievement was compared with known characteristics from the academic data. Differences between early respondents and late respondents also were compared, this time using the actual survey data. Lastly, non-response error was examined using Groves and Couper's bias ratio formula (1998), using both the mean, as called for in the formula, and the median, a modification of the formula, to determine whether there was any indication of nonresponse bias in the survey data, and whether use of the more robust median produced different results than using the mean. The implications of these results could have an impact on decisions to budget (or not) extra funds to encourage a higher response rate, so as to ensure an unbiased representation of results in future studies of this population. Because the research concept of non-response bias is not widely used in evaluation, examining it here will also have important implications for evaluation and evaluation theory. Being aware of nonresponse bias and having the tools to apply it in order to achieve accurate results and make sound interpretations is imperative in research and evaluation alike. Research Design The research design used in this study consisted of a non-experimental relational design, using mixed methods. The research type is descriptive, which lends itself to detailed descriptions of a phenomenon. Descriptive research designs are compatible with the study of behavior and specific attributes of individuals. The study involved all Kalamazoo Promise recipients who are attending or who have attended Western Michigan University (WMU) since 2006, when the first Promise students entered WMU. 33 Some of these students are now juniors at WMU. For the first part of the research, participants were examined in three groups: those who are attending WMU and are in good standing (persisters, iV=200), those on probation (A/=51), and those who are no longer attending WMU (non-persisters, JV=49). For the non-response bias part of the research, participants are examined by respondents (7V=101) and non-respondents (iV=90). The research design involved the administration of an online survey to collect data. Along with the survey, in-depth interviews were conducted with a random sample of students. In addition, academic records were obtained from the Office of Student Academic and Institutional Research for all Kalamazoo Promise Recipients who have attended or are attending WMU. The purpose was to collect data from the sample of Kalamazoo Promise recipients associated with WMU and to be able to make generalizations about this particular population. Surveys were used because of the efficiency and low cost of this type of instrumentation. In-depth interviews were chosen because they offered detailed data. Academic records were used in order to compare respondents, late respondents and non-respondents on key variables to determine the possibility of non-response bias. Sample The population for this study consisted of all Kalamazoo Promise Scholarship recipients who are attending or have attended WMU since the beginning of the scholarship, a total of 307 students. Originally, data were obtained from two sources: the Facilitator of the WMU Kalamazoo Promise and the Office of Student Academic and Institutional Research. Initially, contact names of all WMU Kalamazoo Promise 34 recipients were obtained from the Facilitator, who listed 191 students. After the surveys were returned, these 191 coded names were given to Institutional research to add academic data to the existing data already collected. This procedure was used to comply with the Family Educational Rights and Privacy Act (FERPA). This breakdown is detailed in Table 4. Table 4. Data Obtained from the Facilitator of the WMU Kalamazoo Promise and the Office of Student Academic and Institutional Research Facilitator Institutional Research Persisters 155 200 On Probation 0 51 Non-persisters 36 49 Total 191 307 The breakdown of the population of students by persistence and type of data obtained is detailed in Table 5. Surveys were sent to 191 students, the number initially provided by the Facilitator of the Kalamazoo Promise at WMU. Of these 191 students, 101 responded and 90 did not respond. Of the 72 interviews planned and requested, only 14 agreed to an interview. Academic records and course data were obtained from the Office of Student Academic and Institutional Research on all 307 WMU Promise students. Table 5. Breakdown by Type of Data Collected and Persistence Survey Respondents (191 sent) Interviews Academic Records/Course Data Persisters On Probation Nonpersisters 87 \2 10 2 4 0 101 14 200 51 49 300 Note. N=300 WMU Kalamazoo Promise Recipients. Seven did not have probation status listed. Surveys were sent to 191 students, 101 responded, 90 did not respond. 35 Procedures for Data Collection Subject Recruitment. The subject selection initially was obtained from Patricia Williams, the facilitator of the Kalamazoo Promise support program at Western Michigan University (WMU). Patricia Williams had access to all Kalamazoo Promise students at WMU for her job. She provided the researcher with the names and current contact information of all Kalamazoo Promise recipients who are attending or who have attended WMU to which she had access. Survey. The entire population of Kalamazoo Promise recipients known from the facilitator of the Kalamazoo Promise (7V=191) who are attending or who have attended WMU was asked to participate in a survey titled: Survey of Promise Scholarship Recipients at WMU Spring 2009. This survey was developed by the researcher based on the three factors of Swail's (2003) Geometric model of student persistence and achievement. A copy of the survey questions sorted by the three factors can be found appendix A. The Promise recipients were contacted through an initial e-mail, Introduction Survey E-mail protocol (see Appendix A). Interview. In addition to the survey of the population of WMU's Kalamazoo Promise scholarship recipients, two samples of recipients were invited to learn more about this study. Those who responded and then consented to do so were invited to participate in an in-depth interview. The first sample was all of the Kalamazoo Promise scholarship recipients no longer attending WMU, iV=36. The second sample (JV=36) was a random sample of Kalamazoo Promise scholarship recipients still at WMU. This random sampling 36 occurred by using the contact list of names from the facilitator of the Kalamazoo Promise scholarship program. Names were randomized so they were not alphabetical and every seventh person on the list was chosen until there were 36 names in the random sample. All 72 students in the two samples were contacted via e-mail by the researcher. The e-mail (Interview E-mail Invitation, appendix A) explained the research and asked students, the potential participants, if they were willing to learn more about the project. Those who wanted to learn more about the project were instructed to contact the researcher to set up a time to meet. Once a student contacted the researcher, the researcher contacted the student by phone and arranged a convenient time to meet (Interview Phone Invitation in appendix A). Academic Records. In addition to the survey and interviews, academic records were obtained from all Kalamazoo Promise Scholarship recipients who are attending WMU or who attended WMU in the past (see Table 5). This collection was done after the initial collection of data using the survey in order to link academic data to the survey data without using names. In order to meet the Family Educational Rights and Privacy Act of 1974 (FERPA or the Buckley Amendment) guidelines, no academic records had any identifying information on them (such as student names or social security numbers); therefore a strict coding system was enforced. The data from each sample was compared to determine what differences, if any, there were between recipients WMU retained versus those WMU did not retain. In addition, the question of non-response error was examined by comparing : students who responded to the online survey with those who did not respond in order to see if there was an indication of non-response bias associated with this study. 37 Informed Consent Process There were two separate consent processes, one for the survey and one for the interview. The following section is broken into a consent process for the survey and another consent process for the interview. Survey Consent Process. Students were contacted by e-mail (Introduction Survey E-mail Protocol appendix A) which explained the research and asked the student if he or she were willing to participate. A student who wanted to participate was instructed to click on the link to the Survey Monkey online survey (see appendix A) to read more information about the study and read the informed consent (see appendix A). There was a check box for participants to click if they wanted to proceed, which constituted their consent to use the information they provided for the dissertation research. Participants who did not click the box to confirm they had read and were giving their consent were not able to proceed to the survey. Interview Consent Process. Students were contacted by e-mail (Interview E-mail Invitation, appendix A) which explained the research and asked students, the potential participants, if they were willing to learn more about the project. Those who wanted to learn more about the project were instructed to contact the researcher to set up a time to meet. Once a student contacted the researcher, she contacted the student by phone, and arranged a time that worked for the individual to meet (Interview Phone Invitation, appendix A). At the meeting the consent document was reviewed, with the option to continue and sign the document or discontinue. If the potential participants chose to sign the consent form and agreed to participate in an interview, the interview took place at that 38 time. No one participated in the interview until the consent document had been thoroughly reviewed, questions answered, and the consent document signed. Academic Records. No consent was needed as all records are de-identified, meaning there were no student names attached to this information. Research Procedure Method of Data Collection. Data were collected through an online survey, an interview, or both, as well as from academic records. The survey was sent out first and the interview process started. After this phase had been completed the office of Institutional Research at WMU was contacted to obtain academic records grouped aggregately by categories. No student names were linked to this data. Survey. E-mail addresses were then put into Survey Monkey to enable distribution of the survey electronically. This also allowed follow up e-mails to only go to those who had not yet completed a survey. Only the primary researcher had access to this information. An e-mail was sent to all Kalamazoo Promise recipients with a link to the Survey Monkey survey. (See appendix A for Introduction Survey Email Protocol). Non-response was monitored automatically in Survey Monkey through the e-mail addresses, and non-respondents were sent a reminder email (First Reminder E-mail to Take Survey see appendix A) regarding the survey one week after the survey was emailed. A second e-mail reminder (Second Reminder E-mail to Take Survey appendix A) was sent three days later to those who had not yet filled out the survey. Any survey that came in after this second reminder was considered a late respondent. This nonresponse data was compared to the response data (data that arrived before the second 39 reminder). The survey stayed open until the end of April 2009. At the end of the electronic survey, participants were offered the opportunity to enter a random drawing with a chance to receive one often $20 WMU bookstore gift cards. Participants needed to provide their name and e-mail address, which was not linked to their survey results, to enter the drawing. Interview. Two samples were invited to learn more about the study. They were contacted first through e-mail to invite them to learn more about the study (Interview Email Invitation, appendix A). A week later they were contacted by e-mail again to invite them to participate in an interview if they agreed to participate (Interview E-mail Invitation appendix A). The interview invitation was separate from the survey invitation because of the preference for taking a random sample of Kalamazoo Promise scholarship recipients and not just those who answered the survey. That these respondents might be different was understood and was to be investigated for the non-response bias part of this research. It was hoped that the researcher would be able to start setting up interviews right away. An e-mail reminder of the date, time and location of the interview was sent a few days before the interview, and a reminder call was made the day before the interview. A copy of this E-mail Reminder of Interview and Reminder Phone Call of Interview can be found in the appendix. The interview did not last more than one hour. Participants were given a copy of the study information sheet and the consent form to keep (see appendix A). Both forms were discussed. One copy of the consent form was signed by the participant and kept by the researcher; the other copy, along with the study information sheet, was the participant's to keep. The interview was electronically recorded with a 40 digital audio recorder and transcribed later. After the interview was completed, all participants were given a $20 WMU bookstore gift card as a thank you for their participation. All interviews were completed by the researcher, her business partner, Nakia James, or Katya L. Gallegos Custode and Tammy DeRoo, both graduate research assistants at the College of Education working with Dr. Miron. Both the researcher and Nakia James are partners with Momentum Consulting and Evaluations, L.L.C. and are experienced in interviewing and proficient in working together on extensive projects. All interviewers took and passed the Human Subject Institutional Review Board exam at Western Michigan University. Academic Records. Academic records were obtained through the office of Institutional Research at WMU. All records were de-identified, meaning there were no student names on this information. Records were grouped only by persisters, those on probation, non-persisters, respondents and non-respondents. Instrumentation. The survey and the interview protocol (see appendix A) were developed from research-based factors found in the literature review, principally Swail's Geometric Model, and modified from other validated surveys, including the National Survey of Student Engagement (NSSE), the Community College Survey of Student Engagement (CCSSE), and the College Student Experience Questionnaire (CSEQ). In addition, a separate reliability analysis was run on the newly created survey by each retention factor: cognitive, social and institutional. Items for the continuous/interval level data from the cognitive, social, and institutional factors included from the Survey of Promise Scholarship Recipients at WMU 41 Spring 2009 were grouped to create the following four summated subscales: cognitive engagement, social demands, institutional support, and social engagement. The exact details of each of the scales can be found in Appendix C. The Cognitive Engagement Subscale consisted of 20 items (Items C7 through C27) from the Cognitive Factor of the survey. Participants responded to items on the Cognitive Engagement subscale using a 5-point Likert-type scale where the responses ranged from 1 = not likely to 5 = very likely. The Social Demands subscale consisted of five items (Items S13 through SI7) from the Social Factor of the survey. Participants responded to items on the Social Demands subscale using a 5- point Likert-type scale, where the responses ranged from 1 = never to 5 = often. The Institutional Support subscale consisted of six items (Items 17 through 113) from the Institutional Factor of the survey. Participants responded to items on the Institutional Support subscale using a 5-point Likert-type scale, where the responses ranged from 1 = never to 5 = very often. The Social Engagement subscale consisted of five items (Items 114 through 118) from the Institutional Factor of the Survey. Participants responded to items on the Social Engagement subscale using a 5- point Likert-type scale, where the responses ranged from 1 = very little to 5 = very much. Summated scales offer an advantage over single-item scales in that such scales can be assessed for reliability and for the unidimensionality of the construct being measured (Thorndike, 1967). Items assigned to each scale were summated together to yield total scores. Before running statistical procedures on data from the survey, the 42 researcher assessed the internal consistency of the scales using reliability analysis. Cronbach's coefficient alpha was used to measure the internal consistency of the scales included in the survey (Cohen, 1988; Trochim, 2007). While any test developer hopes to obtain a reliability coefficient that approaches 1.0, such a value is rarely obtained in behavioral and social science research. The significance of the obtained alphas will be tested against the value of alpha = .70, as suggested by Kaplan and Saccuzzo (2005), for the obtained alpha coefficients. The research indicates that values of .70 or greater indicated that a scale is internally consistent (Kaplan & Saccuzzo, 2005; Mertler & Vanatta, 2005). Table 25 presents a summary of the descriptive statistics for the four scales. Location ofData Collection. The location of data collection for the survey happened online. It was the participants' choice where they took this online survey: in their own homes or wherever else they felt comfortable taking it. The location of data collection for the interview took place in conference room 1411 Sangren Hall or in the Multicultural Affairs conference room in Trimpe Hall. Both locations were private but convenient for WMU students. Duration of the Study. The study lasted from the beginning of April until the end of June 2009. Survey respondents took approximately 20 minutes to fill out the survey online. Interview respondents took no more than one hour. All surveying was completed by the end of April, and interviews were completed by mid-May. Data Analysis The purpose of this study was to determine to what extent persisters, those on 43 probation, and non-persisters differ on demographic characteristics, and on each of the following select factors in Swail's (2003) Geometric Model of Student Persistence and Achievement: Cognitive, Social and Institutional retention factors. This study also examined the average number of courses taken per term and number of courses taken the first year of persisters, those on probation, and non-persisters, and of respondents and non-respondents. Additionally, the extent to which respondent, late respondent and nonrespondent Kalamazoo Promise recipients differ on each of Swail's (2003) Geometric Model of Student Persistence and Achievement was examined against known characteristics from the academic data. Differences between early respondents and late respondents also were examined, using the actual survey data. Lastly, non-response error was examined using Groves and Couper's bias ratio formula (1998), using first the mean, as called for in the formula, and then the median, a modification of the formula, to determine whether there was any indication of non-response bias in the survey data and whether using the mean estimator produced different results than using the median estimator. Descriptive and inferential statistics methods, including frequency tables, Multivariate Analysis of Variance analyses (MANOVA) and Chi-square Test of Independence analysis, were used along with the bias formula in order to accomplish this task. MANOVA is a parametric statistical procedure used to measure differences in scores between at least two groups (Mertler & Vanatta, 2007). As such the MANOVA is preferred over the use of several univariate ANOVAs for the following reasons: a) several dependent variables can be assessed simultaneously; b) results may be obtained that may not be detected in univariate tests, such as interaction effects among variables; 44 c) it reduces the overall likelihood of Type I error rate by statistically maintaining the overall rate at the level determined by the researcher. As a parametric statistical procedure, the following assumptions apply to the data (Kilpatrick & Feeney, 2007; Sprinthall, 2007): a) independence, b) normality, and c) homoscedascity. Independence of scores. The researcher assures independence of scores at the outset of participant selection. MANOVA is not very sensitive to this violation, but it must be addressed (Kilpatrick & Feeney, 2007). Because the participants submitted the surveys anonymously online, the researcher presumed that participants completed the surveys independently without the assistance or input from other participants. Multivariate normality. This assumption posits that each groups' patterns of scores should reflect the shape of the normal distribution (Hair, Anderson, Tatham & Black, 1995). The Kolmogorov-Smirnov Test Statistic was used to test this assumption (Hair, et al., 1995; Kilpatrick & Feeney, 2007). Separate test statistics were computed for each dependent variable. A summary of the results is presented in Appendix D, which indicate that the assumption of normality was violated for several variables. This violation, however, does not affect the analysis run as the F-test is robust and violations of the assumptions of normality have minimal effect under certain conditions (Creswell, 2005), including the conditions of this non-experimental relational design. Homogeneity of variance. This assumption posits that there must be equal variances between groups. The Levene Test Statistic (Kilpatrick & Feeney, 2007) was used to test this assumption. Separate test statistics were computed for each dependent variable. A summary of the results is presented in Appendix E. Results indicate the assumption was upheld for all of the academic variables except composite ACT scores. 45 The assumption was also upheld for all four subscale scores of the Survey of Promise Scholarship Recipients at WMU Spring 2009. Interpretations of the assumptions. The assumptions of normality and homogeneity of variance are most critical in the case of experimental research designs (Creswell, 2005). The research for this study was a non-experimental relational design. While the assumptions of normality and homogeneity of variance were not upheld for the data in this study, research suggests that violations of the assumption of normality have little effect under certain conditions. Creswell (2005) also states that the F-test is robust and violations of the assumptions of normality and homogeneity of variance have minimal effect under certain conditions. Specifically, Creswell (2005) states that if the larger group variance is no more than four times the smallest group variance, then violations of the assumption of homogeneity of variance will have minimal effect on the results of the MANOVA procedure. The review of the descriptive statistics for the groups across the dependent variables revealed no cases in which the 4-to-l ratio for the differences in group variances was upheld. Consequently, the researcher deemed that the violations of the assumptions were acceptable considering the exploratory nature of the research. The Chi-square test of independence. This statistical procedure measures the degree to which a sample of data comes from a population with a specific distribution (Mertler &Vanatta, 2007; Rosenberg, 2007; Stevenson, 2007). It tests whether the observed frequency count of a distribution of scores fits the theoretical distribution of scores. This issue was addressed through the use of the Pearson's Chi-square {% ) procedure (Mertler & Vanatta, 2007; Rosenberg, 2007). A non-significant finding 46 indicates no statistically significant differences between the observed and expected frequencies on the variables of interest. Research Questions 1 and 2 All research questions and sub-questions with corresponding variables and analysis are detailed in Table 6. To determine whether differences exist on the selected factors among persisters, those on probation, and non-persisters, by gender and race,, Chi-square Test of independence analysis and MANOVA analysis were applied using the academic data provided by the Office of Student Academic and Institutional Research. In order to examine persisters, those on probation, and non-persisters on average number of courses taken per term and number of course taken the first year, by gender and race, ANCOVA analyses were employed using the course summary data provided by the Office of Student Academic and Institutional Research. First, however, two new variables were created manually. The average number of courses taken per student per term was calculated by adding up all of the fall and spring courses taken per student and dividing by the number of fall and spring terms the student attended. The first-year courses were calculated by adding the courses from the first fall and spring term that students attended WMU. In order to examine respondents, late respondents, and non-respondents of the Survey of Promise Scholarship Recipients at WMU Spring 2009 in average number of courses taken per term and number of course taken the first year, by gender and race, ANCOVA analyses were employed using the course summary data provided by the Office of Student Academic and Institutional Research. The Survey of Promise 47 Scholarship Recipients at WMU Spring 2009 was used only to distinguish between respondents, late respondents, and non-respondents for this question. Research Question 3 All research sub-questions with corresponding variables and analysis is detailed in Table 7. To determine if possible non-response bias exists, several procedures were undertaken. The first was determining whether there was a statistically significant difference between respondents («=52), late respondents («=49), and non-respondents («=84) using known variables of each group from the academic data set (see Table 15). This is one of the suggested ways to determine if bias is a possibility. "After examining the respondents with the non-respondents on known characteristics, if no statistically significant difference is found, then the results can be generalized both to the sample and the population" (Diem, 2004, p. 2). In addition, variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009 were used to compare early respondents and late respondents. This type of comparison is commonly used in social science research to determine an effect, if any, on the statistic being considered when examining nonresponse and possible bias (Miller and Smith, 1983; Smith, 1984). Next the means from each variable were plugged into Groves and Couper's bias ratio formula (1998), which includes the mean for respondents and non-respondents or, in this case, late respondents. Lastly, the median from each variable was plugged into a modified version of Groves and Couper's bias ratio formula (1998), replacing the mean, because the median is a more robust estimate than the mean. When population samples are small, such is the case in this research, the mean is sensitive to extreme scores; the median—the point where half 48 the scores fall below and half above—is less sensitive to extreme scores. If no statistical difference is found among respondents, late respondents and nonrespondents using the academic data set, and no indication of bias exists using the mean or the median calculations using the survey data on the respondents and late respondents, then the results from the survey data can be generalized to the population of Kalamazoo Promise Students across cognitive, social and institutional factors of retention without concern of non-response bias. If an indication of bias is found, the results of the survey data can only speak for the respondents of the survey and not for those who did not respond. Table 6 and 7 organize all of the research questions with their dependent and independent variables along with the source of the data and method of analysis. 49 Table 6. Summary Research Questions 1 and 2 with Independent and Dependent Variables, Data Source and Method of Analysis Research Question Independent Variables Dependent Variables Data Source Method of Analysis 1.1 Difference in Cognitive Factors Persisters = Good Standing On Probation and Nonpersister = Academically dismissed Gender (M,F) Race (White, Black, Other) High school GPA, most recent WMU GPA, ACT composite score Academic Records MANOVA Taking remedial math, reading or writing or taking AP credit Living in a dorm, being and athlete Parental income Academic Records Chi-Square Academic Records Academic Records Chi-Square 1.2 Difference in Social Factors 1.3 Difference in Institutional Factors 2.1 Difference in Persistence 2.2 Difference in Respondents Persisters = Good Standing On Probation and Nonpersister = Academically dismissed Gender (M,F) Race (White, Black, Other) Persisters = Good Standing On Probation and Nonpersister = Academically dismissed Gender (M,F) Race (White, Black, Other) MANOVA First year experience and high school Academic Records Chi-Square Average number of Persisters = Good courses taken per Standing On Probation and Non- term and number of persister = course taken the first Academically dismissed year. Gender (M,F) Race (White, Black, Other) Average number of Respondent, Late & courses taken per Non-respondent term and number of Gender (M,F) Race (White, Black, course taken the first Other) year. Course Summary ANCOVA Gender and race (covariates) Course Summary ANCOVA Gender and race (covariates) 50 Table 7. Summary Research Question 3 with Independent and Dependent Variables, Data Source and Method of Analysis Research Question Independent Variables Dependent Variables Data Source Method of Analysis 3.1 Differences among respondents, late respondents and nonrespondents on dependent variables from academic data Respondents, late respondents and nonrespondent Academic Records MANOVA Academic Records Chi-Square 3.2 Differences between early and late respondents on dependent variables from academic data Early respondent and late respondent Survey MANOVA for subscales; Chi-Square for categorical level data 3.3 Examining indication of non-response bias using bias ratio formula on subscales from survey Early respondent and late respondent (a) high school GPA, (b) most recent WMU GPA, (c) ACT composite score, (d) parental income (e) taking a remedial math course at WMU, (f) taking a remedial reading course at WMU, (g) taking a remedial writing course at WMU, (h) taking AP credit, (i) living in a dorm, (j) being an athlete, or (k) first year experience (FYE) or (1) high school Cognitive, Social and Institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement (See factors/questions table in Appendix) (See factors/questions table in Appendix) Survey Bias Ratio Formula 51 Ethical Considerations Every effort was made to ensure the ethical treatment of all participants. The researcher complied with all standards of the Human Subject Institutional Review Board (HSIRB). It was the responsibility of the researcher to revere the needs and rights of all participants (Locke, Spirduso & Silverman, 2000). The process of informed consent was followed to achieve this. The researcher: (1) acquired written permission of Western Michigan University's Human Subjects Institutional Review Board (HSIRB); (2) clearly explained the objectives of the study in writing to the survey participants and in writing and orally to the interview participants; (3) acquired consent from each interview participant with the consent form found in Appendix A, and from each survey participant with the same consent form, with the survey participants placing a check mark indicating their consent after reading the form in Survey Monkey, an online survey program; and (4) gave each interview participant a $20 WMU book store gift card as a thank you for participating. Risks and Costs to and Protections for Subjects. There were no known anticipated physical, psychological, social or economic risks to the participants. Participants in the survey could complete it in the comfort and privacy of their own homes or wherever was convenient for them. Interview participants who did not live on campus or were no longer attending WMU may have been inconvenienced because interviews were held on campus. For those living on campus or who had class on campus, arrangements were made to schedule the interview at a time most convenient for them. Regardless of whether or not the participant lived on or off campus, the interview 52 was scheduled at a time that was most convenient for him or her. Also, a $20 WMU book store gift card was given for participating in the interview and, it is hoped, offset any inconvenience. Interviews were conducted at each student's convenience; therefore, there was no disruption to any class or administrative function. The interviews took place in a conference room in either Sangren Hall or Trimpe Hall on campus to ensure a neutral environment as well as to take advantage of the excellent audio capabilities. Interviews were digitally recorded. Benefits of Research. This research may benefit institutions working on their own retention issues or those that would like to replicate the Kalamazoo Promise Scholarship. It may also shed some light on possible non-response bias that exists in social science research and evaluation. The knowledge base of the Kalamazoo Promise, college retention, persistence and non-response bias was expanded upon. The participants in the research may have benefited from the study through selfreflection on their academic journey. Since the Kalamazoo Promise Scholarship is free to these students, they may have felt good about "giving back" to help this program succeed. Participants may have also learned more about the Kalamazoo Promise. For example, one question asked if participants were aware that they could still use their Kalamazoo Promise Scholarship even if they took time off from college. Participants not aware of this benefit may have found such information helpful. Confidentiality of Data. Every effort was made to keep all information confidential. Survey Data. Survey data was only linked by e-mail address and was seen only 53 by the researcher and was kept confidential. Students' names were never used. Interview Data. Interview data initially had a student's name attached to it. In order to make the interviewee comfortable the participant was referred to by name during the signing of the consent form. They were told, however, that once the recording device was turned on their name would not be used and that a code number would be given for the transcription of the interview. In that way their names were kept confidential even in the transcription. Consent forms, names and code numbers were only seen by the researcher. It was anticipated that there would be approximately 72 interviews, with the code consisting of numbers 1 through 72. All interview data was reported without using student names. The digital recorders were kept in a locked filing cabinet in the researcher's home until transcribed. No one except the researcher had access to these recorders. Once transcription was completed the digital interview recordings were deleted immediately and no record of the audio files was kept. The transcription was kept in a locked filing cabinet in the researcher's home. The consent forms were kept in a manila envelope in a locked filing cabinet in Dr. Miron's office to which no one except Dr. Miron has access. After the study's completion, the data continues to be stored at WMU in Dr. Miron's office in a locked filing cabinet. In three years these consent forms, transcriptions and data will be disposed of by burning. The doctoral researcher does not work at WMU, has no way to know who these students are, and has no ability to make any decisions regarding their academic performance at WMU. Anyone who reads this dissertation will have no way of knowing who participated other than that participants are recipients of the Kalamazoo Promise Scholarship. Since there are 307 who have attended or attend WMU, it would be 54 impossible to link any aggregate data to any individual student. Confidentiality is assured. Academic Records. Confidentiality was not an issue for academic records as no student names are linked to any of this data. Limitations The limitations of this research, first and foremost, are due to the limited sample size. Ideally, all Kalamazoo Promise recipients would be included in such a study. However, because these recipients have chosen to go to 26 different higher education institutions (Jorth, 2009), it was impossible to include all Promise recipients within the scope of this project. Therefore, the scope of this project was limited to only those Kalamazoo Promise Scholarship recipients who attended Western Michigan University at some point since 2006, when the first recipients attended their first semester at a higher education institution. The entire population of Western Michigan University Kalamazoo Promise recipients was included in this project. Only 4 out of 13, or 31%, of those WMU students who were academically dismissed and therefore are no longer at WMU responded to the survey. This data came from the responses and non-responses of the Survey ofKalamazoo Promise Recipients at WMU Spring 2009 (147 in Good Standing, 25 on Probation, and 13 Academically Dismissed, 6 no probation status listed, 191 total surveyed) and from the academic data obtained from the Office of Student Academic and Institutional Research. This response rate of 31% is considered very low. Considering the nature of this research and the information needed directly from these students to examine differences between 55 persisters, those on probation, and non-persisters, it is essential to take note of this response rate. Generalizations can only be made to the sample that responded. The goal here was to generalize to the population of WMU Kalamazoo Promise recipients, and in order to accomplish this non-response bias was examined in depth. Summary This chapter has provided an overview of the methods and procedures used in compiling this mixed methods research project, including the ethical treatment of all participants. Mixed methods were used based on the needs of the research questions and the types of data used to answer these questions. Using mixed methods allowed the research questions to be answered and not restricted based on the needs of the approach used. "Mixed methods research is formally defined here as the class of research where the researcher mixes or combines quantitative and qualitative research techniques, methods, approaches, concepts or language into a single study" (Johnson & Onwuegbuzie, 2004, p. 17). The research questions lent themselves to each method. Chapter IV presents the findings from the analysis of these data sources. 56 CHAPTER IV RESULTS Due to the complexity of the research questions, the results have been broken into five sections. The initial results are a summary of the academic data received from WMU's Office of Student Academic and Institutional Research, which are included in this first section of the results under the heading Summary Academic Data. The second section is a summary of the Survey of Promise Scholarship Recipients at WMU Spring 2009 under the heading Survey Summary. Each of the other sections was broken up by research question. The research sub-questions are then each answered separately using the academic data, which includes the course summary data and the survey data as needed. Summary Academic Data The summary academic data is intended to give a brief overview of the study's population in terms of students' probation status, gender, race, and survey participation. This data was obtained on all 307 students who have received the Kalamazoo Promise and are attending or have attended WMU. This background provides a description of the sample involved in this study. The socio-demographic characteristics of these 307 Western Michigan Promise scholarship recipients can be seen in Table 8, delineated by whether these students have persisted (still attend), are on probation or have not persisted, meaning that they were academically dismissed from WMU. 57 Table 8. Socio-Demographic Characteristics of WMU Kalamazoo Promise Recipients by Persistence Persister n/% On NonProbation Persister n/% n/% Total by Population n/% % NonPersist within Group3 Male Female 102/34 98/33 29/10 22/7 32/11 17/6 163/53 137/46 19.6 12.4 White Black Asian Hispanic American 130/43 35/12 14/5 8/3 29/10 13/4 1/.3 5/2 24/8 17/6 1/.3 3/1 183/61 65/22 16/5 16/5 13.8 26.2 6.3 18.8 Unknown 21.1 11/4 1/.3 21.1 0/0 4/1 3/1 17/6 0 23.5 Gender Race Indian Note. N=300, not 307 as seven students had no probation status listed. Total percent may be off because of rounding. a % non-persistence within group, male = 32/163= 19.6%. Sources: From academic data from the Office of Student Academic and Institutional Research. Results show that males represented the largest percentage (53.4%) in the sample. In terms of race, the initial data analysis revealed that there were too few participants in the Hispanic, Asian, Native American, and Unknown categories to conduct meaningful statistical comparisons. Therefore the researcher collapsed the responses for those four groups into one group, which was labeled Other. The summary academic data is reported through several different variables; Probation, Group, and Group by Response. The variable Probation had five levels; Good Standing, Academic Dismissal, Extended Probation, Final Probation, Probation, sad Academic Warning. The Probation status variable was collapsed into three levels by combining Extended Probation, Final Probation, Probation, and Academic Warning into a new category, On Probation, after this initial description. 58 Initially, 191 Promise recipients were identified with the help of the WMU Kalamazoo Promise Scholarship facilitator . The 191 were given the opportunity to complete the Survey of Promise Scholarship Recipients at WMU Spring 2009. However, WMU's Office of Student Academic and Institutional Research had 307 Promise students on record as attending or having attended WMU. Of those, 7 had no record of probation status and were excluded from the analysis. Of the remaining 300, the overall percentage of persisters was 66.66%, with another 17% on probation. There are 16% who have been academically dismissed and have therefore not persisted. Table 9 gives an exact count of students in each probation status group with their average high school GPA. Table 9. Average High School GPA by Probation Status at WMU Probation Status Persisters Good Standing On Extended Probation Probation Final Probation Probation Academic Warning NonAcademic persisters Dismissal Total N Mean SD 200 3.496 0.762 95% Confidence Interval Lower Upper Bound Bound 3.390 3.602 6 2.920 0.361 2.541 3.299 2.43 3.46 1 2.990 . . . 2.99 2.99 23 21 3.163 3.209 0.477 0.422 2.957 3.016 3.369 3.401 2.36 2.57 3.98 3.85 49 3.044 0.578 2.878 3.210 0.00 3.89 300 3.363 0.713 3.282 3.444 0.00 4.65 Min Max 0.00 4.65 Note. Seven students had no record of probation status and were not included in this analysis, therefore N= 300 instead of 307. Academic Dismissal = Non-persister. An exploratory analysis using an Analysis of Variance (ANOVA), not surprisingly, found the mean high school GPA based on probation status at WMU to be 59 statistically different [F (5, 294) = 4.711, p =0.000]. The source of this difference was not determined, however. Looking at the same students again grouped by probation status, similar results can be seen with these students' most recent WMU GPA (see Table 10). Table 10. Most Recent WMU GPA by Probation Status at WMU Persisters Probation Status N Mean SD Good Standing 20 0 3.0575 0.5011 95% Confidence Interval Lower Upper Bound Bound 2.9877 3.1274 6 1.7500 0.1249 1.6189 1.8811 1.59 1.94 1 1.9000 . . . 1.90 1.90 23 21 1.0735 2.1943 0.7647 0.1877 0.7428 2.1088 1.4041 0.00 2.2797 2.00 1.98 2.63 49 1.1371 0.5632 0.9754 1.2989 0.00 1.99 30 0 2.5013 0.9755 2.3905 2.6122 4.00 On Extended Probation Probation Final Probation Probation Academic Warning NonAcademic persisters Dismissal Total Min Max 2.06 4.00 0.00 Note. Seven students had no record of probation status and were not included in this analysis, therefore N= 300 instead of 307. Academic Dismissal = Non-persister. An exploratory analysis using ANOVA found that the most recent WMU GPA mean based on probation status was also statistically different [F (5, 294) = 153.391,/? =0.000]. The source of this difference was not determined, however. Students who had low high school GPAs also had low WMU GPAs. Unlike high school GPAs, which regardless of probation status clustered around 3.00, for those WMU students in any of the probation groups, most recent GPAs dropped from the 3.00s into the 1.00s and low 2.00s. 60 Because the exploratory analyses suggest that there are differences, before examining this any further the probation status variable was collapsed into three categories: Good Standing, On Probation, and Academic Dismissal. Extended Probation, Final Probation, Probation, and Academic Warning were all put under one category, On Probation, in order to simplify the analyses from here forward. This data was used not only to answer the research questions but also in an indepth look at non-response error, which can sometimes lead to non-response bias. Table 11 summarizes when each of the 307 Promise students started at WMU, while Table 12 reports the exact incoming first-time, full-time, degree-seeking beginners (FTIAC) cohort counts by persistence, on probation, and non-persistence. FTIAC cohorts are used at WMU when examining retention. Table 11. First Promise Semester 2006 Summer II 2006 Fall 2007 Spring 2007 Fall 2008 Spring 2008 Summer II 2008 Fall 2009 Spring Total Frequency Percentage 1 100 5 90 5 4 100 2 0.3 32.6 1.6 29.3 1.6 1.3 32.6 0.7 307 100.0 It would be suspected that the first FTIAC Cohort group, matriculating in 2006, would have more students on probation or academically dismissed than subsequent year cohort groups. That is because the First Year Experience (FYE), which provided supports 61 for first-year students was just introduced in 2005 and had not officially started until the Fall of 2006. The First Year Experience should be an effective mediator of dropping out and therefore help retain more students; thus, it would not be surprising to see a decrease in dropping out or an increase in retention in the later Cohorts after the FYE was in place. Chi-square analysis reveals, however, that there was no statistically significant difference \X2(6, 300) = 10.139, p = 0.119] between FTIAC Cohort groups and Probation status, which means that the same proportion of students from each FTIAC Cohort group have been academically dismissed or are in good standing. Data was therefore run aggregately and not by FTIAC Cohort. Aggregation of these cohort groups, along with being able to use the 27 students who are not in a cohort, in turn gave more power to the analysis due to the increase in group size. Table 12. FTIAC Cohort by Persistence and Non-Persistence 2006 Fall 2007 Fall 2008 Fall Not in a Cohort Total Persisters n/% On Probation n/% 68/23 56/19 53/18 23/8 200/67 15/5 12/4 23/8 1/.3 51/17 Non-Persisters n/% 17/6 15/5 14/5 3/1 49/16 Note. Af=300, seven students had no probation status listed. About 98% of the WMU Promise students came from Loy Norrix High School (139) and Kalamazoo Central High School (164), with only one student coming from Galesburg to attend WMU. In addition, two students did not have a high school listed on record. 62 Table 13. High School Attended by WMU Kalamazoo Promise Recipients Percentage 45.3 53.4 0.3 0.7 100.0 Frequency Loy Norrix High School Kalamazoo Central High School Galesburg Augusta High School Not Reported Total 139 164 1 2 307 Table 14. Distribution of Promise Students by Race, Gender and High School Kalamazoo Unknown White Black Asian Hispanic American Indian Total Loy Norrix Central (« =139 ) (n =164) Galesburg (B=1) Male Female Male Femal e Male 3 47 11 3 5 0 69 3 45 14 5 3 0 70 7 59 17 5 4 1 93 3 38 21 3 4 2 71 0 0 0 0 0 0 0 Femal e 1 0 0 0 0 0 1 Note. N= 304. One black female and one black male had no high school reported and are therefore not calculated into this analysis. Males and females are pretty evenly distributed among the schools, with no statistically significant difference found [j2(l,153.5) = 1.436,p = 0.231]; race, however, was found not to be evenly distributed [/(5, 51.1) = 4.921,/? = 0.000], (See Table 14). This means that there are higher percentages of White students than any other racial group at both high schools. At Loy Norrix High School, the racial makeup is White 66% and Black 18%; at Kalamazoo Central High School, it consists of White 59% and Black 23%. 63 Students were not evenly distributed across high schools [x(3, 76.5) = 298.209,;? = 0.000] . If Galesburg, which only had one student attending WMU, is taken out along with the two unknown high school students, however, the high school student distribution is found to be evenly distributed between the two remaining schools \x (1, 151.5) = 2.063,/? = .151]. Table 15 gives the details of persistence, on probation, and nonpersistence by high school. Table 15. High School by Persistence and Non-Persistence LoyNorrix Kalamazoo Central Galesburg Augusta Unknown Total Persisters n/% 88/29 109/36 y 3 2/7 200/67 On Probation n/% 23/8 28/9 Q/0 0/0 51/17 Non-Persisters n/% 23/8 25/8 ^ ^-^ 49/16 Note. Percentages add up to only 96.3% due to rounding and percentages calculated by dividing by 307. Seven students had no probation status listed. Three students had no high school listed. Survey Summary This section reports the results of the Survey ofKalamazoo Promise Recipients at WMU Spring 2009. The response rate for this survey was 53% (101/191). It was intended, however, that the survey be sent to the entire population of WMU Kalamazoo Promise recipients. Had this occurred and still only 101 surveys were returned, the response rate would have been 33% (101/307) (see Table 4). Initially, 191 names were obtained from the facilitator of the Kalamazoo Promise at WMU. After the surveys were sent to these students and returned, however, the Office 64 of Student Academic and Institutional Research was contacted. This was done in order to meet FERPA regulations that require keeping student data anonymous, and because of the researcher's desire to link academic data to the survey data. Once data was obtained from the Office of Student Academic and Institutional Research, 307 rows of students without names were listed. An exact description of the distribution of Promise students who answered the survey and their probation status can be seen in Table 16. A total of 101 students responded to the survey, a response rate of 53%. This is quite high for a college student response rate. Unfortunately, only 4 out of the 13 students on academic dismissal responded, a response rate of only 31% within this group of students. This is actually not a bad response rate, either, but because of the nature of this research and wanting to understand what factors influence students to leave the university, it would have been better to have more who had been academically dismissed from WMU respond to the survey. The response rate within the group of students on probation was 40% (10/25). Lastly, the response rate within the group of students in good standing was 59% (87/147). Table 16. Distribution of Promise Students who Answered the Survey by Probation Status Did Not Respond On Time Response Late Response Total Good Standing 60 48 39 ~X\1 On Probation 15 2 8 25 Academic Dismissal 9 2 2 13 Total n/% 84/45 52/28 49/26 185/100% Note. Seven students had no record of probation status and were not included in this table, therefore, N= 300 not 307; and the total of Responded and Did Not Respond does not equal 191 surveyed because of this. 65 The survey, Survey of Kalamazoo Promise Recipients at WMU Spring 2009, consisted of 50 quantitative and qualitative questions with sub-questions. The survey was broken into seven sections: Background questions; Cognitive, Social and Institutional questions; About Kalamazoo Public Schools; questions about the Kalamazoo Promise; and lastly, Changes due to the Kalamazoo Promise. Only highlights of the first two sections are reported here, as they deal directly with retention factors addressed in this dissertation. Please see appendix G for details on the rest of the sections. Table 17. Did You Begin College at WMU or Elsewhere? Response Percent 15.8% Started elsewhere 84.2% Started at WMU If elsewhere, please specify where: Response Count 16 85 16 Most students reported starting at WMU, while almost 16% reported starting elsewhere first; see Table 17. Kalamazoo Valley Community College was the highest reported, while Michigan State University came in second. A few students also reported attending Grand Valley State, Eastern Michigan, and University of Michigan. Most students responding thought they would enroll for an advanced degree after completign their undergraduate degree; see Table 18. This indicates that most of the Kalamazoo Promise recipients at WMU responding have goals, which is a determining factor in retention (Bean & Metzner, 1985). 66 Table 18. Do You Expect to Enroll for an Advanced Degree When, or if, You Complete Your Undergraduate Degree? Response Percent Response Count ~No 34.7% 34 Yes 65.3% 64 Note. Three did not answer this question. Thirty-five percent of students who answered the survey reported that they live in the dorm; see Table 19., The academic data, however, indicates that almost 70% (214/307) of the WMU Kalamazoo Promise recipients live in dorms. Table 19. Where Do You Live During the School Year? Dormitory or other campus housing Residence (house, apartment, etc.) within walking distance of Western Residence (house, apartment, etc.) within driving distance Fraternity or sorority house Response Percent Response Count 35.1% 34 19.6% 19 45.4% 44 0.0% 0 Note. N=97, 4 did not answer. Almost 40% of the parents or guardians of the WMU Promise students do not have a college degree; see Table 20. In other words, 40% of those students who responded to the survey are first-generation college students. 67 Table 20. What is the Highest Level of Education Obtained by Your Father or Mother? Father or Male guardian Not a high school graduate 3 High school diploma or 18 GED Some college, did not 17 complete degree 8 Associate degree 20 Bachelor's degree 9 Master's degree Doctorate degree and/or 4 Professional degree 2 Unknown Mother or Female guardian 3 Response Count 6 13 31 13 30 15 24 14 23 44 23 2 6 3 5 Response ^ 3.57% 18.45 17.86 13.69 26.19 13.69 3.57 2.98 Note. N=96, 5 did not answer. Most students spend anywhere from 1 to 20 hours preparing for class; see Table 21. Almost 22% of the students who responded do not work, while just over 28% work more than 21 hours per week. Of the students who answered this question, a little over 45% do not participate in any college-sponsored activities or organizations. Being involved with these types of activities is a key indicator of retention (Swail, 2003). A little over 47% of the WMU Promise students report that their jobs take time away from their school work; see Table 22. 68 Table 21. About How Many Hours Do You Spend in a Typical 7-day Week Doing Each of the Following? 1-5 6-10 11-20 21-30 31+ Response Count 27 31 30 6 2 97 21 10 14 24 21 6 96 43 38 10 3 1 0 95 67 23 2 0 2 96 17 73 5 0 0 95 None Preparing for class (studying, reading, writing, rehearsing or other activities related to your program Working for pay Participating in collegesponsored activities organizations, campus publications, student government, intercollegiate or intramural sports, etc.) Providing care for dependents living with your (parents, children, spouse, etc.) Commuting to and from classes 0 Note. N=91, 4 did not answer Table 22. If You Have a Job, How Does it Affect Your School Work? I don't have a job My job does not interfere with my school work My job takes some time from my school work My job takes a lot of time from my school work Response Percent 25.8% 26.8% 39.2% 8.2% Response Count 25 26 38 Note. N=91, 4 did not answer Almost 61% of students felt they were academically prepared for classes at WMU. Almost 19%o felt they were not academically prepared and that this might lead to them withdrawing from classes. The rest were neutral. Working too many hours was 69 also a factor that may cause students to withdraw; see Table 23. Table 23. How Likely is it That the Following Issues Would Cause You to Withdraw From Class or From WMU? Working full-time Caring for dependents Academically unprepared Lack of finances Don't fit in Don't offer program of study that I want Not Likely 1 42 39 75 .„ . 16 16 Very Likely Response 5 Count 16 7 97 15 18 6 6 96 17 20 11 7 97 16 13 15 3 12 16 4 0 96 97 11 13 13 12 97 l 3 4 Note. N=97, 4 did not answer Established research suggests that the more involved students are, the more likely they are to persist (Swail, 2003); however, only about 3% of the students responding to the survey are involved in any social fraternity or sorority; see table 24. Table 24. Are You a Member of a Social Fraternity or Sorority? Response Percent 96.8% 3.2% No Yes If yes, which one? Response Count 91 3 4 Note. N=94, 7 did not answer Regarding attending WMU, most students reported having supportive friends and family; see Tables 25 and 26. The more supportive people there are surrounding students, the more likely they are to persist (Swail, 2003). 70 Table 25. How Supportive Are Your Friends of Your Attending WMU? Response Percent 2.1% 15.5% 36.1% 46.4% Not Very Somewhat Quite a bit Extremely Response Count 2 15 35 45 Note. N=94, 7 did not answer Table 26. How Supportive is Your Immediate Family of Your Attending WMU? Response Percent 0.0% 6.2% 23.7% 70.1% Not Very Somewhat Quite a bit Extremely Response Count 0 6 23 68 Note. N=94, 7 did not answer Most students responding to the survey find their relationships with students at WMU friendly and supportive. They feel a sense of belonging. They feel their instructors are available, helpful and sympathetic. Almost 63% of students reported that administrative and office personnel are helpful, informative and flexible; see Tables 27, 28 and 29. 71 Table 27. Which Best Represents the Quality of Your Relationship With Students atWMU? 1 Unfriendly, unsupportive, sense of alienation 2 3 4 5 Friendly, supportive, sense of belonging Response Percent 0.0% 6.2% 19.6% 36.1% 38.1% Response Count 0 6 19 35 37 Note. N=94, 7 did not answer Table 28. Which Best Represents the Quality of your Relationships With Instructors at WMU? 1 Unavailable, unhelpful, unsympathetic 2 3 4 5 Available, helpful, sympathetic Response Percent Response Count 0.0% 4.1% 0 4 32 43 18 33.0% 44.3% 18.6% Note. N=94, 7 did not answer Table 29. Which Best Represents the Quality of Your Relationship With Administrative Personnel & Office Staff at WMU? 1 Unhelpful, Inconsiderate, rigid 2 3 4 5 Helpful, considerate, flexible Response Percent Response Count 0.0% 14.4% 33.0% 33.0% 19.6% 0 14 32 32 19 Note. N=94,7 did not answer Almost 85% of students responding, when asked whether they would still attend 72 WMU if they could start over again, answered "probably yes" or "definitely yes." Only a little over 15% said "probably no" or "definitely no"; see Table 30. Ninety-five percent would recommend WMU to a friend or family member, while 4.3% would not: see table 31. Table 30. If You Could Start Over Again, Would You Still Attend WMU? Response Percent 3.3% 12.0% 53.3% 31.5% Definitely no Probably no Probably yes Definitely yes Response Count 3 11 49 29 Note. N=92, 9 did not answer Table 31. Would You Recommend WMU to a Friend or Family Member? Response Percent 4.3% 95.7% No Yes Response Count 4 88 Note. N=92, 9 did not answer Research Question One Results This section reports the results associated with all three sub-questions of Research Question One. This question seeks to determine to what extent persister, those on probation, and non-persister Kalamazoo Promise recipients differ by demographic characteristics on each of the following selected factors in Swail's (2003) Geometric Model of Student Persistence and Achievement: (a) Cognitive Factors, (b) Social Factors, and (c) Institutional Factors. 73 The Probation variable was used to determine if a student persisted or not. Students who were academically dismissed were considered not to have persisted. Students whose probation status was in Good Standing were considered to have persisted. Those in the On Probation category fall in between these two groups. For now they are persisters, but they look more like non-persisters and as such are examined separately. Research Question 1.1 Results 1.1: Among the three groups of students (persister, those on probation, andnon- persister Kalamazoo Promise recipients, by gender and race), are there any significant differences in the cognitive factors from Swail 's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) high school GPA, (b) most recent WMU GPA, (c) ACT composite score, (d) taking a remedial math course at WMU, (e) taking a remedial reading course at WMU, (f) taking a remedial writing course at WMU, or (g) taking AP credit? The research question was assessed using the MANOVA procedure to assess group differences in the continuous variables (a high school GPA, most recent WMU GPA, and ACT composite score). The Chi-square Test of Independence was used to assess group differences in the categorical variables (taking remedial courses at WMU and taking AP credit in high school). Table 6 delineates the details of each question with related variables and analysis. The first step taken in interpreting the results from the MANOVA procedure was to assess the value of Box's M, which tested the assumption of equal variance across groups (Creswell, 2005). Had the assumption of equal variance been upheld, then the 74 Wilk's Lambda test statistic would have been used to interpret the results. The assumption of equal variance across groups was violated, however, so the Pillai's Trace was used to interpret the results. Results of the data analysis revealed the Box's M = 341.268 to be statistically significant [F (90, 4983.931) = 3.226, p = .000]. Therefore the assumption of equal variance is not upheld. Consequently the Pillai's Trace overall results across three dependent variables were used to interpret results from the MANOVA procedure. The results showed statistically significant differences among groups based on persistence, race, and gender. There were no statistically significant interactive effects due to gender, but there was one statistically significant interaction between persistence and race [F (12, 846) = 2.110,p = .014] (see Table 32). The significant value of Pillai's trace warranted further investigation to determine the source of the statistically significant differences. The obtained test statistic for comparisons of the groups based on persistence revealed a statistically significant difference of [F (6, 562) = 39.234 and/? = .000]. The obtained test statistic for comparisons of the race groups, F (6, 562) = 7.747 andp = .000) and comparisons of the groups based on gender revealed a statistically significant difference of [F (3, 280) = 2.876 andp = .037] between the groups, see Table 17. Lastly, a test statistic for the interaction of persistence and race, F (12, 846) = 2.110 andp = .014] between groups was determined. 75 Table 32. Summary of Omnibus MANOVA Test of Group Differences across Dependent Variables Effect Persistence RaCe Pillai's Trace Hypothesis Partial Eta Observed Value F df Error df Sig. Squared P ° w e r 39.234 6 562.000 .000 .295 1-000 .590 .153 7.747 6.000 562.000 .000 .076 1-000 2.876 3.000 280.000 .037 .030 .684 persistence * race 2.110 12.000 846.000 .014 .029 .942 persistence * gender .687 6.000 562.000 .660 .007 .275 1.425 6.000 562.000 .203 .015 -558 1.428 12.000 846.000 .147 .020 .788 Gender ^ race * gender persistence * race * gender .060 Note. Alpha = .05 Post Hoc analysis. The overall MANOVA revealed statistically significant group differences across the dependent variables and one interaction; follow-up tests were therefore conducted to locate the source of the differences. The follow-up tests consisted of a univariate ANOVA for each dependent variable (Gay, Mills, & Airasian, 2006; Mertler & Vanatta, 2007; Spinthall, 2007). To reduce the occurrence of Type I error when conducting a series of ANOVA's, the Bonferonni correction procedure was used. This procedure set alpha at a more stringent level to keep the alpha across the set of comparisons at a predetermined level. In this case the adjusted alpha equals the overall alpha for the analysis (.05) divided by the number of dependent variables (3). Therefore critical alpha for the post hoc univariate analyses was alpha = .016 (alpha = .05 / 3 = .016). 76 Differences by race. Table 33 presents a summary of the univariate test for the groups based on race. With the adjusted critical value of alpha, .016, only two statistically significant results were found. The obtained test statistic for comparisons of the race groups in terms of ACT composite score revealed a test statistic of F (2, 282) = 9.410 andp = .000. The magnitude of the effect size for the results was n =.063, which according to Cohen (1988) is a small to medium effect. The obtained power of .978 revealed that the differences in ACT composite score between the three groups were large enough to be detected 97.8% of the time. A post hoc analysis and a review of the descriptive statistics revealed that White students, on the average, had significantly higher ACT composite scores (M= 20.989, sd = .492) than Black students (M= 17.368, sd = .698) or students in the Other category (M= 18.874, sd =.925). There were no statistically significant differences between the ACT composite scores of Black students and students in the Other category. Table 33. Summary of Test for Differences by Race across the Dependent Variables Observed Power Dependent Variable df F Sig. Partial Eta Squared High School GPA 2 14.484 .000 .093 .999 Most Recent WMU GPA 2 3.190 .043 .022 .607 ACT Composite Scores 9.410 .000 .063 .978 2 The obtained test statistic for comparisons of the race groups in terms of High School GPA revealed a test statistic of F (2, 282) = 14.484 and/? = .000. A post hoc analysis and a review of the descriptive statistics revealed that White students, on the average, had significantly higher high school GPAs (M= 3.419, sd = .064) than Black 77 students (M= 2.818, sd = .091) or students in the Other category (M =3.188, sd = .121). Students in the Other category had significantly higher high school GPAs than Black students. This means that of the three racial groups, Black student have the lowest high school GPAs. Most recent WMU GPA, although p = .043, is not significant because of the adjusted alpha, .05/3 = .016 to control for Type I error. Difference by gender. Table 34 presents a summary of the univariate test for the groups based on gender. With the adjusted critical value of alpha, .016, no statistically significant result was found. The obtained test statistic for comparisons of the gender groups in terms of ACT composite score revealed a test statistic of F ( l , 282) = 4.385 and p = .037. A post hoc analysis and a review of the descriptive statistics revealed that male students, on average, had significantly higher ACT composite scores (M= 21, sd = 5) than female students (M= 19, sd = 5), however, because of the adjusted critical value of alpha, .016 ACT composite score is considered not statistically significant. This means that statistically males and females score similarly on their ACT composite. There was not a statistically significance difference between males and females across high school GPA or most recent WMU GPA; see Table 34. This means that males and females have similar GPA, both in high school and at WMU. 78 Table 34. Summary of Test for Differences by Gender across the Dependent Variables Dependent Variable High School GPA .003 Observed Power .167 .555 .001 .091 .037 .015 .551 df 1 F .980 Sig. .323 Most Recent WMU GPA 1 .350 ACT Composite Scores 4.385 1 Partial Eta Squared Note. Adjusted alpha = .016, (.05/3) Differences among persisters, those on probation, and non-persisters. Table 35 presents a summary of the results. With the adjusted critical value of alpha = .016, two statistically significant results were found. The obtained test statistic for comparison of the most recent WMU GPA for persisters, those on probation, and non-persisters revealed a test statistic of F(2, 282) = 198.560 and/? = .000. A review of the pair-wise comparisons and descriptive statistics revealed that the most recent WMU GPAs (M = 2.982, sd= .046) of persisters were higher than those of students on probation (M= 1.702 ,sd = .095) and were higher than those of non-persisters (M = \.2\2,sd= .088). Also, those on probation had a statistically higher most recent WMU GPA than did nonpersisters. This means that persisters had the highest most recent WMU GPA, while nonpersisters and the lowest. Students on probation fell between these two groups. 79 Table 35. Summary of MANOVA Results for Persistence across the Dependent Variables Persistence Dependent Variable df 2 High School GPA Most Recent WMU 2 GPA ACT Composite 2 Score F 4.356 Sig. .014 Partial Eta Observed Squared Power .030 .752 198.560 .000 .585 1.00 .183 .001 .078 .833 Note. Alpha is .016; (.05/3) The obtained test statistic for comparison of high school GPA for persisters, those on probation and non-persisters revealed a test statistic of F (2, 282) = 4.356 and/? = .014. A review of the pair-wise comparisons and the descriptive statistics revealed that persisters had higher high school GPAs (M= 3.50, sd = .76) than did non-persisters (M = 3.04, sd = .58). There was no statistically significant difference between those on probation (M- 3.15, sd = .44) and non-persisters or those on probation and persisters. Students on probation had high school GPAs that fell in between persisters and nonpersisters. Students who are persisters had the highest high school GPAs. There was not a statistically significant difference in terms of ACT composite scores among persisters, those on probation, or non-persisters. Statistical comparison for persistence by race interaction. Table 36 presents a summary of the univariate test for the groups based on interaction. With the adjusted critical value of alpha, .016, only one statistically significant result was found. The test for interaction between persistence and race in terms of most recent WMU GPA revealed a test statistic of F (2, 282) = 6.257 and/? = .000. A review of the descriptive statistics revealed that White students in good standing (persisters) had higher most recent WMU 80 GPAs (M= 3.136, sd = .047) than Black students (M = 2.826, sd = .091) or students in the Other category (M= 2.985, sd =.090) who were also in good standing. Students from the Other category who were in good standing had higher WMU GPAs than Black students in the same category. There were few differences in the WMU GPAs of students classified as being on probation. Black students who were dismissed (nonpersisters) had higher WMU GPAs (M = 1.3 3 6 , ,s J = . 13 5) than students in the White (M = .982, sd= A15) or other (M = 1.316, sd = . 195) racial groups who were also dismissed. Table 36. Summary of MANOVA Results for Persistence and Race Interaction across the Dependent Variables Persistence Dependent Variable High School GPA Most Recent WMU GPA ACT Composite Score df F Sig. Partial Eta Observed Squared P o w e r 2 .548 .701 .008 -182 2 6.257 .000 .082 -988 2 .933 .445 .013 -295 Note. Alpha is .016; (.05/3) The second part of research question 1.1 was run separately using the Chi-square Test of Independence analysis because of the nominal variables; remedial courses, advanced placement credit, gender, and race. A crosstabs procedure, using the Chi-square Test of Independence, revealed a statistically significant difference, ^2(2, 300) = 21.854 and/> =.000, among persisters, those on probation, and non-persisters in terms of whether the students had taken remedial courses at WMU. The data revealed that approximately 35% of non-persisters had taken at least one remedial course at WMU, approximately 22% of those on 81 probation had taken at least one remedial course at WMU, while only 9% of the persisters reported having taken a remedial course. This means that students who took remedial classes were more likely not to persist. The results also revealed a statistically significant difference,^ (2, 300) = 10.696 andp =.005, among the persisters, those on probation, and non-persisters in terms of haven taken AP courses in high school. The data revealed that 12% of the persisers had taken at least one AP course in high school, while those on probation did not take any AP course, and 2% of non-persisters took AP courses in high school. This means the students who take AP courses in high school are more likely to persist. 82 Table 37. Crosstab for Tests of Differences Among Persisters, Those on Probation, and Non-persisters Across the Variables of Persistence and Taking Remedial Courses at WMU, AP Credit, Gender, and Race Total 182 On NonProbation . Persister 40 32 91% 78.4% 65.3% 85% 18 11 17 46 9% 21.6% 34.7% 15% 200 51 49 300 176 51 48 275 88% 100% 98% 91.7% 24 0 1 25 12% 0% 2% 8.3% 200 51 49 300 102 29 32 163 % within persistence 51% 56.9% 65.3% 54.3 Count 22 17 137 % within persistence 49% 43.1% 34.7% 45.7% Total Count 200 51 49 300 Count 130 29 24 183 % within persistence 65% 56.9% 49.0% 61.0% Count 13 17 65 % within persistence 17.5% 25.5% 34.7% 21.7% Count 9 8 52 % within persistence 17.5% 17.6 16.3% 17.3% Total Count 51 49 300 Persister No Remedial Course Count % within persistence Yes Remedial Course Count % within persistence Total Count No AP Credit Count Yes AP Credit % within persistence Count % within persistence Total Count Male Female White Black Others Count 98 35 35 200 254 The results failed to reveal a statistically significant difference in terms of gender, 83 X2(2, 300) = 3.405 andp =.182. A crosstabs procedure, Chi-square Test of Independence, also failed to reveal a statistically significant difference,^2(4, 300) = 7.648 andp =.105, among persisters, those on probation, and non-persisters in terms of race. See table 37 for details. Research Question 1.2 Results 1.2 Among the three groups of students: persister, those on probation, and nonpersister Kalamazoo Promise recipients by gender and race, are there any differences in the social factors from Swail 's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) living in a dorm, (b) being an athlete or (c) parental income? This research question was addressed through the use of the One-way Analysis of Variance procedure (ANOVA) and a Chi-square Test of Independence. The ANOVA procedure was used to assess the differences among the groups on the interval level data (parental income). The Chi-square Test of Independence was used to assess the differences among the groups on the categorical level data of living in the dorm. The One-way ANOVA failed to reveal a statistically significant difference, F (2, 299) = 1.926 andp = .148, among the three groups on parental income, which means that students were persisters, on probation, or non-persisters regardless of income: students whose parents had high incomes were not found to have persisted more than students whose parents had low incomes. There were four athletes, all male; one white male was in good standing, two 84 white males were on probation, and the fourth athlete was a black male who had been academically dismissed. This was too small of a sample to examine statistically, so the variable was not addressed in the statistical procedures. Lastly, the variable living in a dorm was examined using the crosstabs procedure. There was no statistically significant difference,^ (2, 300) = .445 and/? =.801, found in the observed and expected frequency count of persisters, those on probation, and nonpersisters in terms of living in the dorm. This means that living in a dorm made no difference on whether a student persisted, was on probation, or did not persist. Research Question 1.3 Results 1.3 Among the three groups of students: persister, those on probation, and nonpersister Kalamazoo Promise recipients, by gender and race, are there any differences in the institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement using the following dependent variables from the academic data: (a) first year experience (FYE), and (b) which high school Promise students came from? This research question was addressed through the use of the Chi-square Test of Independence to assess the differences among the groups on the categorical levels of which high school the students attended and First Year Experience (FYE). The Chi-square Test of Independence indicated that there was not a statistically significant difference, y?(6, 299) = 2.479 andp =.871 in the observed and expected frequency count of persisters and non-persisters in terms of which high school they attended. One non-persister did not have a high school listed, therefore n = 299 instead 85 of n = 300. This means that approximately the same number of students from Loy Norrix High School and Kalamazoo Central High School persisted, were on probation, or did not persist. One high school did not produce more students who were non-persisters than the other high school. The Chi-square Test of Independence, however, did reveal a statistically significant difference, x2(2, 300) = 10.101 and/? =.006, among the three groups in terms of First Year Experience at WMU, which means that there was a difference among students who persisted, who were on probation, or who did not persist in terms of whether they participated in the first year experience or not. Only 37.7% of the WMU Kalamazoo Promise students participated in the FYE, the other 62.3% did not, see table 38. Of those students who did participate in FYE, about half of those on probation and about half of the non-persisters participated. Only about 30% of persisters participated. This means that a higher percentage of those on probation and the non-persisters participated in the FYE than the percentage of persisters. Table 38. Crosstab for Tests of Differences Among Persisters, Those on Probation, and Non-persisters Across the Variables of Persistence and Taking Remedial Courses at WMU, AP Credit, Gender, and Race Persister On NonTotal Probation Persister 137 24 26 187 % within No Persistence 68.5% 47.1% 53.1% 62.3% Count 63 27 23 113 % within Yes Persistence Total Count 31.5% 52.9% 46.9% 37.7% 200 51 49 300 Count 86 Research Question Two Results Research question two is broken into two sections, both of which examine the number of courses taken per term and the number of courses taken the first year. Research question 2.1 examines persisters, those on probation, and non-persisters, while research question 2.2 examines respondents, late respondents, and non-respondents controlling for race and gender in both questions. Research Question 2.1 Results 2.1 Is there a difference in the average number of overall courses taken per term and number of courses taken the first year among the three groups of students: Persister, On probation and Non-persister WMU Kalamazoo Promise recipients, controlling for gender and race, using the course summary data? An analysis of covariance (ANCOVA) was conducted to assess the differences in the average number of overall courses taken per term by persisters, those on probation, and non-persisters, when controlling for gender and race. Table 39 presents a summary of the results. Results indicate that there was no statistically significant [F{\, 300) = 1.027,/> = .312] differences in the average number of classes taken per term for the three groups across the variable gender. Results, however, indicated that there were statistically significant [F(l, 300) .4.148, p = .043] differences in the average number of classes taken per term for the three groups across the variable of race. The descriptive statistics showed that students in the Other category took an average of 5.26 courses per term, while Black students took an average of 5.00 courses per term. White students took, on average, the fewest courses per term, 4.94; see table 87 40. The final analysis revealed statistically significant [F(l, 300) = 4.440,/? = .013] differences in the average number of courses taken per term by persisters, those on probation, and non-persisters. Persisters took more average courses per term than Nonpersisters. There were no other differences. The ANCOVA failed to indicate any interactive effects between, gender, race, persistence, and the average number of classes taken per term. Table 39. ANCOVA Results of Persisters and Non-persisters, Controlling for Race and Gender for Average Number of Courses Taken Per Term Sig. Partial Eta Squared Observed Power 4.148 .043 .014 .528 1.393 1.027 .312 .003 .173 6.021 4.440 .013 .029 .760 df Mean Square F Race 2 5.625 Gender 1 Persistence 2 88 Table 40. Descriptive Statistics for Gender, Race and Persistence for Average Number of Courses Taken Per Term Average Number of Courses per Term Persistence Race Academic Gender Standard Deviation 1.30 Variance Total N 200 1.70 On Probation 4.85 .83 .70 51 Non-persister 4.65 .84 .70 49 Total 5.00 1.18 1.39 300 White 4.94 1.01 1.02 189 Black 5.00 1.18 1.40 66 Other 5.26 1.63 2.66 52 Total 5.01 1.18 1.38 307 Male 5.05 1.10 1.21 164 Female 4.97 1.26 1.59 143 Total 5.01 1.18 1.38 307 Persister Mean 5.13 An analysis of covariance (ANCOVA) was conducted to assess the differences in the number of courses taken the first year by persisters, those on probation, and nonpersisters, when controlling for gender and race. Table 41 presents a summary of the results. Results indicate that there was no statistically significant [F(\, 300) = .999, p = .318] differences in the number of courses taken the first year for the three groups across the variable gender. Nor was a statistically significant result found across the variable race [F(l, 300) = .3.204, p = .074]. Results, however, indicated that there was a statistically significant [F(l, 300) 3.574, £> = .029] difference in the number of courses taken the first year among persisters, those on probation, and non-persisters. The descriptive statistics, table 42, showed that students who are persisters took on average more courses (n = 10.17) their first year than those on probation (n = 9.12) 89 and non-persisters (n = 9.31). What is interesting is that Non-persisters took on average slightly more courses than those on probation. The ANCOVA failed to indicate any interactive effects between, gender, race, persistence, and the average number of classes taken per term. Table 41. ANCOVA Results of Persisters, Those on Probation, and Non-persisters, Controlling for Race and Gender for Number of Courses Taken the First Year Sig. Partial Eta Squared Observed Power 3.204 .074 .011 .528 10.122 .999 .318 .003 .173 6.021 4.440 .013 .029 .760 df Mean Square F Race 1 32.470 Gender 1 Persistence 1 Table 42. Descriptive Statistics for Gender, Race and Persistence for Number of Courses Taken the First Year Number of Course First Year persistence Race Academic Gender Mean Standard Deviation Variance Total N 10.17 3.53 12.48 200 On Probation 9.12 2.73 7.43 51 Non-persister 9.31 1.91 3.63 49 Total 9.85 3.22 10.34 300 White 9.51 3.08 9.47 189 Black 10.08 3.19 10.19 66 Other 10.29 3.81 14.52 52 Total 9.76 3.24 10.52 307 Male 9.94 3.01 9.05 164 Female 9.56 3.49 12.19 143 Total 9.76 3.24 10.52 307 Persister 90 Research Question 2.2 Results 2.2 Is there a difference in the average number of overall courses taken per term and number of courses taken the first year among the three groups of students: Respondents, late respondents and non-respondents of the Survey of Promise Scholarship Recipients at WMU Spring 2009 controlling for gender and race using the course summary data? An analysis of covariance was conducted to assess the differences between the average number of courses per term taken by respondents, late respondents, and nonrespondents, controlling for gender and race. Table 43 presents a summary of the results. Results indicate that there were no statistically significant differences among the three groups in the number of classes taken per term when compared by race [F(l, 185) = 2.305,/? = .131] or by gender [F(l, 185) = .344,/? = .558] and also no statistically significant [F(2, 185) = .314,p = .731] difference among the three groups in the average number of classes taken per term. The ANCOVA also failed to indicate any interactive effects between, gender, race, category of respondent, and the average number of classes taken per term. Table 43. ANCOVA Results of Respondents, Late Respondents and Non-respondents, Controlling Across Race and Gender for Average Number of Classes Taken Per Term Observed Power df Mean Square F Race 1 4.011 2.305 .131 .013 .327 Gender 1 .598 .344 .558 .002 .090 GroupByResponse 2 .547 .314 .731 .003 .099 91 Sig. Partial Eta Squared Table 44. Descriptive Statistics for Gender, Race and Response for Average Number of Courses Taken Per Term Average Number of Course per Term Survey Group by Response Race Academic Gender Respondent Mean 5.13 Standard Deviation 1.07 Variance Total N 1.15 52 Late Respondent 5.06 1.06 1.13 49 Non-respondent 5.22 1.57 2.47 84 Total 5.15 1.32 1.73 185 White 4.94 1.01 1.02 189 Black 5.00 1.18 1.40 66 Other 5.26 1.63 2.66 52 Total 5.01 1.18 1.38 307 Male 5.05 1.10 1.21 164 Female 4.97 1.26 1.59 143 Total 5.01 1.18 1.38 307 An analysis of covariance was conducted to assess the differences in the number of courses taken the first year by respondents, late respondents, and non-respondents, controlling for gender and race. Table 45 presents a summary of the results. Results indicate that there were no statistically significant differences among the three groups in the number of courses taken for by race [F(l, 185) = 2.305,/) = .131] or gender [F(l, 185) = .015, p = .903],. The final analysis revealed that there was no statistically significant [F(2, 185) = .271, p = .763] difference in the number of courses taken the first year for the three groups. The ANCOVA also failed to indicate any interactive effects between, gender, race, category of respondent, and the number of classes taken the first year. 92 Table 45. ANCOVA Results of Respondents, Late Respondents and Non-respondents, Controlling across Race and Gender for the Number of Courses Taken the First Year Observed Power df Mean Square F Race 1 13.987 1.139 .287 .006 .186 Gender 1 .181 .015 .903 .000 .052 .3.325 .271 .763 .003 .092 GroupByResponse 2 Sig. Partial Eta Squared Table 46. Descriptive Statistics for Gender, Race and Response for Number of Courses Taken the First Year Number of Courses First Year Survey Group by Response Race Academic Gender Mean Standard Deviation Variance Total N Respondent 10.23 2.56 6.53 52 Late Respondent 9.98 3.81 14.48 49 Non-respondent 10.39 3.80 14.41 84 Total 10.24 3.48 12.12 185 White 9.51 3.08 9.47 189 Black 10.08 3.19 10.19 66 Other 10.29 3.81 14.52 52 Total 9.76 3.24 10.52 307 Male 9.94 3.01 9.05 164 Female 9.56 3.49 12.19 143 Total 9.76 3.24 10.52 307 Research Question Three Results Research question three examines to what extent respondent, late respondent, and non-respondent Kalamazoo Promise recipients differ on each of the following selected factors in Swail's (2003) Geometric Model of Student Persistence and Achievement: (a) 93 Cognitive Factors, (b) Social Factors, and (c) Institutional Factors, using known characteristics from the academic data. In addition, to what extent do early respondents differ from late respondents on variables from the Survey of Promise Scholarship Recipients at WMUSpring 20091 Lastly, using Groves and Couper's bias ratio formula, was there an indication of non-response bias? Data Analysis for Research Question 3 The data analysis for research question three was accomplished in several phases. First, the Chi-square Test of Independence was used to assess the differences among the groups on the categorical level variables in the cognitive, social, and institutional factors included from the Survey of Promise Scholarship Recipients at WMU Spring 2009. The data were assessed using the Pearson Chi-square Test of Independence, which is a nonparametric statistical procedure that addresses the differences between observed and expected frequency counts for a distribution of scores (Mertler &Vanatta, 2007; Rosenberg, 2007; Stevenson, 2007). Nonparametric tests are not as powerful as parametric statistical procedures; however the Pearson Chi-square Test of Independence is an appropriate procedure for assessing the differences in the observed and expected frequency counts of the two groups across the dependent variables. Second, items for the continuous/interval level data from the cognitive, social, and institutional factors included from the Survey of Promise Scholarship Recipients at WMU Spring 2009 were grouped to create the following four summated subscales: cognitive engagement, social demands, institutional support, and social engagement. The exact details of each of the scales can be found in Appendix C. 94 The Cognitive Engagement Subscale consisted of 20 items (Items C7 through C27) from the Cognitive Factor of the survey. Participants responded to items on the Cognitive Engagement subscale using a 5-point Likert-type scale where the responses ranged from 1 = not likely to 5 = very likely. The Social Demands subscale consisted of five items (Items S13 through SI 7) from the Social Factor of the survey. Participants responded to items on the Social Demands subscale using a 5- point Likert-type scale, where the responses ranged from 1 = never to 5 = often. The Institutional Support subscale consisted of six items (Items 17 through 113) from the Institutional Factor of the survey. Participants responded to items on the Institutional Support subscale using a 5-point Likert-type scale, where the responses ranged from 1 = never to 5 = very often. The Social Engagement subscale consisted of five items (Items 114 through 118) from the Institutional Factor of the Survey. Participants responded to items on the Social Engagement subscale using a 5- point Likert-type scale, where the responses ranged from 1 = very little to 5 = very much. Summated scales offer an advantage over single-item scales in that such scales can be assessed for reliability and the unidimensionality of the construct being measured (Thorndike, 1967). Items assigned to each scale were summated together to yield total scores. Before running statistical procedures on data from the survey, the researcher assessed the internal consistency of the scales using reliability analysis. Cronbach's coefficient alpha was used to measure the internal consistency of the scales included in the survey (Cohen, 1988; Trochim, 2007). While any test developer hopes to obtain a 95 reliability coefficient that approaches 1.0, such a value is rarely obtained in behavioral and social science research. The significance of the obtained alphas will be tested against the value of alpha = .70, as suggested by Kaplan and Saccuzzo (2005), for the obtained alpha coefficients. The research indicates that values of .70 or greater indicated that a scale is internally consistent (Kaplan & Saccuzzo, 2005; Mertler & Vanatta, 2005). Table 47 presents a summary of the descriptive statistics for the four scales. The results indicate that Social and Cognitive Engagement Subscales obtained coefficient alphas were statistically significant atp < .05. The results further revealed that the other two scales obtained acceptable internal consistency estimates for the scores obtained from this study. The results indicate that the four scales collected reliable data from the participants in this study. Table 47. Summary of Results from the Reliability Analysis and Descriptive Statistics for Subscales of the Survey of Promise Scholarship Recipients 95% Confidence Interval F Test with True Value .7 Lower Upper Alpha Bound Bound Value dfl 1° .687 Engagement Institutional 7 g 2 Support Social 74g Demands Cognitive g30 Engagement .959 87 dft Sig n 261 .582 4 .565 .782 JQ4 g45 L3?4 g6 5J6 mi+ 66Q g20 U 9 1 93 465 126 774 877 L?61 89 1513 mo* Mean sd 12.352 3.594 ? 23.207 4.698 6 12.4647 5.165 lg 54.111 10.308 Note. * Significant atp<.05 In addition to assessing the internal consistency of the scales contained in the 96 survey, an item-analysis was performed on the individual items in a scale. This statistical analysis provided information of the internal consistency of single items as they related to the homogeneity of items contained in a scale (Thorndike, 1967). Appendix F presents a summary of the results. The item analysis was conducted by investigating the item-total correlation for each item in a scale. Items with a correlation of .30 or higher were retained for inclusion in subsequent analytic procedures. This value was chosen because it represents the critical value of r with alpha set at .01 and df- 100 (Ary, Jacobs & Razavieh, 1996). Items with lower correlations were excluded from the subsequent statistical procedures, if excluding the items did not decrease alpha of the scale to which the item was assigned. In addition, items with correlation less than .30 were considered for either modification or removal from the questionnaire (Ary, Jacobs & Razavieh, 1996; Thorndike, 1967). Items were considered for removal if removing the item did not decrease the alpha for the scale. A review of the results, as presented in Appendix F, indicated that all items on the Institutional Support subscale and Social Demands subscale were good items as all achieved inter-item correlations that exceeded the cut score of .30. However, one item from the Social Engagement subscale (Attended Art Exhibit, Play, Dance, or Music) was excluded from subsequent statistical analyses due to a low inter-item correlation. Deleting the item resulted in four items remaining for the scale: Often Exercise Participated in Physical Education ; Often Participated in Spiritual Activities; Often Tried Understand Someone Else's Eyes and Often Learned Something Changed Way Understand Issue. A review of the results for the Cognitive Engagement subscale resulted in two 97 items, (C8, Come to class without completing readings or assignments, C23 Skipped class) being excluded from subsequent statistical analyses due to negative inter-item correlations with other items in the scale. The researcher initially recoded reverse coding the items in an effort to improve the total inter-item correlation; however, doing so did not raise the correlations above the critical cut score of .30. Deleting the items resulted in 18 items remaining for the scale. All items in each subscale can be seen in detail in Appendix F. Research Question 3.1 Results 3.1 Among the three groups of students: respondent, late respondent, and nonrespondent Kalamazoo Promise recipients, are there any differences in the following known dependent variables from the academic data: (a) high school GPA, (b) most recent WMU GPA, (c) ACT composite score, (d) taking a remedial math course at WMU, (e) taking a remedial reading course at WMU, (f) taking a remedial writing course at WMU, (g) taking AP credit, (h) living in a dorm, (i) being an athlete, (j) parental income, (k) first year experience (FYE) or (I) high school that could indicate possible non-response bias? As academic data obtained from the Office of Student Academic and Institutional Research provides data on all students regardless of whether they responded to the Survey of Promise Scholarship Recipients at WMU Spring 2009 or not, this data was used to determine if these students are similar. If these students are similar, then the survey data could be generalized to the population of Western Michigan University Kalamazoo Promise Scholarship recipients and not just to those students who responded to the 98 survey. If students who responded to the survey are different from those who did not respond to the survey, then possible non-response bias must be examined in more depth. Table 48. Distribution of Promise Students who Answered the Survey by Probation w = 185 Did Not Respond On Time Response Late Response Total Good Standing 60 48 39 147 On Probation 15 2 8 25 Academic Dismissal 9 2 2 13 Total 84 52 49 185 Note, Seven students had no record of probation status and were not included in this table, there for, N= 300 not 307; and the total of Responded and Did Not Respond does not equal 191 surveyed because of this. Non-respondents (JV=84) in this research are considered to be those survey recipients who did not fill out the online survey. Late respondents are those respondents who only after more than two reminders responded (April 20 and later was the cutoff date). "Persons who respond in later waves are assumed to have responded because of the increased stimulus and are expected to be similar to non-respondents" (Armstrong & Overton, 1977, p.2). As non-respondents did not fill out the survey, non-respondents and respondents could not be compared using the survey data. Therefore using the successive wave method of examining non-response bias, on-time respondents (N=52) were compared to late respondents (/V=49) using the data obtained from the administration of the survey. Successive waves refer to the stimulus done over time, i.e. reminder emails, post cards, follow-up calls (Armstrong & Overton, 1977). In addition, because additional data was obtained on all students through academic records, non-respondents (7V=84) were compared to respondents (iV=101) and non-respondents were compared to late respondents, to detect if any differences exist for 99 the population of WMU Promise students. This research question was addressed through the use of the MANOVA procedure and a Chi-square Test of Independence. The MANOVA procedure was used to assess the differences among the groups on the interval level data (high school GPA, ACT composite score, most recent WMU GPA, and parents AGI). The Chi-square Test of Independence was used to assess the differences among the groups on the categorical level data (taking remedial courses, taking AP credit, living in the dorm, high school attended and first year experience). Table 49 presents a summary of the MANOVA procedure. The results revealed two statistically significant differences among the three groups. These statistical differences were found in ACT composite score and most recent WMU GPA. The comparison of ACT composite scores across the three groups—on time response, late response, and no response—revealed a test statistic oiF(2, 298) =4.597 andp = .011. The magnitude of the effect size for the results was n2 =.048, which according to Cohen (1988) is a small effect. The obtained power of .773 revealed that the differences in the ACT composite scores across the three groups were large enough to be detected 77.3% of the time. The null hypothesis for this research question was rejected. A review of the descriptive statistics revealed that on-time respondents reported significantly higher ACT composite scores (M= 21.8, sd = 6.672) = than did nonrespondents (M= 19.05, sd= 6.381). There were no statistically significant differences between the scores (M= 21.04, sd = 4.975) of late respondents and the scores of the other two groups. 100 Table 49. Summary of MANOVA Comparison of Survey of Promise Scholarship Recipients at WMU Spring 2009 across the Demographic Variables Dependent Variable df F Sig. Partial Eta Observed Power Squared ACT Composite Score 2 4.597 .011 .048 High School GPA 2 1.646 .196 .018 .773 .344 Most Recent WMU GPA 2 8.796 .000 .088 .969 Parents Aggregate Income 2 1.006 .368 .011 .223 The most recent WMU GPA across the three groups revealed a test statistic of F(2, 298) =8.796 and/? = .000. The magnitude of the effect size for the results was r\ =.088, which according to Cohen (1988) is a medium. The obtained power of .969 revealed that the differences in the most recent WMU GPA for the three groups were large enough to be detected 96.9% of the time. The null hypothesis for this research question was rejected. A review of the descriptive statistics revealed that on-time respondents reported significantly higher WMU GPA (M= 3.076, sd = .672) = than did non-respondents (M= 2.542, sd = .813). There were no statistically significant differences between the scores (M= 2.856, sd = .669) of late respondents and the scores of the other two groups. The second part of research question 3.1 was examined using the Chi-square Test of Independence due to the categorical nature of these variables. A summary of the results is presented in Table 50. There were no statistically significant differences between the observed and expected frequency counts among the three groups. The null hypothesis for this research question was upheld. 101 Table 50. Test Groups Differences across the Categorical Variables / First Year Experience 4.491 a a df Asymp. Sig. (2-sided) 2 .106 Remedial courses .894 2 .640 Being an athlete 2.431a 2 .297 Living in the dorm 4.239a 2 .120 Taking AP credit 1.018a 2 .601 Gender a 4.685 2 .096 Race 2.145a 4 .709 High school attended 3.782a 4 .436 Note. a. 0 cells (.0%) have expected count less than 5. The minimum expected count is 6.36. Research Question 3.2 Results 3.2 Between the two groups of students: early respondent and late respondent Kalamazoo Promise recipients, are there any differences in the cognitive, social, or institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement using the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009 indicating possible non-response bias? Non-respondents are not used for this question as this data is not available for them. Late respondents are those who answered after the second e-mail reminder but before the third e-mail reminder. Late respondents were considered to be similar to nonrespondents for the purpose of this research. "Persons who respond in later waves are assumed to have responded because of the increased stimulus and are expected to be similar to non-respondents" (Armstrong & Overton, 1977, p. 2). This research question was addressed through the use of the MANOVA and the 102 Chi-square Test of Independence. The MANOVA procedure was used to assess the differences among the groups on the four summated subscales of the Survey of Promise Scholarship Recipients at WMUSpring 2009. Pillai's Trace was used to interpret results from the MANOVA procedure. Table 51 presents a summary of the MANOVA results. The value of Pillai's trace was not statistically significant [F (4, 73) = 1.080, p = .373]. There was not enough evidence to reject the null hypothesis for this research question. There were not statistically significant differences between the two groups for the data obtained from the Survey of Promise Scholarship Recipients at WMU Spring 2009. Table 51. Summary of MANOVA Results for Early and Late Respondents of the Survey of Promise Scholarship Recipients on the Four Summated Subscale Scores ,f „. F lg Partial Eta Squared Observed Power Dependent Variable S Social Demands Cognitive Engagement 1 2.749 .101 .035 1 .751 .389 .010 .374 .137 Institutional Engagement 1 .080 .778 .001 .059 Social Engagement 1 1.662 .201 .021 .247 ' 103 Table 52. Summary of Comparison of Early Respondents and Late Respondents across the Categorical Variables of the Survey of Promise Scholarship Recipients Variable ssDegreePursuing I2 80.9823 df 88 P .689 a ssCurrentMajor 85.987 84 .419 ssAdvancedDegree 1.972a 1 .160 ssIfYesDescribe 53.001a 52 .435 ssDescribeCareerGoals 95.662a 95 .462 ssFreeReducedLunch 1.834a 1 .176 sGenderfromSurvey a 3.414 1 .065 ssLivingWhereDuringSchoolYear 5.433a 2 .066 ss WhoLiveWithS chool Year 2.893a 6 .822 ssHighestEducationFather 4.159a 5 .527 ssHighestEducationMother a 5.827 5 .323 siLevelAwarenessPromise 5.827a 5 .323 siPercentScholarshipEligibleFor 4.515a 7 .719 scHoursPreparingForClass 5.055a 5 .409 ssHoursWorkingForPay a 3.774 5 .582 scHoursCollegeActivities 4.204a 4 .379 ssHoursDependentCare 4.849a 5 .435 ssHoursCommuting 5.175a 2 .075 a 3 .431 ssJobEffectSchoolWork 2.755 Note. a. 0 cells (.0%) have expected count less than 5. The Chi-square Test of Independence was used to assess the differences between the groups on the categorical level data. The results failed to reveal any statistically significant differences between the two groups across the various categorical variables; see Table 52. 104 Research Question 3.3 Results 3.3 Using and modifying Groves and Couper 's bias ratio formula (1998,) is there an indication of non-response bias, and is there a difference between using the mean or the median, a more robust statistic, in determining a bias estimate on the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009? The null hypothesis for this research question was addressed by using and modifying Groves and Couper's bias ratio formula (1998). The formula was applied twice: once with the means and once with the medians obtained from the descriptive data for the four subscale scores of the Survey of Promise Scholarship Recipients at WMU Spring 2009. The calculations were performed using the following formula: Bias = 1 - Response Rate % (Respondent Mean - Late-respondent Mean) Ky r) = ( l - r ) ( yr- y„r) Table 53 gives the descriptive statistics used to use the bias ratio formula. 105 Table 53. Descriptive Statistics for the Subscale Scores of the Survey of Promise Scholarship Recipients Survey Group by Response Mean On Time Respondent N 13.104 52.783 14.844 22.073 48 46 45 41 5.832 10.840 4.106 4.698 34.010 117.507 16.862 22.070 Median 12.000 52.500 14.000 21.000 Mean 11.761 55.500 15.256 24.217 N 46 44 43 46 4.321 9.648 3.971 4.511 18.675 93.093 15.766 20.352 Median 12.000 54.500 15.000 24.000 Mean 12.447 54.111 15.046 23.207 N 94 90 88 87 Std. Deviation Variance Late Respondent Std. Deviation Variance Total Social Cognitive Institutional Social Engagement Demands Engagement Support Std. Deviation Variance 5.165 10.308 4.023 4.698 26.680 106.257 16.182 22.073 Median 12.000 54.000 14.500 23.000 The results of the comparison of means are presented in Table 54. The bias estimates ranged from a low of-1.2884% to a high of 5.0853%. If the response rate is 53% and the mean Social Demands subscale for the respondents is 13.1042 and the mean for the Non-respondent is 12.4468, then the non-response error is (1-.5288)(13.104212.4468) = 0.0509. This means that the non-response bias is 5% with regards to the total sample mean for the Social Demands subscale score. The same bias percent calculations were completed for each of the subscales— 106 Cognitive Engagement, Institutional Support and Social Engagement—with the resulting percentages of 2%, 1% and 4% respectively. These results failed to reveal any large (10% or higher) differences in the response rates between respondents and nonrespondents. Table 54. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Mean Scores of the Participants Using the 53% Response Rate Response Mean Social Demands Cognitive Engagement Institutional Support Social Engagement 13.1042 52.7826 14.8444 22.0732 Late Response Mean 11.7609 55.5000 15.2558 24.2174 Total Mean 12.4468 54.1111 15.0455 23.2069 Nonresponse Bias 0.0509 -0.0237 -0.0129 -0.0435 Bias Percent 5.0853 -2.3663 -1.2884 -4.3536 Note. Response rate = 101/191 = .5288 = 53% The results of the comparison based on the median are presented in Table 55. The bias estimates ranged from a low of 0% to a high of 6.1461%. If the response rate is 53%, the mean Social Demands subscale for the respondents is 11.7610, and the mean for the Non-respondent is 11.7610, then the non-response error is (1-.5288)(11.761011.7610) = 0. This means that the bias is 0% with regards to the total sample mean for the Social Demands subscale score. The same bias percent calculations were completed for each of the subscales, with the resulting percentages of 2%, 3% and 6% respectively. These results failed to reveal any large differences in the response rates between respondents and non-respondents. 107 Table 55. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Median Scores of the Participants Using the 53% Response Rate Median Response Social Demands 11.761 Cognitive Engagement 52.500 Institutional Support 14.000 Social Engagement 21.000 Median Late Response 11.761 54.500 15.000 24.000 Total NonMedian response Bias 12.000 0 54.000 0.0175 14.500 0.0325 23.000 0.0615 Bias Percent 0 -1.7452 -3.2497 -6.1461 Note. Response rate = 101/191 = .5288 = 53% The same analysis was run using the response rate of 33% using the entire population, instead of the 53% rate of those who were e-mailed the survey and responded. The results of the comparison of means are presented in Table 56. The bias estimates ranged from a low of 3.3697%) to a high of 26.0868%. If the response rate is 33%, the mean Social Demands subscale for the respondents is 13.1042, and the mean for the Non-respondent is 12.4468, then the non-response error is (1-.3289)(13.104211.7609) = 0.0724. This means that the bias is 7% with regards to the total sample mean for the Social Demands subscale score. The same bias percent calculations were completed for each of the subscales, with the resulting percentages of 3%, 8% and 26% respectively. These results failed to reveal any large differences in the response rates between respondents and late respondents, except for the subscale Social Engagement. For the subscale of Social Engagement, there is a 26% bias percent between respondents and late respondents 108 Table 56. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Mean Scores of the Participants Using the 33% Rate Response Mean Social Demands Cognitive Engagement Institutional Support Social Engagement 13.1042 52.7826 14.8444 22.0732 Late Response Mean 11.7609 55.5000 15.2558 24.2174 Total Mean 12.4468 54.1111 15.0455 23.2069 Nonresponse Bias 0.0724 -0.0336 -0.0772 -0.2609 Bias percent 7.2418 3.3697 7.7202 26.0868 Note. Response rate = 101/307 = .3289 = 33% The results of the comparison based on the median are presented in Table 57. The bias estimates ranged from 0% to 26.0870%. If the response rate is 33%, the mean Social Demands subscale for the respondents is 11.7610, and the mean for the late respondent is 11.7610, then the non-response error is (1-.32899)(11.7610-11.7610) = 0. This means that the bias is 0% with regards to the total sample mean for the Social Demands subscale score. The same bias percent calculations were completed for each of the subscales, with the resulting percentages of, 4%, 7% and 26% respectively. These results failed to reveal any large differences in the response rates between respondents and late respondents except for the subscale; Social Engagement; for the subscale of Social Engagement, there is a 26% bias percent between respondents and late respondents. 109 Table 57. Response Bias Estimates for the Subscale Scores of the Survey of Promise Scholarship Recipients Based on Median Scores of the Participants Using the 33% Response Rate Social Demands Cognitive Engagement Institutional Support Social Engagement Median Response Median Late Total Response Median 11.761 52.500 14.000 21.000 11.761 54.500 15.000 24.000 12.000 54.000 14.500 23.000 i > UJU- response Bias 0 -0.0370 -0.0690 -0.2609 Bias Percent 0 -3.7037 6.8966 -26.0870 Note. Response rate = 101/307 = .3289 = 33% This chapter presented the summary results of both the academic data as well as survey data. Along with this, the results of all three research questions with subquestions were presented in detail. 110 CHAPTER V CONCLUSIONS This concluding chapter consists of three sections. First, the central findings of the three research questions and sub-questions are discussed. Second, suggestions for future research are enumerated. Lastly, the potential implications of these results for future studies dealing with Kalamazoo Promise recipients are discussed, with implications for WMU, KPS, the Kalamazoo community, researchers and evaluators to consider. Central Findings The central findings of this dissertation are several, as each of the three research questions had a different focus. Research question one focused on persisters, those on probation, and non-persisters in terms of the Cognitive, Social and Institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement, using the academic data obtained from the Office of Student Academic and Institutional Research. The second research question also focused on persisters, those on probation, and non-persisters, but this time in terms of the average number of courses taken per term and the number of courses taken the first year. This data was also obtained from the Office of Student Academic and Institutional Research. The last research question focused on examining non-response bias and whether there was an indication of bias in this research that would limit the generalizations that can be made back to the population of WMU Kalamazoo Promise Scholarship recipients. Ill Part one of this question examined respondents, late respondents, and non-respondents in terms of academic data obtained on all. The second part examined early respondents with late respondents, as the literature shows that late respondents are very similar to non-respondents using the Survey of Promise Scholarship Recipients at WMU Spring 2009. Persons who respond in later waves are assumed to have responded because of the increased stimulus and are expected to be similar to non-respondents" (Armstrong & Overton, 1977, p.2). Non-respondents could not be looked at using the Survey of Promise Scholarship Recipients at WMU Spring 2009. Non-respondents did not answer this survey; therefore no data on them exists from this survey. The last part of question three used the bias ratio formula of Groves and Cooper(1998) to determine a bias ratio on the dependent variables for the Survey of Promise Scholarship Recipients at WMU Spring 2009. This formula was also modified, using the median in place of the mean, in hopes of determining a better indication of bias. Because of the number and scope of these questions, the central findings are organized into three sections: one for each research question and its corresponding sub-questions. Research Question One Conclusion Difference by race. Although high school GPAs for all Kalamazoo Promise participants on average are around 3.00 regardless of probation status, there is still a significant difference between those in good standing (persisters), those on probation, and those that were academically dismissed (non-persisters) from WMU. A closer examination revealed that White students had on average significantly higher high school GPAs than Black students or students in the Other category. Also, students in the Other 112 category had significantly higher high school GPAs than Black students. Black students have the lowest high school GPAs, while White students have the highest. White students also scored significantly higher on their ACT composite score than did Black students or students in the Other category. There was no significant difference between Black students and students in the Other category on ACT composite score. The interesting thing here is not that the White students scored higher on their ACTs and had higher high school GPAs than the Black students or students in the Other racial group; rather it is that there was no significant difference found among racial groups on most recent WMU GPAs. This indicates that Western Michigan University Kalamazoo Promise recipient students perform similarly regardless of race, which is not the case for these same students in high school. This prompts the question: what is different in high school from the institutional level in terms of race? This area needs more research in order to determine exactly what is indicated here. Difference amongpersisters, those on probation, and non-persisters. By race, there was a significant difference for high school GPA and ACT composite score. Yet there were not significant differences by race for most recent WMU GPA. The absence of differences by race was found among those who persisted, those on probation, and those who did not persist. It would be expected that persisters had higher GPAs and higher ACT composite scores, just as was found in terms of race. This, however, was not the case for ACT composite score. The opposite of what holds true for race in terms of ACT composite score was found between persisters, those on probation and non-persisters. There was no significant difference found between persisters, those on probation and non-persisters in 113 terms of ACT composite score. There was a significant difference found between persisters, those on probation and non-persisters with regard to high school GPA and most recent WMU GPA. Persisters had higher high school GPAs than non-persisters. Those on probation performed similar to both persisters and non-persisters falling right in between the two groups. This means that students who persist had the highest high school GPAs. This makes sense; it is assumed that students who have higher high school GPAs are better prepared for a post-secondary education than students who have low high school GPAs. Persister also had higher most recent WMU GPAs than either those on probation or non-persisters. In addition, those on probation had higher most recent WMU GPAs than non-persisters. Persisters have the highest WMU GPAs, while non-persisters have the lowest WMU GPAs. Obviously, this would be the case as those at WMU whose GPAs are low are academically dismissed from WMU. In addition, in the group of persisters, White students had higher most recent WMU GPAs than either Black students or students in the Other category. Students from the Other category who were also persisters had higher WMU GPAs than Black students in the same category. There were little differences in the WMU GPAs of students classified as being on probation. Out of students in the non-persister category, Black students had higher WMU GPAs than non-persister students in the White or Other racial group. This suggests some Black students may have been on the edge of not being academically dismissed, and could have persisted with a little academic support. Students coming to WMU start off very similar in terms of their high school GPA, whether they persist, find themselves on probation, or do not persist: all three 114 groups average in the 3.00 range, and ACT composite scores are statistically no different. Thus, neither high school GPA nor ACT composite scores appear to be predictors of whether a student will persist or not. The second part of this question examined persisters, those on probation, and nonpersisters in terms of whether the participants had taken remedial courses at WMU. Thirty-five percent of non-persisters had taken at least one remedial course at WMU, approximately 22% of those on probation had taken remedial courses at WMU, while only 9% of the persisters reported taking a remedial course. This means that nonpersisters took the most remedial courses. In addition, 12% of persisters had taken at least one AP course in high school, while those on probation did not take any, and only 2% of non-persisters took AP courses in high school. Thus, students who take AP courses in high school are more likely to persist than students who do not. Obviously, students need to be academically prepared before they come to WMU in order to have a higher success rate with their post-secondary education. By gender and race, there were no statistically significant differences between persisters, those on probation, and non-persisters, which indicates that statistically the same number of males and females persisted, and regardless of racial group, persistence or lack thereof was evenly distributed. Research question 1.2. Although factors of retention found in the literature indicate that higher parental income and living in the dorm increase retention rates (Astin, 1999; Seidman, 2005), this was not found to be true with this population. This study determined that there was no statistically significant difference between persisters, those on probation, and non-persisters in terms of parental income or living in a dorm. 115 Research suggests that living in a dorm increases the social involvement of students, helping them stay involved with their institution and, in turn, helping to retain them. Such was not the case for WMU Kalamazoo Promise students, however. Western Michigan University and Kalamazoo Promise Scholarship recipients are a special case; all of these students have gone to high school and lived in the community for at least four years, fostering for them community connections that students from outside areas might not have. "Students who have money, if it came from their parents, also probably have high levels of social and human capital. Isolating the effects of money from cultural capital is difficult. These factors are likely to be mutually supportive in terms of retention, and a student who has both money and cultural capital will benefit in terms of social integration and institutional fit" (Seidman, 2005, p 235). These students may or may not have high levels of "social and human capital" through their parents, as the research suggests. Promise students obtain their financial support through a scholarship, which is much different from what existing research discusses. Yet despite existing research that associates retention with parental income, no such connection is evident here. Existing research also indicated that living in a dorm would make an impact on retention. There was no difference, however, among Promise students at WMU in terms of whether students lived in a dorm or not. Literature associating living in a dorm with improved retention seems more relevant in the cases of students who move away from home and lack other connections in the college community. For Kalamazoo Promise students and WMU, however, there are special circumstances to account for. Promise students attending WMU have already gone to high school in the local community. They 116 presumably have many social connections and social supports not found among students in previous research connecting dorm living with improved retention; such students are more likely to have moved away from home to a new community. Supposing WMU had decided to offer free dorm rooms to Kalamazoo Promise recipients on the assumption that dorm living would improve their retention, in keeping with previous research on the subject, the evidence indicates such policy would have made no difference in retention of this population. The Promise scholarship appears to affect this population in ways that existing factors of retention cannot predict. Research Question 1.3. There was no statistically significant difference determined in terms of which high school students attended prior to attending WMU. Statistically the same number of students from each high school succeeds or fails at WMU. This is good news for the high schools: it indicates that one high school is not doing better or worse than the other in preparing students for WMU. First Year Experience that Western Michigan University provides new students seems to make a difference, although not a strong one. Out of all of the WMU Promise students, only 37% participated in FYE; about half of those on probation and about half of non-persisters participated in it. By comparison, only about 30% of persisters participated. That only 37% of Promise recipients participated in FYE, while half of those on probation and half of non-persisters participated in it, suggests that WMU may need to rethink the program for Promise recipients and that FYE does not seem to be functioning as intended. WMU may want to examine in greater depth what services the university provides actually for Promise students. 117 Research Question Two Conclusion Research question two was answered using the course data provided by the Office of Student Academic and Institutional Research at WMU. The first sub-question looks at the average number of courses taken per term, controlling for gender and race, by persisters, those on probation, and non-persisters. It also examined the number of courses taken the first year, controlling for gender and race. The second sub-question examines the respondents, late respondents, and non-respondents of the Survey of Promise Scholarship Recipients at WMU Spring 2009 in terms of the average number of courses taken per term, controlling for gender and race. It also examined the number of courses taken the first year, controlling for gender and race. Question 2.1, Average number of courses taken per term by persistence. The difference in the average number of courses taken per term between persisters, those on probation and non-persisters controlling for gender and race was examined. Not surprisingly, there was a significant difference in the average number of courses taken per term by persisters (M= 5.13), compared with the average number those on probation (M = 4.85) and non-persisters (M= 4.65) each took. This concurs with existing research, which suggests that taking more courses per course relates to persistence. There was no difference found across gender, but there was a difference found across race. Students in the Other racial category took the most classes per term (M= 5.26), while Black students took an average of 5.00 courses per term. White students took on average the least amount of courses per term (M= 4.94). This was surprising, because minority students in question one were found to have lower high school GPAs and lower ACT composite 118 scores than White students. Question 2.1, Number of courses taken the first year by persistence. The difference in the number of course taken the first year among persisters, those on probation, and non-persisters, controlling for gender and race, was examined. Not surprisingly, there was a significant difference between the number of courses taken the first year by persisters (M = 10.17), compared with the number those on probation (M = 9.12) and non-persisters (M= 9.31) each took. Existing research indicates that when higher number of courses are taken, students are more likely to persist, and this is evident here. Question 2.2, Average number of courses taken per term by response and the number of classes taken the first year. The difference in the average number of courses taken per term and the number of courses taken the first year among respondents, late respondents, and non-respondents, controlling for gender and race, was examined. There were no statistically significant differences among respondents, late respondents, and non respondents in the average number of courses taken per term or the number of courses taken the first year. This means that non-response error is not an issue here, indicating that respondents, late respondents, and non-respondents are similar or at least statistically no different. This means that in terms of non-response bias and generalizing to the population, based on the course data given on the entire population, it can be confirmed that generalizations can be made from the responses to the population of WMU Kalamazoo Promise students. 119 Research Question Three Conclusion If it is believed that non-respondents are different from respondents in ways that are critical to the research or evaluation questions being asked, then non-response bias should be examined thoroughly in order to make accurate generalizations from the population being examined. The key phrase here is "critical to the research or evaluation questions." With the Kalamazoo Promise Scholarship being so new and having such enormous implications for Kalamazoo students, its community, and other cities replicating this universal scholarship program, it was imperative to know whether nonresponse error or bias would prevent study findings from being able to be generalized to the larger population of Promise recipients attending WMU.. Research Question 3.1 This research question looked at the three groups of students—respondents, late respondents, and non-respondents—in terms of the academic data provided by the Office of Student Academic and Institutional Research at WMU. One method for controlling non-response error is to compare respondents to non-respondents (Miller & Smith, 1983). If no statistically significant difference is found between respondents and nonrespondents on known characteristics, then the results can be generalized both to the sample and the population (Diem, 2004). In other words, if the two groups of students are similar on known variables, then assumptions can be made for the unknown variables. In examining respondents, late respondents, and non-respondents, two items were found to be statistically significant: ACT Composite Scores and Most Recent WMU 120 GPA. Respondents had higher ACT Composite Scores than the non-respondents. The late respondents, however, were statistically no different than either the respondents or the non-respondents. For the respondents to have higher scores than the other two groups might be expected; they presumably are more ambitious, as indicated by answering the survey. This is in keeping with the tendency of respondents to have taken the most number of courses, compared with non-respondents. It would have also been unsurprising for non-respondents to either have had lower scores, or, perhaps, the opposite: to have possibly much higher scores. In other words, they might be either lower-achieving and less ambitious, or higher- achieving, with little time to respond to a survey. Regardless, research suggests that late respondents and non-respondents should be similar (Armstrong & Overton, 1977, p.2). In this case, late respondents were found to be not statistically different from nonrespondents, as the research suggests. Late respondents were also found to not be statistically different from respondents, however. The scores of late respondent scores fell right between those of respondents and non-respondents. This finding suggests the population in question departs in some way from the norm described in existing literature in some way, which suggests a further avenue for research. Respondents also had higher Most Recent WMU GPAs than the non-respondents. Again as with the ACT Composite Scores, the late respondents were statistically no different from either the respondents or the non-respondents. Again as well, there was no statistically significant difference between the late respondents and either respondents or non-respondents. The Most Recent WMU GPA of late respondents fell right in between those of respondents and non-respondents. This coincides with existing research that 121 suggests that late respondents and non-respondents are similar (Armstrong & Overton, 1977, p.2). This study shows that the similarity of respondents and non-respondents may not always be the wisest assumption, despite reliance on previous literature. Late respondents in this study were similar to both respondents and nonrespondents. This would be fine if respondents and non-respondents were similar; in this study, however respondents and non-respondents are indeed different in two of the variables. This suggests that researchers seeking to ensure there is no response bias by comparing respondents with late respondents are making assumptions and generalizing to the population in ways that may not be accurate. Surprisingly, no significant difference was found among respondents, late respondents, and non-respondents on the rest of the known variables: High School GPA, Parents' Aggregate Income, First Year Experience, Remedial Courses, Being an Athlete, Living in the Dorm, Taking AP credit, Gender, Race or High School Attended. Using the course data in question 2.2 found that there was a difference across Race. Non-response error should be looked at when the variables being examined are critical to the interpretations. Here I think the course data is not as critical as the academic data. Using established literature, it can be cautiously concluded from this data that respondents to the Survey of Kalamazoo Promise Recipients at WMU 2009 are more likely to persist at WMU as they have higher ACT composite scores as well as higher WMU GPAs than do non-respondents to the survey. The question of response error and therefore possible bias is still open, as respondents are statistically different from non-respondents on two variables, which 122 would allow only for generalizations made to the sample, not to the population. However, because this study is examining the factors of retention, and the two factors ACT composite scores and most recent WMU GPAs were the only two variables found to be different between respondents and non-respondents, this may be acceptable for this population, because ACT composite scores were not found to be associated with whether a student persisted or not. By contrast, the most recent WMU GPA is an obvious variable that, of course, affects persistence. Another method for examining non-response error is by comparing early or ontime respondents with late respondents. Comparing early or on-time respondents with either late or reluctant respondents is commonly done in social science research to determine the effect, if any, of non-response on the data under consideration (Armstrong & Overton, 1977; Diem, 2004; Miller & Smith, 1983; Smith, 1984). Late respondents are not statistically different from either respondents or nonrespondents. In order for there to be no further examination into non-response bias, nonrespondents should have been found to be similar to late respondents. In this case, however, late respondents were found to be similar to both respondents and nonrespondents, which would be fine if there were no statistically significant difference between respondents and non-respondents. Further examination into non-response bias is indicated here. As it stands, the data can only be generalized to the sample of students who responded and not to the population of WMU Kalamazoo Promise recipients. Research Question 3.2 This research question looked at the two groups of students, early respondents and 123 late respondents, in terms of the cognitive, social and institutional factors from Swail's (2003) Geometric Model of Student Persistence and Achievement, using the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009. Another way of examining non-response error and possible non-response bias is by comparing early or on-time respondents with late or reluctant respondents. This is commonly done in social science research to determine the effect, if any, on nonresponse on the statistics considering (Miller & Smith, 1983; Smith, 1984). Nonrespondents and late respondents are assumed to behave similarly in research question 3.2, (Armstrong & Overton, 1977, p.2). The late respondents' data from the survey are as assumed to be similar to that of non-respondents for this question, as no data are, of course, available on non-respondents. This is commonly done in survey research due to declining response rates to surveys in the richer parts of the world (de Leeuw and de Heer, 2002). No statistically significant difference was found between the two groups, early and late respondents, in any of the four subscales of the survey: Social Demands, Cognitive Engagement, Institutional Engagement and Social Engagement. There was also no statistically significant difference found in any of the categorical variables. It could therefore be assumed that respondents and late respondents, and therefore nonrespondents, are similar. Subsequently, generalizations could be made to the population of WMU Kalamazoo Promise recipients from the data obtained from the Survey of Kalamazoo Promise Recipients at WMU Spring 2009. As can be seen, depending on the method chosen to look at non-response error, different conclusions can be made. In question 3.1 it was determined that generalizations 124 could only be made to the sample of those who responded to the survey. In question 3.2, however, using a different method to look at non-response error, it was determined that generalizations could be made to the population, and not just the sample of those who responded. Question 3.3, using yet another method to look at non-response error, offers an opportunity to clear this up. Research Question 3.3 This research question examined the dependent variables from the Survey of Promise Scholarship Recipients at WMU Spring 2009 using Groves and Couper's bias ratio formula (1998). No high bias ratios were found in any of the subscale scores of the Survey ofPromsie Scholarship Recipients at WMU Spring 2009 using Groves and Couper's bias ratio formula (1998) and the 53% response rate. No high bias ratios were found with the modified formula using the 53% response rate again and the median in place of the mean, as the median is more robust. Using the median, however, did change the percentages somewhat; see Table 58. Table 58. Results of Mean and Median used in Bias Ratio Formula at the 53% Response Rate Social Demands Cognitive Engagemei Institutional Support Social Engagement Bias Percent Mean 53% Response Rate 5.0853 -2.3663 -1.2884 -0.0435 Note. 53% response rate is the 101 returned surveys/ 191 sent out. 125 Bias Percent Median 0 -1.7452 -3.2497 -6.1461 This 53% response rate was calculated using the 101 students who answered the survey divided by the number of surveys sent out (191). The actual population of this group totals 307, however, this discrepancy occurred at the onset of the research which was explained in detail earlier under question three results. A high bias ratio was found in one of the subscale scores of the Survey of Promise Scholarship Recipients at WMU Spring 2009 using Groves and Couper's bias ratio formula (1998) and the 33% response rate. The same high bias ratio was also found with the modified formula using the 33% response rate again and the median in place of the mean as the median is more robust. Using the median, however, did change the percentages somewhat. Table 59 shows the bias percents for the mean and median using the 33% response rate, which is the 101 surveys returned divided by the population of 307; along with the 53% response rate for comparison. Table 59. Results of Mean and Median used in Bias Ratio Formula at the 53% and 33% Response Rate Social Demands Cognitive Engagement Institutional Support Social Engagement Bias Percent Bias Percent Mean Median 53% Response Rate Bias Percent Bias Percent Median Mean 33% Response Rate 5.0853 -2.3663 -1.2884 -0.0435 7.2418 3.3697 7.7202 26.0868 0 -1.7452 -3.2497 -6.1461 0 -3.7037 6.8966 -26.0870 Note. 53% response rate is the 101 returned surveys/191 sent out. 33% response rate is the 101 returned surveys/ 307 population. Using the 33% response rate indicates that generalizations of the data from the Survey of Kalamazoo Promise Recipients at WMU Spring 2009 can be made in three of the four subscales of the survey: Social Demands, Cognitive Engagement and 126 Institutional Support. Generalizations to the population cannot be made for the Social Engagement subscale, however. This information can only be generalized to the sample of those who answered the survey. The only place where there was a difference found that should be noted is with the modified bias ratio formula using the 33% response rate. This response rate is theoretical, however, as the actual response rate of the surveys sent out was 53%. With that said, generalizations from the survey can be made to the population with perhaps the exception of the items on the Social Engagement subscale. The results from the academic data obtained from the Office of Student Academic and Institutional research is on the population of WMU Promise students, however, as data was obtained on all 307 of them. Therefore these results can be generalized to the population of WMU Kalamazoo Promise recipients. Very briefly these results are: • Significant differences: o White students have higher High school GPAs and higher ACT composite scores than Black students or Other students o Students in the Other category had higher high school GPAs than Black students o Persisters had higher high school GPAs than non-persisters o Persisters had higher WMU GPAs than those on probation and nonpersisters o In the persister category, White students had the highest WMU GPA o In the probation category, there was no difference by race 127 o In the non-persister category, White students had the lowest most recent WMUGPA o Non-persisters took more remedial courses o Persisters took more AP courses o Students on probation took no AP courses o Only 37.7% of WMU Promise students participated in FYE o Half of students on probation and half of those who did not persist participated in First Year Experience program (FYE). Only 30% of persisters participated in FYE o Persisters took on average more classes per term and more classes the first year than those on probation or non-persisters o White students took on average the least amount of courses per term • No difference found o Between racial groups and WMU GPA • Interesting because a difference was found for high school GPA and ACT composite scores o Between persisters, those on probation and non-persisters in • FTIAC Cohort group • ACT composite score • Parental income • Living in a dorm • High school attended • Good news for high schools as both produced same amount 128 of persisters, those on probation, and non-persisters o Between respondents, late respondents, and non-respondents on the average number of courses taken per term or the number of courses taken the first year Some existing predictors known from a whole body of established research, such as parental income, living in a dorm, or ACT scores, are not functioning as research suggests they will. It is not replicated here. Perhaps the playing field has changed with the scholarship money in place. When financial issues are evened up, the findings of past research does not appear to hold true. Future Research Two types of future research are suggested here. The first set of suggestions is based on the data obtained for this research that gave rise to more questions and thus the need for additional research and data. The second set of suggestions for future research uses the data already gathered for this project to examine questions beyond the scope of this project. • Research using new data: o Larger sample including ALL recipients from all 26 institutions and their college retention factors. o Ideally, interviewing all recipients, regardless whether they are attending, have attended, or are no longer attending any higher education institution, to determine their individual circumstances to examine any developing trends. 129 o Investigate whether there is grading bias among racial groups at the high school level or higher education institution level. o Examine high school student records in greater detail for information on high school courses taken and preparation for higher education. High school GPA obtained from the Office of Student Academic and Institutional Research at WMU does not indicate whether it is based on higher level classes or lower level classes taken in high school. Students who took harder classes could have received lower grades, but presumably would have a higher chance of persisting at a higher education institution than would students who took lower level courses in high school. • Research using the existing data from the survey and academic data; o Non-response error • Examine each individual survey item and run through the bias ratio formula of Groves and Couper (1998). • Examine each individual variable from the academic data and run through the bias ratio formula of Groves and Couper (1998). o Compare self-reported data from the survey with the academic data obtained from the Office of Student Academic and Institutional Research. Potential Implications The implications of this dissertation are threefold. First, it offers Kalamazoo Public School district, its community and WMU suggestions on how these results could affect them. Second, it examines non-response bias and the implication of this type of 130 analysis on small-scale research as well as on evaluation. Lastly, it examines retention factors and how they functioned here. The issue for Kalamazoo school district is that that more minority and lowincome students are staying in high school because of the Promise (Miron & Cullen, 2008). Whereas many of these low-income and minority students used to drop out of high school, the evidence now suggests that they are staying in school with plans to graduate. This will have implications for WMU as well, because a high proportion of these students are likely to attend either KVCC or WMU. As more and more students graduate who perhaps otherwise would not have without the incentive of the Promise, more students will struggle with the requirements and demands of higher education. Minority students—in particular Black students and those with low high school GPAs— may need particular attention such as WMU provides with the Multicultural Affairs office or the First Year Experience. Also, supports and services currently provided by an array of community organizations offer examples that can be expanded to a larger scale involving more stakeholders. All of the needs covered in Social, Cognitive and Institutional factors that affect retention should, ideally, be addressed by community groups, schools and higher education institutions. The trends suggest that at least for the next couple of years, graduation or completion rates are likely to go up, although a higher proportion of the students may be less well prepared for attending a university. Although WMU is presently serving at-risk and minority students very well, as this population increases WMU will need to better prepare for meeting the needs of this special population in hopes of retaining them and ensuring their success. Perhaps one option could be making the First Year Experience 131 mandatory for Promise students. This dissertation determined that minority students, in particular Black students, had lower high school GPAs and lower ACT scores than White students and, therefore, are in need of services provided by WMU, the community, and KPS school district in order for them to persist at post-secondary education. Systematic services need to be in place for these students, including not only academic services but social services as well. These students are also likely to benefit from mentors and a go-to person for questions and support. Students who were in elementary or middle schools when the Promise was announced will have more time to improve their performance in school before they reach critical decisions that need to be taken in high school regarding the extent to which they will take college preparatory classes. These students will have had much more time to prepare and think about post-secondary education in terms of what they are doing in high school, where the first cohorts of Promise recipients did not have this time. It is predicted that more students will take advanced placement classes in high school; as seen in this study with this population, students who took AP courses persisted at a much higher rate than those who did not take AP courses. Until then, however, additional services and supports need to be provided to the minority and low-income students as well as students who have low high school GPAs or are first-generation college attendees. Forty percent of the WMU Promise students reported being first-generation college students. WMU also needs to encourage students to take more courses per term at WMU, as existing research indicates this as a factor of persistence and, in this study, persisters took more courses than non-persisters or students who were on probation. In addition to implications for KPS, the community and WMU to consider, 132 another implication of this dissertation is that it illustrates how non-response error and non-response bias can be examined in small scale and relatively low budget research studies and program evaluations. Determining the quality of outcomes has been and still is of utmost importance in research and evaluation work. Assuming a representative sample has been obtained, research and evaluators alike would like to generalize their findings to the population and not just to the sample for which the data was obtained. Even with a representative sample and a reasonably high response rate, response error can still occur. Taking the extra steps needed to ensure that an indication of non-response error and therefore possible response bias does not exist is a minimal measure to help ensure the accuracy of the findings. This dissertation took extra steps to determine if the results found here could be generalized to the population of the WMU Kalamazoo Promise Scholarship recipients without the concern of response error and possible nonresponse bias. It was hoped that this non-response bias analysis would help contribute to evaluation. As an example to researchers and evaluators, this dissertation illustrates how this can be done, even with relatively small numbers and a limited budget. In order to ensure the accurate use of research data, anyone making assumptions from existing research needs to be concerned with the generalizations made, how these were made, and if they are appropriate in the context in which they are used. When stakes are high, it is better to take measures to ensure accuracy. Currently, non-response error is more commonly used in large-scale research; small-scale research and program evaluation have not seen the value in this type of analysis. In order to ensure the quality and validity of the findings, this dissertation illustrated that it can easily be completed and that it is imperative to examine non-response error in order to generalize with confidence. 133 One final implication that requires attention is the finding that some of the established factors that typically predict persistence or non-persistence do not appear to function with this population as the existing literature indicates they should. Cognitive factors such as high school GPA, WMU GPA, and taking remedial or advanced placement courses did function as expected. Other factors of retention—ACT composite scores, parental income, living in a dorm, and Eirst Year Experience provided by WMU, which according to the literature would be expected to have an effect on persistence—did not. Interestingly, most of these factors are Social factors or Institutional factors, which were not found to differ among persisters, those on probation, and non-persisters. This is a fascinating finding. Factors that normally have a relationship with persistence are not exhibited here among scholarship recipients, which really asks more questions than it answers. The factors that do seem to be associated with persistence among Promise scholarship recipients are academic ones; those that do not deal with more than just academics. One possible explanation is that the funding and support provided by the Kalamazoo Promise, along with the supports provided by WMU, KPS, and the community, have mitigated or neutralized many of the factors that typically contribute to non-persistence in higher education. This possible explanation has huge implications and, of course, deserves further consideration in future research. 134 T?T5T7T3T?CW-'T7C Adelman, C. (1999). Answers in the toolbox: Academic intensity, attendance patterns, and bachelor's degree attainment. Washington, D.C.: U.S. Department of Education Office of Educational Research and Improvement. Armstrong, S.J., & Overton, T.S. (1977). Estimating non-response bias in mail surveys. Journal of Marketing Research, 14, 396-402. Ary, D., Jacobs, L.C., & Razavieh, A. (1996). Introduction to research in education (5X ed.). Fort Worth, TX: Harcourt Brace Court. Astin, A.W. (1971). Predicting academic performance in college: Selectivity data for 2300 American colleges. New York: The Free Press. Astin, A.W. (1975). Preventing students from dropping out. San Francisco: JosseyBass.Astin, A.W. (1984). Student involvement: A developmental theory for higher education. Journal of College Student Personnel, 25, 297-308. Astin, A. W. (1999). Student involvement: A developmental theory for higher education. Journal of College Student Development, 40(5), 518-529. Bean, J. P., & Eaton, S. B. (2000). A psychological model of college student retention. In Braxton, J. M., (Ed.), Reworking the student departure puzzle. Nashville, TN: Vanderbilt University Press. Bean, J.P. & Metzner, B. (1985). A conceptual model of nontraditional undergraduate student attrition. Review of Educational Research 55, (4), 485-540. Bonham, L. A. & Luckie, J.A.I. (1993). Taking a break in schooling: Why community college students stop out. Community College Journal of Research and Practice 77(3), 257-70. Braxton, J.M. (Ed.). (2004). Reworking the student departure puzzle. Nashville, TN: Vanderbilt University. Braxton, J.M., & Lien, L. A. (2004). The viability of academic integration as a central construct in Tinto's interactionalist theory of college student departure. In J. M. Brick, J. M., Bose, W., & Bose, J. (2001, Aug. 5-9). Analysis of potential nonresponse bias. Proceedings of the Annual Meeting of the American Statistical Association. Rockville, MD. Cabrera, A. F., Nora, A., & Castaneda, M.B. (1993). College persistence: Structural equations modeling test of an integrated model of student retention. Journal of Higher Education, 64(2), 123-139. 135 Cabrera, A. F., Stampen, J. O. & Hansen, W. L. (1990). Exploring the effects of ability to pay on persistence in college. The Review of Higher Education, 13(3), 303- 336. Carroll, J. (1988). Freshman retention and attrition factors at a predominantly Black urban community college. Journal of College Student Development, 29, 52-59. Cataldi, E.F., Laird, J., and KewalRamani, A. (2009). High School Dropout and Completion Rates in the United States: 2007(NCES 2009-064). National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Washington, DC. Retrieved from http ://nces. ed. go v/pubsearch/pubsinfo. asp ?pubid=2 009064 Chan, E. (2002). Patterns of full-time/part-time attendance and their effects on retention and graduation. Showcase presented at the Association for Institutional Research 42 nd Annual Forum, Toronto, Canada. Retrieved from http://ocair.org/files/presentations/paper2002_03/poster2_EvaChan.pdf Cohen, J., & Cohen, P. (1988). Applied multiple regressions/correlation analysis for the behavioral sciences (2 nd ed.). Hillsdale, NJ: Lawrence Erlbaum, Associates. Creswell, J. W. (2003). Research design: Qualitative, quantitative, and mixed methods approaches. Thousand Oaks, CA: Sage Publications. Cresswell, J. W. (2005). Educational research: Planning, conducting, and evaluating quantitative and qualitative research (2 nd . ed.). Upper Saddle River, NJ: Merrill/Prentice Hall. Cunningham, A.F. (2005). Changes in Patterns of Prices and Financial Aid (NCES 2006-153). U.S. Department of Education. Washington, DC: National Center for Education Statistics. De Heer, W. & De Leeuw, E. (2002). Trends in household survey nonresponse; A longitudinal and international comparison, in Goves R. et al. (eds.) Survey Nonresponse, New York: John Wiley, pp. 41-54. Diem, K. G., PhD. (2004). Maximizing response rate and controlling non-response error in survey research. Rutgers Cooperative Research & Extension, JAES, Rutgers, The State University of New Jersey. Eberts, R., & Kitchens, R. (2008). Communities investing in education and economic development. Kalamazoo, MI. (pp. 1-40), Retrieved from: www.upjohninst.org/promise/2008_promisenet_proceedings.pdf 136 Educational Policy Institute (EPI), (2007). Institutional Student Retention Assessment: Program manual. Retrieved from: http://www.isra-online.com/assets/ISRA_Manual.pdf Feldman, M. J. (1993). Factors associated with one-year retention in a community college. Research in Higher Education 34(4), 503-12. Gall, M. D., Borg, W.R., & Gall, J.P. (1996). Educational research: An instruction (6l ed.). Whit plains, NY: Longman. Gay, L. R., Mills, G. E., & Airasian, P. (2006). Educational research: Competencies for analysis and application (8 th ed.). Upper Saddle River, NJ: Merrill/Prentice Hall. General Accounting Office (GAO). (1995). Higher education: Restructuring student aid could reduce low-income student dropout rate. (GAO/HEHS-95-48). Washington DC: U.S. Government Printing Office. Grosset, J. M. (1991). Patterns of integration, commitment, and student characteristics and retention among younger and older students. Research in Higher Education, 32(2), 159-78. Groves, R. M., & Couper, M.P. (1998). Non-response in household interview surveys. New York: Wiley and Sons. Guloyan, E. V. (1986). An examination of white and non-white attitudes of university freshmen as they relate to attrition. College Student Journal, 20, 396-402. Hair, J.F., Jr., Anderson, R.E., Tatham, R. L., & Black, W. C. (1995). Multivariate data analysis: With readings (4 th ed.); 617-671. Englewood Cliffs, NJ: Prentice-Hall, Inc. Hagedorn, L. (2006). How to Define Retention: A New Look at an Old Problem. Los Angeles: Transfer and Retention of Urban Community College Students, 26, Dialog, ERIC, ED 493674. Hayman, D., (2007). Rising above the gathering storm: Engineering undergraduate student affairs 2007 report. College of Engineering, University of Illinois at Chicago. Retrieved from: http://www.namepa.org/region_c/programs/uic_undergraduate_student_affairs_20 07report.pdf Jayson, S. (2009, January 7). Getting the most bang for you college buck. USA Today. Retrieved from http://www.usatoday.com/news/education/2009-01-07-best-valuecolleges_N.htm 137 Johnson, R. B. & Onwuegbuzie, A., (2004, October). Mixed methods research: A research paradigm whose time has come. Educational Researcher, 33 (7). Retrieved from http://www.aera.net/uploadedFiles/Journals_and_Publications/ Journals/Educational_Researcher/Volume_33_No_7/03ERv33n7_Johnson.pdf Jorth, R., (2009). Data from the Administrator of the Kalamazoo Promise Scholarship. Kano, JVL, Franke, T., Afifi, A., & Bourque, L., (2008). Adequacy of reporting results of school surveys and nonresponse effects: A review of the literature and a case study. Educational Researcher, 37 (8) 480-490. Kaplan, R. M., & Saccuzzo, D. P. (2005). Psychological testing: Principles, applications, and issues (5 th ed.). Belmont, CA: Thomson-Wadsworth. Kilpatrick, L. A., & Feeney, B. C. (2007). SPSS for windows step by step: A simple guide and reference 15.0 update (8 th ed.). Pearson Education, Inc. Lanni, J. C. (1993, March 18-22). The longitudinal student success study: The entering student survey. Paper presented at the 17th Annual Meeting of the National Association for Equal Opportunity in Higher Education, Washington, DC. (ED350 017). Levin, J. R. & Levin, M. E. (1991). A critical examination of academic retention programs for at-risk minority college students. Journal of College Student Development 32, 323-334. Locke, L. F., Spirduso, W. W., & Silverman, S. J. (2000). Proposals that work (4th ed.). Thousand Oaks, CA: Sage Publications. Mertler, C. A., & Vanatta, R. A. (2005). Advanced and multivariate statistical methods (3rd ed.) Glendale, CA; Pyrzcak Publishing. Miller-Adams, M. (2009). The power of a promise education and economic renewal in Kalamazoo. W.E. Upjohn Institute Employment Research, Kalamazoo, Michigan. Miller, L. E., & Smith, K.L. (1983). Handling nonresponse issues. Journal of Extension, 27(5), 45-50. Miron, G., & Cullen, A., (2008, October). Trends and patterns in student enrollment for Kalamazoo public schools working paper #4. Western Michigan University: The Evaluation Center. Retrieved from http://www.wmich.edu/evalctr/promise/documents/WorkingPaper4.pdf 138 Miron, G., & Evergreen S. (2007, January). The Kalamazoo Promise as a catalyst for change in an urban school district: A theoretical working paper #1. Western Michigan University: The Evaluation Center. Retrieved from http://www.wmich.edu/evalctr/promise/documents/WorkingPaperl.pdf Miron, G., Spybrook, J., & Evergreen S. (2008, April). Key finding from the 2007 survey of high school students. Evaluation of the Kalamazoo promise working paper #3. Western Michigan University: The Evaluation Center. Retrieved from http ://www. wmich. edu/evalctr/promise/documents/WorkingPaper3 .pdf Moore, N. (1995). Persistence and attrition at San Juan college. Farmington, NM: Office of Institutional Research, Grant Development, and Planning, San Juan College. (ED 380 159). Naretto, J. A. (1995). Adult student retention: The influence of internal and external community. NASPA Journal, 32(2), 90-97. National Center for Education Statistics (NCES). (June, 1998). First generation students: Undergraduates whose parents never enrolled in postsecondary education. (U.S. Department of Education Office of Educational Research and Improvement/ NCES 98-082). Washington DC: U.S. Government Printing Office. National Court Reporters Association (NCRA). (May, 2006). Report education commissioner: report to the members. Retrieved from: http://www.ncraonline.org/NR/rdonlyres/9370CllD-8D57-41CE-BB883D94439AD3AC/0/rec_report.pdf Panos, R. J. & Astin, A. W. (1968). Attrition among college students. American Education Research Journal, 5, 57-72. Pascarella, E. T., & Terenzini, P. T. (1980). Predicting freshman persistence and voluntary dropout decisions from a theoretical model. The Journal of Higher Education, 51(1), 60-75. Price, L. A. (1993). Characteristics of early student dropouts at allegany community college and recommendations for early intervention. Cumberland, MD: Allegany Community College, 1993. (ED 361 051). Rendon, L. I., Jalomo, R. E., & Nora, A. (2004). Theoretical considerations in the study of minority student retention in higher education. In J. M. Braxton (Ed.), Reworking the student departure puzzle (p. 127-156). Nashville, TN: Vanderbilt University Press. 139 Rosenberg, K. M. (2007). The excel statistics companion. Belmont, CA; Thomson Higher Education. Seidman, A., (ed.) (2005). College student retention: formula for student success. Westport, CT ACE/ Praeger. Smith, T. W. (1984). Estimating non-response bias with temporary refusals. Sociological Perspectives, 27 (4), 473-489. Spady, W. G. (1970). Dropouts from higher education: An interdisciplinary review and synthesis. Interchange, 7(1), 64-85 Sprinthall, R. C. (2007). Basic statistical analysis (8 th ed.). Boston, MA: Pearson Allyn & Bacon. Stage, F. K. (1989). Motivation, academic and social integration, and the early dropout American Educational Research Journal, 26(30, 385-402. St. John, E. P. (1990). Price response in persistence decisions: Analysis of the high school and beyond senior cohort. Proceedings for the Seventh Annual Conference of the NASSGP/NCHELP Research Network, 29-56. New Jersey Higher Education Assistance Authority: Trenton. St. John, E. P., Cabrera, A. F., Castaneda, M. B., Nora, A., & Asker, E. (2004). Economic influences on persistence reconsidered: How can finance research inform the reconceptualization of persistence models? In J. M. Braxton (Ed.), Reworking the student departure puzzle (p. 29-47). Nashville, TN: Vanderbilt University Press. Stevenson, J. P. (2007.) Applied multivariate statistics for the social sciences (5 ed.).New York, NY: Routledge. Stoecker, J., Pascarella, E., & Wolfe. L. (1988). Persistence in higher education: A 9-year test of a theoretical model. Journal of College Student Development, 29, 196-209. Swail, W., Education., A., ERIC Clearinghouse on Higher Education, W., & George Washington Univ., W. (2003, January 1). Retaining Minority Students in Higher Education: A Framework for Success. ASHE-ERIC Higher Education Report. Jossey-Bass Higher and Adult Education Series. (ERIC Document Reproduction Service No. ED483024) Retrieved from ERIC database. Terenzini, P. T., & Pascarella, E. T. (1980). Toward the validation of Tinto's model of college student attrition: A review of recent studies. Research in Higher Education, 12, 271-282. 140 Thomas, R. O. (1990). Programs and activities for improved retention. In Hoissler, Bean,and Associates, The strategic management of college enrollments: Chap. 11, 186-201. San Francisco: Jossy-Bass. Thorndike, R. L. (1967). Reliability. In D. N. Jackson & S. Messick (Eds.) Problems in human assessment (201-214). New York, NY: McGraw-Hill Book Company. Tinto, V. (1975). Dropouts from higher education: A theoretical synthesis of recent research. Review of Educational Research, 45, 89-125. Tinto, V. (1987). Leaving college: Rethinking the causes and cures of student attrition. Chicago: University of Chicago Press. Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition. Chicago, IL: Chicago Press. Trochim, W. MN. K, & Donnelly, J. P. (2007). Research methods knowledge base (3ed.). Macon, OH; Thomson. U.S. Census Bureau. (2009). Kalamazoo county quick facts from the U.S. census bureau. Washington, DC: Author. Retrieved from http://quickfacts.census.gov/qfd/states/26/26077.html Western Michigan University. (2009). A comprehensive report of retention rates. Office of Student Academic and Institutional Research. Unpublished report. Kalamazoo, MI. Windham, P. (1994, August 1-3). The relative importance of selected factors to attrition at public community colleges. Paper presented at the 23rd Annual Conference of the Southeastern Association for Community Colleges, Savannah, Georgia. (ED 373 833). Yung, K. (2007, August 20). Kalamazoo school chief leaves with promise secret intact. Detroit Free Press. Retrieved from: http://www.usatoday.com/news/education/ 2007-08-20-kalamazoo-promise_N.htm 141 Appendix A Participant Paperwork Survey of Promise Scholarship Recipients at WMU Spring 2009 (Note: This survey was formatted for use as an online survey, rather than a paper and pencil survey.) Please complete the following survey to help us understand the impact of the Kalamazoo Promise on Kalamazoo Public Schools as recent graduates. Your responses also will help us understand how improvements can be made to ensure that all students will have the opportunity to benefit from this scholarship program and succeed in college. The survey is part of a dissertation study on the factors related to success in college. This is an anonymous survey, so please do NOT write your name or any identifying information on the questionnaire. Thank you for your assistance. B\CK(iROlM)IMOK\l\lION Which high school did you attend? 1. • Loy Norrix In which year did you graduate high school? • Kalamazoo Central • 2006 • 2007 • Phoenix • 2008 2. In which month and year did you start studying at Western Michigan University? _ 3. Did you begin college at Western Michigan University (WMU) or elsewhere? • Started at WMU • Started elsewhere. Where? 4. If you graduated in 2006 or 2007, describe where you have gone to school or what you did prior to the current school year. 5. What is your classification in college? •Freshman/first-year • Sophomore • Junior 142 • Senior • Unclassified 6. How many credit hours are you taking this term? credits 7. Please estimate the total credit hours you have earned at WMU, not including those you are taking this semester. credits 8. When do you most frequently take classes? (Mark only ONE) • Day classes (morning or afternoon) • Evening classes • Weekend classes 9. What degree(s) do you wish to pursue? 10. What is your current major? 11. Do you expect to enroll for an advanced degree when, or if, you complete your undergraduate degree? • No • Yes If yes, describe the next degree you wish to pursue. 12. Briefly describe your career goals. 13. When enrolled in high school, did you qualify for the free/reduced lunch program at your school? • No • Yes 14. What is your gender? • Female 15. What is your marital status? separated • Male Qnot married • married • divorced • • widowed 16. What is your race/ethnicity? (Mark the one ethnic group with which you most identify) • American Indian or other Native American • Asian, Asian American, or Pacific Islander 143 • Black or African American • White (non-Hispanic) • Mexican or Mexican American • Puerto Rican • Other Hispanic or Latino • Multiracial • I prefer not to respond • Other: 17. Is English your native (first) language? • No • Yes 18. Where do you live during the school year? Q Dormitory or other campus housing • Residence (house, apartment, etc.) within walking distance of Western • Residence (house, apartment, etc.) within driving distance • Fraternity or sorority house 19. With whom do you live during the school year? (Fill in all that apply) • No one, I live alone • One or more other students • My spouse or partner • My child or children • My parents • Other relatives _J Friends who are not students at WMU G Other people, who 20. What is the highest level of education obtained by your: Father or Male guardian Not a high school graduate High school diploma or GED Some college, did not complete degree Associate degree Bachelor's degree Master's degree Doctorate degree and or Professional Degree gree Unknown 144 • • • • • •• • Mother or Female guardian • • • • • • • • 21. Estimate your grade point average (GPA) in high school Straight "A "s are equivalent to a 4.0 GPA. A "B " average would be 22. Estimate your grade point average (GPA), thus far at equal to 3.0 GPA. A "C" WMU 23. Please rate your level of awareness about the Kalamazoo Promise. Not at all Very Familiar familiar © © © 24. What additional information would you like to have regarding the Kalamazoo Promise? Length of Attendance....Benefit 25. How much tuition scholarship are you eligible for under the Kalamazoo Promise? (Please fill in the blank) 0/ /o 145 K-12 ....100% 1-12 95% 2-12 ....95% 3-12 ....95% 4-12 ....90% 5-12 ....85% 26. About how many hours do you spend in a typical 7-day week doing each of the following? IS on limn s p( r week 1-5 6- 11- 21-' 3 i Preparing for class (studying, reading, writing, rehearsing or a 0 © © © ® © 0 © © © ® © 0 © © © ® © 0 © © © ® © 0 © © © ® © other activities related to your program b Working for pay Participating in college-sponsored activities (organizations, c campus publications, student government, intercollegiate or intramural sports, etc.) Providing care for dependents living with your (parents, d children, spouse, etc.) e Commuting to and from classes 27. If you have a job, how does it affect your school work? • I don't have a job • My job does not interfere with my school work • My job takes some time from my school work • My job takes a lot of time from my school work 28. How likely is it that the following issues would cause you to withdraw from class or f r o m WMU?(MarA: the most appropriate response for each item, where l=Not Likely & 5=Verv Likely.) 1 , ' , . . ? " *• ,, Likely" ,. -'f,,* Likely'i © © © © © 6 Caring for dependents © © © © © c © © © © © © © © © © © © © © © a Working full-time Academically unprepared d Lack of finances e Don't fit in 146 / © Don't offer program of study that I want 29. Are you a member of a social fraternity or sorority? © © © © • No • Yes 30. Are you a student athlete on a team sponsored by WMU's athletics department? • No • Yes If yes, on what team(s) are you an athlete (i.e., football, swimming)? 31. Are you involved in student associations or organizations, if so please list: 32. How supportive are your friends of your attending WMU? • Not Very • Somewhat • Quite a bit • Extremely 33. How supportive is your immediate family of your attending WMU? • Not Very • Somewhat • Quite a bit • Extremely 34. Mark the box that best represents the quality of your relationships with people at WMU. Your relationship with: a. Other Students Unfriendly, unsupportive, Friendly, supportive sense of alienation ID 2Q 3Q 4U 5Q 6U sense of belonging b. Instructors Unavailable, unhelpful, unsympathetic Available, helpful !• 2Q 3Q 4Q 5Q 6Q sympathetic c. Administrative Personnel & Office Staff Unhelpful, inconsiderate, 1Q Rigid 2Q 3Q 4Q 5Q 6Q Helpful, considerate, flexible 35. In your experiences at WMU during the current school year, about how often have you 147 done each of the foiiowing? (Mark the most appropriate response for each item, where l=Never & 5=Very Often.) m m ! K * i Asked questions in class or contributed to class #5! & ,-w © © © © © Q © © © © © © © © S © © © © © © © © © © © © © © © Q a 'OfferiS *» ^* ^, i discussions b Made a class presentation Come to class without completing readings or c assignments Worked with classmates outside of class to prepare d class assignments e g Tutored or taught other students (paid or voluntary) Participated in a community-based project as a part Used instant messaging to work on an assignment h Used e-mail to communicate with an instructor © © © © © © © © i Discussed grades or assignments with an instructor © © © © © J Talked about career plans with an instructor or Discussed ideas from your readings or classes with Received prompt feedback (written or oral) from © © © © © © © ft) © (D © © © © © © © © © © © © © © © © © © © © © © © © © f k I Worked harder than you thought you could to meet m (2) © © © © © © © © © © an instructor's standards or expectations Worked with instructors on activities other than n coursework 0 Discussed ideas from your readings or classes with P ' Had serious conversations with students who differ q Had serious conversations with students of different race or ethnic background than your own r Skipped class 148 Included diverse perspectives (different races, s religions, genders, political beliefs, etc.) in class © © © © © © © © ® © discussions or writing assignments t Put together ideas or concepts from different courses 36. To what extent does WMU emphasize each of the following? (Mark the most appropriate response for each item, where 1 = Very Little & 5=Very Much.) Tart eX- " '^-J 3V J- , i -. *<4,fti - ^Muca -»-* % Jtfav» "» "T © © © © © © © © © © c Encouraging contact among students from different economic, social, and racial or ethnic backgrounds © © © © © d Helping you cope with your non-academic responsibilities © © © © © © © © © © © © © © © © © © © © a Spending significant amounts of time studying Providing the support you need to help you succeed b academically e Providing the support you need to thrive socially Attending campus events and activities (special speakers, f cultural performances, athletic events, etc.) g Using computers in academic work 37. During the current school year, about how often have you done each of the following? (Mark the most appropriate response for each item, where l = Very Little & 5—Very Much.) i C, • * \\-t ' T *" "Seldom Attended an art exhibit, play, dance, music, theater, or other performance © © Exercised or participated in physical fitness activities © 149 •is Often. © © © © © Participated in activities to enhance your spirituality (worship, meditation, prayer, etc.) Tried to better understand someone else's views by imagining how an issue looks from his or her perspective Learned something that changed the way you understand an / © (2) © © © © © © issue or concept : 38. Overall, how would you evaluate the quality of academic advising you have received at WMU? • Poor • Fair • Good • Excellent 39. How would you evaluate your entire educational experience at WMU? • Poor • Fair • Good • Excellent 40. If you could start over again, would you go still attend WMU? • Definitely no • Probably no • Probably yes • Definitely yes 41. Would you recommend WMU to a friend or family member? • No UYes 42. What can WMU do better? QUESTIONS ABOUT KAI. \M AZOO PI BUC SCHOOLS 43. To what extent do you agree or disagree with the following statements about your high school? (Mark the most appropriate response for each item, where l=Strongly Disagree & 5=Strongly Agree.) Strongly Strongly Do Disagree Agree n't Kn •$<wM TEA CHER-STUDENT RELA TION'S 150 A Teachers were patient when a student had trouble learning © © © © B Teachers made extra efforts to help students © © © © © o C Teachers understood and met the needs of each student © © © © © V) D Teachers were fair to students © © © © /= ^ () STbDENT ACADEMIC ORIENTATION Students at my high school understood why they were in A © © © © 0 © © © © O school At my high school, students were interested in learning B new things Students at my high school had fun but also worked hard C © © © © a ') © © © © (i () on their studies Students at my high school worked hard to complete their D school assignments STUDENT ASPIRATIONS - " >•',? i A Getting good grades in high school was important to me © © © © a O b I pushed myself in high school to do better academically © © © © Q () c In high school, I believed that I could be successful © © © © () d Going to college was important to my future © © © fc O fTEACHEis' EXPECTATIONS *OF, STUDENTS My high school teachers believed that I would graduate a © © © © (5 0 © © © © a O © © © © ft P from high school My high school teachers believed that I would succeed in b college c My high school teachers had high expectations of me in 151 class d © © © © I had a teacher who was a positive role model for me GUIDANCE/ COLLEGE READINESS ' - a Teachers or counselors encouraged students to think about b Teachers or counselors helped students plan for future c Teachers or counselors helped students with personal d Students at my high school could get help and advice from e I received the assistance I needed to go to college f My high school prepared me well for my future © © © © © © © © © ® © ® © o . © © © © © © © © © © © © © () © , O'J © Ov © ; o © 0© 0 44. What changes could be made by Kalamazoo Public Schools to better prepare students for college and other post-secondary options. Q l LSI IONS \ B ( ) l I IIIF KA! VM VZOO PROMISL 45. To what extent do you agree or disagree with the following statements regarding the Kalamazoo Promise? (Mark the most appropriate response for each item, where l=Strongly Disagree & 5= Strongly Agree.) 1 *> , £ i l~;, , , -s. strongly** i * > ; .*|Strongl V Disagree! U *• ' *l C * V "'' s V '•» ' i y Jt / *^ •• fcrf i .'>• *v-s-* & Agree •^."'I-^.V'-/;^ Teachers and/or school staff at my high school spoke to me about the Kalamazoo Promise 152 © © © d My parents/guardians have spoken with me about the Kalamazoo © © © © © © © © d (J) (|) (J) (4) (5 © © © © d © © © © d © © © © d h I was confident before the Promise that I could afford to go to college, © © © © d i I wasn't sure that I could afford college before the Promise. I didn't © © © © d J I still am not sure if I can afford college, because I am not eligible for © © © © d © © © © Promise My parents/guardians encouraged me to work harder in school because d of the Promise The Kalamazoo Promise gives me more flexibility about which college The Promise hasn't really made a difference to my educational goals or plans / I changed my career goals because of the Kalamazoo Promise I worked harder in high school because I knew that the Promise would pay for college I wanted to go to college even before the announcement of the d Kalamazoo Promise C I 1 \ \ ( . I > 1 ) I L TO I H K K \ L \ M \ / O O P K O M I S L 46. To what extent do you agree or disagree with the following statements « ' ' »*Strph*gly-^',vXf, - V&SfrongU T Disagree lV. < .*"^ f T1 % ""|l . y «* . ;C •^.AgreK* a My attendance in high school improved © © © © © & More academic support was provided after school © © © © © c My school started offering more college prep courses © © © © © 153 .* ; d I enrolled in more college prep courses © © © © e Teachers expected that more students would go to college © © © ® © f The amount of homework increased © © © © © g Students became better behaved and were getting into less trouble © © © © © h More information was provided about higher education opportunities © © © © © i My peers were more motivated to succeed in school © © © © © J I talked about college more often with peers © © © © © k The quality of student academic performance improved © © © © © © © © © © More support from community organizations was provided to students I and families 47. Describe changes in your family as a result of the Kalamazoo Promise 48. How has the Kalamazoo Promise changed your life? Thank you for your assistance with this survey! Interview Protocol - Kalamazoo Promise Students Hi, I am (your name), thank you so much for coming today for this interview. Let's go over the study information sheet and the consent form first and then we can start the interview. (Go over informed consent, giving participant a copy to keep and a copy to sign.) Project: Kalamazoo Promise Scholarship Recipients: A Comparative Analysis of Higher Education Retention and Non-response Bias Date: Time of Interview: Location: 154 Interviewer: Interviewee: Semi-Structured Questions: ' . -_ Background r ..-'4&^^..',l^. *..•:.: Ii2' Name: Age: Gender: Race: Last School: 1. Are you currently attending college? a. If yes, where? b. If no, do you plan to? i. If yes, how, where, when? ii. If no, why not? \Note: Participants who indicate that they have left WMU should answer questions 2-4. Those that are still at WMU should skip to question 5.] 2. What were the circumstances that influenced your decision to leave WMU? 3. Is there anything someone could have done to help you stay in college? 4. Did you know that you could return to college and still use the Kalamazoo Promise Scholarship? 5. Where do get information regarding the Kalamazoo Promise? 6. Do you have a support system around you? Please describe this support system. Cognitive Factors 7. Do you (or did you) enjoy college? a. What did you enjoy the most? b. What did you enjoy the least? 8. I am going to ask you to rate the difficulty level of various things on a scale of 15, where a 1 means "very difficult" and a 5 means "very easy." A 3 would be in the middle, which would indicate that it was just right. [Show illustration of the 155 scale to the interviewee.] You are welcome to provide examples or explain the reasoning for the rating you give. [Ask for examples when the score is 1 or 5.] a. How would you rate course work at WMU? b. How would you rate course work at your high school? i. [If there is a difference ask them to explain] c. For you studying is... d. For you learning is... Social Factors e. Making new friends at college is/was... 9. Did getting the Kalamazoo Promise Scholarship solve your financial issues as they relate to college expense? Please explain. 10. How do you feel about learning? 11. Are/were you involved in any extra curricular activities at WMU? Please explain. 12. Are you involved in your community? Please explain. 13. Do you have goals? If so, can you share them. 14. What does your family think about your decision to attend (not attend) WMU? 15. What do your friends think about your decision to attend (not attend) WMU? 16. Please finish the following sentences... a. On the weekend I like to... b. After a long day at work I like to... c. My friends are the... d. My family always... e. If I could do anything I wanted I would... Institutional Factors 17. Did WMU provide you with the services that you needed? Please explain. 18. Did you attend an orientation? 156 19. How do you feel about the admissions process at WMU? 20. How do you feel about the instructors at WMU? 21. How do you feel about the classes you took at WMU? 22. How do you feel about the programs offered at WMU? 23. Is there anything WMU could do better? Please explain. Kalama/oo Promise 24. What changes would you suggest to KPS to help more students get what they need so they can be prepared for college? (Should they be doing anything differently?) a. Academically b. Socially c. Other support (health, counseling, mentoring, tutoring, technology) 25. Is there anything else you are thinking about with the Promise that I haven't asked you about yet? This completes the interview. Do you have any follow-up questions or comments? Thank you for participating in this interview. Remember, if you have any follow-up comments, concerns, or questions, please contact me or Dr. Miron. Our contact information is on the consent form. Give participant their phone card and thank them again. 157 Interview Rating Scale Rating Difficulty Level m Very Difficult No Change 158 Very Easy V UMi K-HK:. :«^- S £i::>1w-few«:.&ir. Date: Marchi8,20G9 To: Gary Miron, Principal Investigator Michelle Ann Bakerson. Student Investigator for dissertf-Uor; From: Amy Nauglc, P a J ^ ^ Q h a i r M f W W « W ~ Re: HSIRBProjectNumber. 09-03-10 This letter will serve as confirmation that your research project entitled "Kstaaasoo Promise Scholarship Recipients: A Comparative Analysis of Higher Edsf&iioti Retesnion aadNonrespoase Bias" has been approved tinderfeeexpedited category si" rsstew ty the Human Subjects Institutional Review Board. The conditions and durc.&>n z>T this approval arc specified in the Policies of "Western Mchigan University. Yc« sanrnow begin to implement the research as described in tie application. Please note that yon may only conduct this research exactly in the fern-; it «*5 approved. You most seek specific board approvalforany changes in mis project Ycx r.«;si ?Jso seekreapprovalif me project extends beyond meterminationdate noted itJcv.% la addition if there are any unanticipated adverse reactions or unanticipated events associated with the conduct of this research, you should immediately rasp-tec' tfte project and contact the Chair of the HSIRB for consultation. The Board wishes you successin the pursuit of your research goals. Approval Termination: March 18,2010 mm 159 Vkbsxt Hall, Kateraaw, M! 4900S-i<SS tK9)28742S3 ft& (289)387-82/6 Date: March 18, 2009 To: Gary Miron, Principal Investigator Michelle Ann Bakerson, Student Investigator for dissertation From: Amy Naugle, Ph.D., Chair Re: HSIRB Project Number: 09-03-10 This letter will serve as confirmation that your research project entitled "Kalamazoo Promise Scholarship Recipients: A Comparative Analysis of Higher Education Retention and Nonresponse Bias" has been approved under the expedited category of review by the Human Subjects Institutional Review Board. The conditions and duration of this approval are specified in the Policies of Western Michigan University. You may now begin to implement the research as described in the application. Please note that you may only conduct this research exactly in the form it was approved. You must seek specific board approval for any changes in this project. You must also seek reapproval if the project extends beyond the termination date noted below. In addition if there are any unanticipated adverse reactions or unanticipated events associated with the conduct of this research, you should immediately suspend the project and contact the Chair of the HSIRB for consultation. The Board wishes you success in the pursuit of your research goals. Approval Termination: March 18, 2010 160 Study Information Sheet/Consent Form for Online Survey Western Michigan University Department of Educational Leadership, Research and Technology Dr. Gary Miron, Principal Investigator Michelle Ann Bakerson, Student Investigator Title: Kalamazoo Promise Scholarship Recipients': A Comparative Analysis of Higher Education Retention and Non-Response Bias You are invited to participate in a study entitled, "Kalamazoo Promise Scholarship Recipients: A Comparative Analysis of Higher Education Retention and Non-Response Bias." The study is being conducted by Michelle Ann Bakerson, a doctoral student in the Evaluation, Measurement and Research doctoral program at Western Michigan University, under the direction of Dr. Gary Miron, her dissertation chair. The following information is being provided for you to decide whether you wish to participate in this study as well as to inform you that you are free to decide not to participate in it, or to withdraw at any time, without affecting your relationship with the researchers, Western Michigan University or the Kalamazoo Promise Scholarship. The purpose of the study is to examine retention factors of Kalamazoo Promise recipients who are attending and who have attended Western Michigan University (WMU) to determine if there is a difference between those students that WMU retained and those it did not retain. If you agree to participate you will be asked to complete an on-line survey regarding your experiences in high school and at WMU. It should only take you about 20 minutes to complete and you will be one of approximately 300 subjects to participate. Your name will not be associated with the research findings in any way, and your identity as a participant will be known only to the researcher. The survey information that you provide will be maintained online secured with a password and then when data is aggregated with no identifying information it will be located in a locked file cabinet in the residence of the researcher for a period of three years. At that time all data will be destroyed. There are no known risks and/or discomforts associated with this study. While there are no direct benefits to you from participating in the study, we hope to learn about your academic experiences and the quality of education and support provided by WMU and KPS may improve based on these findings. You also have the option to enter a random drawing to receive one often $20 WMU book store gift cards if you choose to fill out the survey. You can withdraw from participating at anytime and can skip any question you do not wish to answer. This will 161 not affect your opportunity to enter the drawing or your chances of receiving one of the gift cards. If you have any questions about this study, you may contact the primary researcher, Michelle Ann Bakerson, M.A., at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at michelle.a.bakerson(a),wmich.edu. You may also contact the Dissertation Chair, Gary Miron, Ph.D., (269-387-3883), Human Subjects Institutional Review Board (269387-8293) or the Vice President for Research (269-387-8298) if questions or problems arise during the course of the study. This consent document has been approved for use for one year by the Human Subjects Institutional Review Board (HSIRB) as indicated by the stamped date and signature of the board chair in the upper right corner. Do not participate in this study if the stamped date is older than one year. If you decide to participate and give your consent, please put a check in the box below [this box will be online] and continue on to answer the questions of this electronic survey. 162 Study Information Sheet/Consent Form for Interview Western Michigan University Department of Educational Leadership, Research, and Technology Dr. Gary Miron, Principal Investigator Michelle Ann Bakerson, Student Investigator Title: Kalamazoo Promise Scholarship Recipients: A Comparative Analysis of Higher Education Retention and Non-Response Bias You are invited to participate in a study about the "Kalamazoo Promise Scholarship Recipients': A Comparative Analysis of Higher Education Retention and Non-Response Bias." The study is being conducted by Michelle Ann Bakerson, a doctoral student in the Evaluation, Measurement and Research doctoral program at Western Michigan University, under the direction of Dr. Gary Miron, her dissertation chair. The following information is being provided for you to decide whether you wish to participate in this study as well as to inform you that you are free to decide not to participate in it, or to withdraw at any time, without affecting your relationship with the researchers, Western Michigan University or the Kalamazoo Promise Scholarship. The purpose of the study is to examine retention factors of Kalamazoo Promise recipients who are attending and who have attended Western Michigan University (WMU) to determine if there is a difference between those students that WMU retained and those they did not retain. If you agree to participate you will be asked to participate in an in depth interview regarding your experiences in high school and at WMU. This interview will be held on WMU's campus and should take about 50 minutes to complete. You will be one of about 72 subjects to participate. Digital audio recording equipment will be used to ensure accuracy of the information received and written transcripts of all interviews will be produced. You may request the interviewer to turn off the audio recorder at any time during the interview. Your interview will be given a number for transcription. Only the interviewer will know your name. Your name will not be on the transcription and your name will not be reported in any way. There will be no way to identify any individual student. 163 Do not hesitate to ask any questions about the study either before participating or during the time that you are participating. Your name will not be associated with the research findings in any way, and your identity as a participant will be known only to the researcher during the interview. The digital recorders will be kept in a locked filing cabinet in the researcher's home until transcribed. No one except the researcher will have access to these recorders. Once transcription has been completed the digital interview recordings will be deleted immediately. No record of the audio files will be kept once the transcription is completed. The transcription will be kept in a locked filing cabinet in the doctoral researcher's home. The consent forms will be kept in a manila envelope and locked in a filing cabinet in the doctoral researcher's home as well. No one except the researcher will have access to this cabinet. After the study's completion the data will be stored at WMU in Dr. Miron's office in a locked filing cabinet. In three years these consent forms, transcriptions and data will be disposed of by burning. There are no known risks and/or discomforts associated with this study. While there are no direct benefits to you from participating in the study we hope to learn about your academic experiences. As a thank you for participating you will be given a $20 WMU book store gift card that can be used for books, merchandise, phone cards or beverages at the WMU book store. You can withdraw from participating at anytime and can skip any question you do not wish to answer. This will not affect your receiving the gift card. If you have any questions about this study, you may contact the primary researcher, Michelle Ann Bakerson, M.A, at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at [email protected]. You may also contact the Dissertation Chair, Gary Miron, Ph.D., (269-387-3883), Human Subjects Institutional Review Board (269-387-8293) or the Vice President for Research (269-387-8298) if questions or problems arise during the course of the study. This consent document has been approved for use for one year by the Human Subjects Institutional Review Board (HSIRB) as indicated by the stamped date and signature of the board chair in the upper right corner. Do not participate in this study if the stamped date is older than one year. A copy of this consent form will be given to you to keep for your own records. Participant Date Interviewer/Researcher Date 164 Introduction Survey Jb-mail Protocol: Kalamazoo Promise Recipients Hi "Name of Student", My name is Michelle Bakerson and I am a doctoral student at WMU. I am conducting a study on Kalamazoo Promise Scholarship Recipients. I am a doctoral student in the Evaluation, Measurement and Research doctoral program at the College of Education. I would love for you to click on the following link (http: ) to complete a brief survey so I can gather information about recipients of the Kalamazoo Promise who are attending or have attended WMU. You of course are free to participate or not. The choice you make will not affect your relationship with the researcher, Western Michigan University or the Kalamazoo Promise Scholarship at all. If you are interested in learning more, you will be given more information about the study and your rights before you take the survey. The survey needs to be completed by April 10th and will only take about 20 minutes to fill out. At the end of the survey you will be given the opportunity to enter a drawing to receive one often WMU book store gift cards worth $20 that can be used for anything at the WMU book store, such as; books, magazines, merchandise, phone cards or beverages at the WMU book store. If you have any questions about this study, you may contact me at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at michelle.a.bakerson(a),wmich.edu. Thank you so much for your time and I look forward to hearing from you. Thank you, Michelle Bakerson Doctoral Student WMU 165 First Reminder Email to l ake Survey Hi "Name of Student", My name is Michelle Bakerson and I am a doctoral student at WMU. I sent you an e-mail last week asking for your participation in a research project I am doing to finish my graduate degree. Your participation it important to my research and won't take much time at all. Please consider going to the following link (http: ) and taking a brief survey about Kalamazoo Promise Scholarship Recipients. . You of course are free to participate or not. The choice you make will not affect your relationship with the researcher, Western Michigan University or the Kalamazoo Promise Scholarship at all. If you are interested in learning more, you will be given more information about the study and your rights before you take the survey. The survey needs to be completed by April 10th and will only take about 20 minutes to fill out. At the end of the survey you will be given the opportunity to enter a drawing to receive one often WMU book store gift cards worth $20 that can be used for anything at the WMU book store, such as; books, magazines, merchandise, phone cards or beverages at the WMU book store. If you have any questions about this study, you may contact me at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at [email protected]. Thank you so much for your time and I look forward to hearing from you. Thank you, Michelle Bakerson Doctoral Student WMU 166 Second Reminder Email to Take Survey Hi "Name of Student", It's me again. My name is Michelle Bakerson and I am a doctoral student at WMU. I have sent you a couple of e-mail requests to complete a survey regarding Kalamazoo Promise Recipients. Please consider completing this quick survey at the following link (http: ). Your help is greatly appreciated. You of course are free to participate or not. The choice you make will not affect your relationship with the researcher, Western Michigan University or the Kalamazoo Promise Scholarship at all. If you are interested in learning more, you will be given more information about the study and your rights before you take the survey. The survey needs to be completed by April 10th and will only take about 20 minutes to fill out. At the end of the survey you will be given the opportunity to enter a drawing to receive one often WMU book store gift cards that can be used for anything at the WMU book store, such as; books, magazines, merchandise, phone cards or beverages at the WMU book store. If you have any questions about this study, you may contact me at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at [email protected]. Thank you so much for your time and I look forward to hearing from you. Thank you, Michelle Bakerson Doctoral Student WMU 167 Interview E-mail Invitation Hi "Name of Student", My name is Michelle Bakerson and I am a student at WMU. I am conducting a study on Kalamazoo Promise Scholarship Recipients. I am a doctoral student in the Evaluation, Measurement and Research doctoral program at the College of Education. I am inviting you to learn more about this project. If after this information you are still interested and agree to participate I would like to invite you to an interview so that I can gather information about recipients of the Kalamazoo Promise who are attending or have attended WMU. You of course are free to participate or not. The choice you make will not affect your relationship with the researcher, Western Michigan University or the Kalamazoo Promise Scholarship at all. If you are interested in learning more, you will be given more information about the study and your rights before you complete the interview. Your name will never be used in reporting any information. All information will be reported aggregately only. Your name and participation will be held strictly confidential. At the end of the interview you will be given a $20 WMU book store gift card that can be used for anything at the WMU book store, such as; books, magazines, merchandise, phone cards or beverages at the WMU book store as a thank you for your time. If you have any questions about this study or would like to schedule an interview, you may contact me at (269-362-1620) (office), (269-684-5566) (home), or by e-mail at michelle.a.bakerson(q)wmich.edu. Thank you so much for your time and I look forward to hearing from you. Thank you, Michelle Bakerson Doctoral Student WMU 168 Reminder E-Mail for Interview Hello (student's name), this is Michelle Bakerson. I would like to express my appreciation for your willingness to meet with me and learn more about the project for my Kalamazoo Promise study. This message is a reminder of the meeting which will be held on (date) at (time) in (location). If you decide to participate you will receive a $20.00 WMU book store gift card as compensation for your participation. I look forward to meeting with you. If you have any questions in the meantime, please e-mail me at [email protected] or call me at 269-362-1620. Also here are the directions to get to (location)...describe directions. 169 Reminder Phone Call for Interview Hello (student's name), this is Michelle Bakerson. I would like to express my appreciation for your willingness to meet with me and learn more about the project for my Kalamazoo Promise study. This message is a reminder of the meeting which will be held on (date) at (time) in (location). Give directions for location. If you decide to participate you will receive a $20.00 WMU book store gift card as compensation for your participation. I look forward to meeting with you. If you have any questions in the meantime, please e-mail me at michelle.a.bakerson(a),wmich.edu or call me at 269-362-1620. 170 Appendix B Cognitive, Social and Institutional Factor of Retention and Corresponding Survey Items, along with Academic Data Variable Names with Measurement Type and Cognitive, Social and Institutional Factors of Retention and Corresponding Survey Items with Subscales Identified. Cognitive, Social and Institutional Factors of Retention and Corresponding Survey Items Factor Name Cognitive Items Estimate your grade point average (GPA) in high school. Estimate your grade point average (GPA), thus far at WMU. About how many hours do you spend in a typical 7-day week preparing for class (studying, reading, writing, rehearsing or other activities related to your program)? About how many hours do you spend in a typical 7-day week participating in college-sponsored activities (organizations, campus publication, student government, intercollegiate or intramural sports, etc)? How likely is it that being academically unprepared would cause you to withdraw from class or from WMU? In your experience at WMU during the current school year, about how often have you done each of the following? o Asked questions in class or contributed to class discussions o o o o o o o o o o o o o o Made a class presentation Come to class without completing readings or assignments Worked with classmates outside of class to prepare class assignments Tutored or taught other students (paid or voluntary) Participated in a community-based project as a part of a regular course Used instant messaging to work on an assignment Used e-mail to communicate with an instructor Discussed grades or assignments with an instructor Talked about career plans with an instructor or adviser Discussed ideas from your readings or classes with instructors outside of class Received prompt feedback (written or oral) from instructors on your performance Worked harder than you thought you could to meet an instructor's standards or expectations Worked with instructors on activities other than coursework Discussed ideas from your readings or classes with others 171 o o Social Institutional outside of class (students, family members, co-workers, etc.) Had serious conversations with students who differ from you in terms of their religious beliefs, political opinions, or personal values Had serious conversations with students of different race or ethnic background than your own Skipped class Included diverse perspectives (different races, religions, genders, political beliefs, etc.) in class discussions or writing assignments Put together ideas or concepts from different courses when completing assignments or during class discussions What degree(s) do you wish to pursue? What is your current major? Do you expect to enroll for an advanced degree when, or if, you complete your undergraduate degree? Briefly describe your career goals. When enrolled in high school, did you qualify for the free/reduced lunch program at your school? Where do you live during the school year? With whom do you live during the school year? What is the highest level of education obtained by your father or male guardian? What is the highest level of education obtained by your mother or female guardian? About how many hours do you spend in a typical 7-day week working for pay? About how many hours do you spend in a typical 7-day week providing care for dependents living with you (parents, children, spouse, etc.)? About how many hours do you spend in a typical 7-day week commuting to and from class? If you have a job, how does it affect your school work? How likely is it that working full-time would cause you to withdraw from class or from WMU? How likely is it that caring for dependents would cause you to withdraw from class or from WMU? How likely is it that non fitting in would cause you to withdraw from class or from WMU? How likely is it that the lace do finances would cause you to withdraw from class or from WMU? Are you a member of a social fraternity or sorority? Are you a student athlete on a team sponsored by WMU's athletics department? Are you involved in student associations or organizations? How supportive are you friends of your attending WMU? How supportive is you r immediate family of your attending WMU? What is the quality of your relationship with other students? Please rate your level of awareness about the Kalamazoo Promise. 172 What additional information would you like to have regarding the Kalamazoo Promise? How much tuition scholarship are you eligible for under the Kalamazoo Promise? How likely is it that not offering the program of study would cause you to withdraw from class or from WMU? What is the quality of relationship with instructors at WMU? What is the quality of relationship with the administrative personnel and office staff at WMU? To what extent does WMU emphasize: o Spending significant amounts of time studying? o Providing the support you need to help you succeed academically? o Encouraging contact among students from different economic, social, and racial or ethnic backgrounds? o Helping you cope with your non-academic responsibilities (work, family, etc.)? o Providing the support you need to thrive socially? o Attending campus events and activities (special speakers, cultural performances, athletic events, etc.)? o Using computers in academic work? During the current school year, about how often have you done each of the following? o Attended an art exhibit, play, dance, music, theater, or other performance o Exercised or participated in physical fitness activities o Participated in activities to enhance your spirituality (worship, meditation, prayer, etc.) o Tried to better understand someone else's views by imagining how an issue looks from his or her perspective o Learned something that changed the way you understand an issue or concept Overall, how would you evaluate the quality of academic advising you have received at WMU? How would you evaluate your entire educational experience at WMU? If you could start over again, would you still attend WMU? Would you recommend WMU to a friend or family member? 173 Academic Data Variable Names with Measurement Type Variable Name StudentID FirstPromiseSemester FTIACCohort DualEnrolled HSGPA HSGradDate HSName ACTComposite Gender Race TransHrsPass Probation MostRecentWMUGPA MostRecentGPASemester RemedialMath RemedialReading RemedialWriting ParentsAGI Athelete Dorm APCredit FYE GPA200630SumII GPA200640Fall GPA200710Spring GPA200720SumI GPA200730SumII GPA200740Fall GPA200810Spring GPA200820SumI GPA200830SumII GPA200840Fall GPA200910Spring Group DateResponded GroupByResponse GroupByLateResponse Measurement Nominal Nominal Nominal Nominal Scale Nominal Nominal Scale Nominal Nominal Scale Nominal Scale Nominal Nominal Nominal Nominal Scale Nominal Nominal Nominal Nominal Scale Scale Scale Scale Scale Scale Scale Scale Scale Scale Scale Nominal Nominal Nominal Nominal 174 Cognitive, Social and Institutional Factors of Retention and Corresponding Survey Items with Subscales Identified. Factor Name Cognitive Items CI Estimate your grade point average (GPA) in high school. C2 Estimate your grade point average (GPA), thus far at WMU. C3 About how many hours do you spend in a typical 7-day week preparing for class (studying, reading, writing, rehearsing or other activities related to your program)? C4 About how many hours do you spend in a typical 7-day week participating in college-sponsored activities (organizations, campus publication, student government, intercollegiate or intramural sports, etc)? C5 How likely is it that being academically unprepared would cause you to withdraw from class or from WMU? In your experience at WMU during the current school year, about how often have you done each of the following? Cognitive Engagement Subscale o C6 Asked questions in class or contributed to class discussions o C7 Made a class presentation o C8 Come to class without completing readings or assignments o C9 Worked with classmates outside of class to prepare class assignments o CIO Tutored or taught other students (paid or voluntary) o CI 1 Participated in a community-based project as a part of a regular course o CI2 Used instant messaging to work on an assignment o CI3 Used e-mail to communicate with an instructor o C14 Discussed grades or assignments with an instructor o CI 5 Talked about career plans with an instructor or adviser o C16 Discussed ideas from your readings or classes with instructors outside of class o CI7 Received prompt feedback (written or oral) from instructors on your performance o CI8 Worked harder than you thought you could to meet an instructor's standards or expectations o CI9 Worked with instructors on activities other than coursework o C20 Discussed ideas from your readings or classes with others outside of class (students, family members, co-workers, etc.) o C21 Had serious conversations with students who differ from you in terms of their religious beliefs, political opinions, or personal values o C22 Had serious conversations with students of different race or ethnic background than your own o C23 Skipped class o C24 Included diverse perspectives (different races, religions, genders, political beliefs, etc.) in class discussions or writing assignments o C25 Put together ideas or concepts from different courses when completing assignments or during class discussions 175 Social • • • • • • • • • • • • • SI What degree(s) do you wish to pursue? S2 What is your current major? S3 Do you expect to enroll for an advanced degree when, or if, you complete your undergraduate degree? S4 Briefly describe your career goals. S5 When enrolled in high school, did you qualify for the free/reduced lunch program at your school? S6 Where do you live during the school year? S7 With whom do you live during the school year? S8 What is the highest level of education obtained by your father or male guardian? S9 What is the highest level of education obtained by your mother or female guardian? S10 About how many hours do you spend in a typical 7-day week working for pay? SI 1 About how many hours do you spend in a typical 7-day week providing care for dependents living with you (parents, children, spouse, etc.)? S12 About how many hours do you spend in a typical 7-day week commuting to and from class? S13 If you have a job, how does it affect your school work? Social Demands Subscale • • Institutional S14 How likely is it that working full-time would cause you to withdraw from class or from WMU? SI5 How likely is it that caring for dependents would cause you to withdraw from class or from WMU? 516 How likely is it that non fitting in would cause you to withdraw from class or from WMU? 517 How likely is it that the lace do finances would cause you to withdraw from class or from WMU? 518 Are you a member of a social fraternity or sorority? 519 Are you a student athlete on a team sponsored by WMU's athletics department? 520 Are you involved in student associations or organizations? 521 How supportive are you friends of your attending WMU? 522 How supportive is you r immediate family of your attending WMU? S23What is the quality of your relationship with other students? 11 Please rate your level of awareness about the Kalamazoo Promise. 12 What additional information would you like to have regarding the Kalamazoo Promise? I3How much tuition scholarship are you eligible for under the Kalamazoo Promise? 14 How likely is it that not offering the program of study would cause you to withdraw from class or from WMU? 15 What is the quality of relationship with instructors at WMU? 16 What is the quality of relationship with the administrative personnel and office staff at WMU? 176 • To what extent does WMU emphasize: (Institutional Support Subscale) o 117 Spending significant amounts of time studying? o 118 Providing the support you need to help you succeed academically? o 119 Encouraging contact among students from different economic, social, and racial or ethnic backgrounds? o 120 Helping you cope with your non-academic responsibilities (work, family, etc.)? o 121 Providing the support you need to thrive socially? o 122 Attending campus events and activities (special speakers, cultural performances, athletic events, etc.)? o I 23Using computers in academic work? • • • • • During the current school year, about how often have you done each of the following? (Social Engagement Subscale) o 124 Attended an art exhibit, play, dance, music, theater, or other performance o 125 Exercised or participated in physical fitness activities o 126 Participated in activities to enhance your spirituality (worship, meditation, prayer, etc.) o 127 Tried to better understand someone else's views by imagining how an issue looks from his or her perspective o 128 Learned something that changed the way you understand an issue or concept 129 Overall, how would you evaluate the quality of academic advising you have received at WMU? 130 How would you evaluate your entire educational experience at WMU? 131 If you could start over again, would you still attend WMU? I32Would you recommend WMU to a friend or family member? 177 Appendix C Summary Demographic Data on Interval Level Data Across Descriptive Statistics Std. Academic Academic persistence Race Gender Mean persistence 1 Male 3.5815 .59654 71 Female 3.8051 .36166 59 Total 3.6830 .51419 130 Male 2.6940 1.16989 15 Female 3.1065 .87016 20 Total 2.9297 1.01469 35 Male 3.5506 .48484 16 Female 3.2147 1.19125 19 Total 3.3683 .94011 35 Male 3.4462 .75492 102 Female 3.5481 .76941 98 Total 3.4961 .76185 200 Male 3.3411 .42700 18 Female 3.3127 .42408 11 Total 3.3303 .41848 29 Male 2.7275 .27035 4 Female 2.9178 .38986 9 Total 2.8592 .35771 13 Male 2.9143 .34708 7 Female 3.2450 .19092 2 Total 2.9878 .34084 9 Male 3.1534 .45519 29 Female 3.1450 .42789 22 Cognitive Deviation N High School 2 GPA 3 - Total On 1 Probation 2 3 Total 178 Non- 1 persister 2 3 Total Total 1 2 3 Total Academic persistence 1 Cognitive Most Recent Western Michigan 2 Total 3.1498 .43926 51 Male 3.2231 .28922 16 Female 3.2500 .41689 8 Total 3.2321 .32805 24 Male 2.8964 .29354 11 Female 2.5650 1.29157 6 Total 2.7794 .77575 17 Male 2.8560 .40321 5 Female 3.3467 .62083 3 Total 3.0400 .51722 8 Male 3.0534 .34518 32 Female 3.0253 .87726 17 Total 3.0437 .57763 49 Male 3.4857 .54944 105 Female 3.6787 .43376 78 Total 3.5680 .51114 183 Male 2.7727 .84100 30 Female 2.9651 .86303 35 Total 2.8763 .85178 65 Male 3.2675 .54098 28 Female 3.2337 1.07128 24 Total 3.2519 .82024 52 Male 3.3170 .66554 163 Female 3.4185 .76457 137 Total 3.3633 .71305 300 Male 3.1145 .49243 71 Female 3.1578 .52141 59 Total 3.1342 .50427 130 Male 2.6513 .40438 15 Female 3.0005 .38093 20 179 University GPA 3 Total On 1 Probation 2.8509 .42327 35 Male 3.0463 .43124 16 Female 2.9237 .56227 19 Total 2.9797 .50325 35 Male 3.0357 .49464 102 Female 3.0803 .50934 98 Total 3.0576 .50113 200 Male 1.7906 .58799 18 .9482 .85781 11 Total 1.4710 .80358 29 Male 1.9700 .16207 4 Female 1.9233 .58402 9 Total 1.9377 .48420 13 Male 1.6329 .87036 7 Female 1.9450 .12021 2 Total 1.7022 .76739 9 Male 1.7772 .62104 29 Female 1.4377 .85564 22 Total 1.6308 .74318 51 Male .9613 .71620 16 1.0038 .58977 8 Total .9754 .66393 24 Male 1.2745 .41972 11 Female 1.3983 .30407 6 Total 1.3182 .37778 17 Male 1.0020 .31027 5 Female 1.6300 .45398 3 Total 1.2375 .46855 8 Male 1.0753 .58234 32 Female 1.2535 .52198 17 Total 1.1371 .56319 49 Femg|e 2 3 Total Non- Total 1 persister Fema|e 2 3 Total 180 Total 1 2 3 Total Academic persistence 1 Cognitive ACT Composite 2 Score 3 Total On 1 Probation 2 3 Male 2.5594 .99966 105 Female 2.6253 1.10646 78 Total 2.5875 1.04406 183 Male 2.0557 .74725 30 Female 2.4489 .78855 35 Total 2.2674 .78891 65 Male 2.3279 1.02104 28 Female 2.6804 .71120 24 Total 2.4906 .90085 52 Male 2.4269 .97665 163 Female 2.5899 .97026 137 Total 2.5013 .97551 300 Male 23.04 4.695 71 Female 21.46 5.361 59 Total 22.32 5.050 130 Male 17.00 5.529 15 Female 17.00 4.611 20 Total 17.00 4.947 35 Male 21.38 7.228 16 Female 15.95 7.884 19 Total 18.43 7.968 35 Male 21.89 5.639 102 Female 19.48 6.243 98 Total 20.71 6.050 200 Male 20.78 4.152 18 Female 19.91 2.879 11 Total 20.45 3.690 29 Male 20.00 4.082 4 Female 16.56 1.424 9 Total 17.62 2.873 13 Male 19.29 3.684 7 181 Total Non- 1 persister 2 3 Total Total 1 2 3 Total Female 18.50 2.121 2 Total 19.11 3.296 9 Male 20.31 3.947 29 Female 18.41 2.754 22 Total 19.49 3.580 51 Male 21.50 2.757 16 Femg|e 19.25 2.121 8 Total 20.75 2.739 24 Male 16.82 1.662 11 Female 16.83 2.483 6 Total 16.82 1.912 17 Male 19.80 2.588 5 Female 18.33 3.215 3 Total 19.25 2.712 8 Male 19.63 3.170 32 Female 18.24 2.538 17 Total 19.14 3.014 49 Male 22.42 4.428 105 Female 21.01 4.876 78 Total 21.82 4.664 183 Male 17.33 4.310 30 Female 16.86 3.647 35 Total 17.08 3.942 65 Male 20.57 5.827 28 Female 16.46 7.126 24 Total 18.67 6.721 52 Male 21.17 5.037 163 Female 19.15 5.477 137 Total 20.25 5.329 300 182 Appendix D Tests of Normality and Persistence, those on Probation and Non-persistence, Subscale Scores of the Survey of Promise Recipients by Response Category Persistence, those on probation and non-persistence Tests of Normality Kolmogorov-Smirnov3 persistence Statistic Academic Cognitive High persistence df Shapiro-Wilk Sig. Statistic df Sig. .165 200 .000 .708 200 .000 „ „ , On Probation .084 51 .200* .968 51 .186 Non-persister .137 49 .022 .747 49 .000 Academic Cognitive ACT persistence .147 200 .000 .862 200 .000 Composite Score Probation .188 51 .000 .886 51 .000 Non-persister .138 49 .021 .961 49 .107 .060 200 .075 .975 200 .001 .225 51 .000 .833 51 .000 .111 49 .182 .936 49 .010 .210 200 .000 .644 200 .000 .137 51 .018 .904 51 .001 .146 49 .010 .932 49 .007 School GPA 0n Academic Cognitive Most persistence Recent Western Michigan _ . - , . . . a On Probation University GPA Non-persister Academic Social Parents persistence Aggregate Income On Probation Non-persister a. Lilliefors Significance Correction *. This is a lower bound of the true significance. 183 Subscale scores of the Survey of Promise Recipients by Response Category Kolmogorov-Smirnova Survey Group by Response Social Demands Statistic On Time Respondent Late Respondent(so nonresponse) Cognitive Engagement On Time Respondent Late Respondent(so nonresponse) Social Engagement On Time Respondent Late Respondent(so nonresponse) Institutional Support On Time Respondent Late Respondent(so nonresponse) 184 df Sig. .147 37 .042 .153 41 .017 .106 37 .200* .145 41 .029 .138 37 .071 .164 41 .007 .144 37 .052 .105 41 .200* Appendix E Tests of Homogeneity of Variance and Subscales of the Subscale Scores of the Survey of Promise Recipients by Response Category Levene's Test of Equality of Error Variances3 F Academic Cognitive High df1 df2 Sig. 2.608 17 282 .001 3.080 17 282 .000 1.442 17 282 .116 School GPA Academic Cognitive Most Recent Western Michigan University GPA Academic Cognitive ACT Composite Score Tests the null hypothesis that the error variance of the dependent variable is equal across groups. a. Design: Intercept + persistence + ethnicity + aGender + persistence * ethnicity + persistence * aGender + ethnicity * aGender + persistence * ethnicity * aGender Test of Homogeneity of Variances Academic Social Parents Aggregate Income Levene Statistic 3.473 df2 df1 2 Sig. 297 .032 Subscales of the Subscale scores of the Survey of Promise Recipients by Response Category 185 F dfl df2 Sig. 2.267 1 76 .136 Cognitive Engagement .625 1 76 .432 Social Engagement .915 1 76 .342 Institutional Support .280 1 76 .598 Social Demands 186 Appendix F Summary Results of Item Analysis, Institutional Support, Social Engagement, Social Demands and Cognitive Engagement Institutional Support Item Statistics Mean siWMUEmphasizeTimeStudyin Std. Deviation N 3.5747 .99571 87 3.5172 .98668 87 2.9310 1.17921 87 2.4828 1.06599 87 2.9080 1.10635 87 3.5632 .99652 87 4.2299 .75792 87 g siWMUEmphasizeSupportAcad emically siWMUEmphasizeContactStud entsDifferentEconSocRace siWMUEmphasizeHelpCopeNo nacademics siWMUEmphasizeProvidingSu pportThriveSocially siWMUEmphasizeAttendingCa mpusEventsActivities siWMUEmphasizeUsingComp utersAcademicWork Summary Item Statistics Maximum / Mean Item Means Item Variances Minimum Maximum Range Minimum Variance N of Items 3.315 2.483 4.230 1.747 1.704 .335 7 1.040 .574 1.391 .816 2.421 .065 7 187 Item-Total Statistics siWMUEmphasizeTimeSt udying Corrected Item- Squared Cronbach's Scale Mean if Scale Variance Total Multiple Alpha if Item Item Deleted if Item Deleted Correlation Correlation Deleted 19.6322 16.979 .500 .317 .755 19.6897 16.217 .614 .423 .733 20.2759 14.574 .678 .528 .715 20.7241 15.993 .580 .446 .739 20.2989 15.445 .622 .462 .729 19.6437 18.813 .262 .164 .799 18.9770 19.627 .279 .128 .790 siWMUEmphasizeSuppor tAcademically siWMUEmphasizeContac tStudentsDifferentEconSo cRace siWMUEmphasizeHelpC opeNonacademics siWMUEmphasizeProvidi ngSupportThriveSocially siWMUEraphasizeAttendi ngCampusEventsActivitie s siWMUEmphasizeUsing Computers AcademicWor k Scale Statistics Mean 23.2069 Variance 22.073 Std. Deviation N of Items 7 4.69819 188 Social Engagement Case Processing Summary % N Cases Valid 88 28.7 Excludeda 219 71.3 Total 307 100.0 a. Listwise deletion based on all variables in the procedure. Item Statistics Mean Std. Deviation N ssOftenAttendedArtExhibitPlayDanceMusic 2.6932 1.39257 88 ssOftenExercisedParticipatedPhysEd 3.4545 1.30348 88 ssOftenParticpatedSpiritualActivities 2.0568 1.35916 88 ssOftenTriedUnderstandSomeoneElsesView 3.2955 1.27900 88 3.5455 1.03845 88 ssOftenLearnedSomethingChangedWayUnd erstandlssue Summary Item Statistics Maximum / Mean Item Means 3.009 Minimum 2.057 Maximum Range 3.545 189 1.489 Minimum 1.724 Variance .394 N of Items 5 Item-Total Statistics Corrected Item- Squared Cronbach's Scale Mean if Scale Variance Total Multiple Alpha if Item Item Deleted if Item Deleted Correlation Correlation Deleted ssOftenAttendedArtExhib itPlayDanceMusic 12.3523 12.920 .132 .021 .687 11.5909 11.693 .313 .130 .592 12.9886 10.287 .464 .237 .510 11.7500 10.351 .510 .460 .488 11.5000 11.494 .513 .434 .507 ssOftenExercisedParticipa tedPhysEd ssOftenParticpatedSpiritu alAotivities ssOftenTriedUnderstandS omeoneElsesView ssOftenLearnedSomething ChangedWayUnderstandl ssue Scale Statistics Mean Variance Std. Deviation 16.182 15.0455 N of Items 4.02266 5 Cronbach's Alpha Based on Cronbach's Alpha Standardized Items .687 .697 N of Items 4 190 Item Statistics Mean ssOftenExercisedParticipatedPhy sEd ssOftenPartiopatedSpiritualActiv Std. Deviation N 3.4545 1.30348 88 2.0568 1.35916 88 3.2955 1.27900 88 3.5455 1.03845 88 ities ssOftenTriedUnderstandSomeon eElsesView ssOftenLearnedSomethingChang edWayUnderstandlssue Summary Item Statistics Maximum / Mean Item Means 3.088 Minimum Maximum 2.057 Minimum Range 3.545 1.489 Variance 1.724 N of Items 4 .483 Item-Total Statistics Corrected Item- Squared Cronbach's Scale Mean if Scale Variance if Total Multiple Alpha if Item Item Deleted Item Deleted Correlation Correlation Deleted ssOftenExercisedParticipa tedPhysEd 8.8977 8.645 .336 .130 .708 10.2955 7.544 .473 .229 .622 9.0568 7.411 .556 .460 .564 ssOftenParticpatedSpiritu alActivities s sOftenTriedUnderstandS omeoneElsesView 191 Item-Total Statistics Corrected Item- Squared Cronbach's Scale Mean if Scale Variance if Total Multiple Alpha if Item Item Deleted Item Deleted Correlation Correlation Deleted ssOftenExercisedParticipa tedPhysEd 8.8977 8.645 .336 .130 .708 10.2955 7.544 .473 .229 .622 9.0568 7.411 .556 .460 .564 8.8068 8.502 .551 .433 .586 ssOftenParticpatedSpiritu alActivities ssOftenTriedUnderstandS omeoneElsesView ssOftenLearnedSomething ChangedWayUnderstandl ssue Scale Statistics Mean Variance 12.3523 Std. Deviation N of Items 3.59450 12.920 4 Intraclass Correlation Coefficient F Test with True Value .7 95% Confidence Interval Intraclass Correlationa Single Measures Average Measures Lower Bound Upper Bound Value dfl df2 Sig .355b .245 .473 .309 87 261 1.000 .687c .565 .782 .959 87 261 .582 192 Social Demands Case Processing Summary % N Cases Valid 94 30.6 Excluded3 213 69.4 Total 307 100.0 a. Listwise deletion based on all variables in the procedure. Reliability Statistics Cronbach's Alpha Based on N of Items Cronbach's Alpha Standardized Items .748 .752 6 Item Statistics Mean Std. Deviation N 2.2660 1.37704 94 1.9362 1.21645 94 2.1809 1.27813 94 ssWithdrawLackFinances 2.4681 1.51482 94 ssWithdrawDontFitln 1.3404 .74130 94 2.2553 1.48060 94 ssWithdrawWorkingFullTime ssWithdrawCaringForDependent s ssWithdrawAcademicallyUnprep ared siWithdrawDontOfferProgramW anted 193 Summary Item Statistics Maximum / Mean Item Means Minimum 2.074 Maximum 2.468 1.340 Minimum Range 1.128 Variance 1.841 N of Items .159 6 Item-Total Statistics Corrected Item- ssWithdrawWorkingFullT ime ssWithdrawCaringForDep endents Squared Cronbach's Scale Mean if Scale Variance if Total Multiple Alpha if Item Item Deleted Item Deleted Correlation Correlation Deleted 10.1809 18.085 .572 .340 .687 10.5106 19.392 .542 .330 .698 10.2660 18.713 .573 .356 .688 9.9787 17.333 .559 .340 .691 11.1064 23.601 .351 .138 .747 10.1915 19.640 .369 .147 .750 ssWithdrawAcademically Unprepared ssWithdrawLackFinances ssWithdrawDontFitln siWithdrawDontOfferPro gramWanted Scale Statistics Mean 12.4468 Variance 26.680 Std. Deviation N of Items 5,16526 6 194 Cognitive Engagement % N Cases 90 29.3 Excluded2 217 70.7 Total 307 100.0 Valid a. Listwise deletion based on all variables in the procedure. Reliability Statistics Cronbach's Alpha Based on Cronbach's Alpha Standardized Items .830 N of Items 18 .831 Item Statistics Mean Std. Deviation N scl 3.8000 .92651 90 sc2 3.2778 1.14193 90 sc4 3.2556 1.25922 90 sc5 2.0556 1.19325 90 sc6 1.7667 1.02825 90 sc7 2.0667 1.27904 90 sc8 4.0667 .93376 90 sc9 3.6222 1.05551 90 sclO 2.7444 1.25027 90 sell 2.3111 1.25102 90 scl2 3.3222 1.00368 90 sc!3 3.2667 1.09954 90 195 scl4 1.8556 1.11739 90 scl5 3.5000 1.15389 90 scl6 3.1333 1.27376 90 ssl7 3.4889 1.17315 90 scl9 3.1667 1.09391 90 sc20 3.4111 1.01555 90 Summary Item Statistics Maximum / Mean Item Means 3.006 Minimum Maximum 4.067 1.767 Minimum Range 2.300 Variance 2.302 N of Items .489 18 Item-Total Statistics scl sc2 sc4 sc5 sc6 sc7 sc8 sc9 sclO sell Scale Mean if Item Scale Variance if Corrected Item- Deleted Item Deleted Total Correlation Squared Multiple Cronbach's Alpha if Correlation Item Deleted 50.3111 97.183 .450 .327 .820 50.8333 95.287 .434 .351 .821 50.8556 93.833 .444 .481 .820 52.0556 96.120 .372 .392 .824 52.3444 94.610 .529 .395 .816 52.0444 97.256 .292 .324 .829 50.0444 99.728 .303 .507 .827 50.4889 97.758 .354 .594 .825 51.3667 95.785 .364 .496 .825 51.8000 89.061 .662 .632 .807 196 scl2 scl3 scl4 scl5 scl6 ssl7 scl9 sc20 50.7889 98.483 .340 .390 .825 50.8444 96.470 .397 .368 .822 52.2556 93.855 .515 .425 .816 50.6111 95.656 .411 .439 .822 50.9778 95.325 .374 .508 .824 50.6222 94.800 .441 .515 .820 50.9444 94.323 .505 .601 .817 50.7000 97.583 .381 .423 .823 Scale Statistics Mean 54.1111 Variance 106.257 Std. Deviation N of Items 18 10.30811 197 Appendix G Survey Summary Tables Table Gl Which High School Did You Attend? Loy Norrix Kalamazoo Central Phoenix Response Percent 40.6% 59.4% 0.0% Response Count 41 60 0 Response Percent 39.6% 33.7% 26.7% Response Count 40 34 27 Response Percent 15.8% 84.2% Response Count 16 85 16 Table G2 In which Year Did You Graduate High School? 2006 2007 2008 Table G3 Did You Begin College at WMU or Elsewhere? Started elsewhere Started at WMU If elsewhere, please specify where: 198 Table G4 What is your classification in college? Freshman/first-year Sophomore Junior Senior Unclassified Response Percent Response Count 25.0% 31.0% 38.0% 6.0% 0.0% 25 31 38 6 0 Response Percent 96.0% 4.0% 0.0% Response Count 96 4 0 Note. One person did not respond. Table G5 When do you most frequently take classes? Day classes (morning or afternoon classes) Evening classes Weekend classes Table G6 Do you expect to enroll for an advanced degree when, or if you complete your undergraduate degree? Response Percent 34.7% 65.3% No Yes Note. Three did not respond. 199 Response Count 34 64 Table G7 When enrolled in high school, did you qualify for the free/reduced lunch program at your school? No Yes Response Percent 70.1% 29.9% Response Count 68 29 Response Percent 47.9% 52.1% Response Count 46 50 Note. N=91, 4 did not answer. Table G8 What is your gender? Female Male Note. N=96, 5 did not answer. Table G9 What is your race/ethnicity? (Mark the one ethnic group with which you most identify American Indian or other Native American Asian, Asian American, or Pacific Islander Black or African American White (non-Hispanic) Mexican or Mexican American Puerto Rican Other Hispanic or Latino Multiracial I prefer not to respond Other (please specify) Note. N=93, 8 did not answer. 200 Response Percent 0.0% 6.5% 18.3% 61.3% 3.2% 0.0% 1.1% 5.4% 4.3% Response Count 0 6 17 57 3 0 1 5 4 3 Table G10 Is English your native (first) language? No Yes Response Percent 9.4% 90.6% Response Count 9 87 Response Percent 35.1% Response Count 34 19.6% 19 45.4% 44 0.0% 0 Note. N=96, 5 did not answer. Table Gl 1 Where to you live during the school year? Dormitory or other campus housing Residence (house, apartment, etc.) within walking distance of Western Residence (house, apartment, etc.) within driving distance Fraternity or sorority house Note. N=91, 4 did not answer. Table G12 With whom do you live during the school year? (Fill in all that apply Response Percent 6.2% 54.6% 5.2% 1.0% 34.0% 3.1% 5.2% No one, I live alone One or more other students My spouse or partner My child or children My parents Other relatives Friends who are not students at WMU Other people, who Note. N=97, 4 did not answer. 201 Response Count 6 53 5 1 33 3 5 2 Table G12 What is the highest level of education obtained by your Father or Mother Not a high school graduate High school diploma or GED Some college, did not complete degree Associate degree Bachelor's degree Master's degree Doctorate degree and/or Professional degree Unknown Father or Male guardian Mother or Female guardian Response Count 3 18 17 8 20 9 3 13 13 15 24 14 6 31 30 23 44 23 4 2 6 2 3 5 Note. N=96, 5 did not answer. Tabled 3 Please rate your level of awareness about the Kalamazoo Promise. Response Percent 1.0% 0.0% 10.4% 41.7% 46.9% Not at all familiar Not familiar Neutral Familiar Very familiar Note. N=96, 5 did not answer. 202 Response Count 1 0 10 40 45 Table G 1 4 About how many hours do you spend in a typical 7-day week doing each of the Preparing for class (studying, reading, writing, rehearsing or other activities related to your program Working for pay Participating in college-sponsored activities organizations, campus publications, student government, intercollegiate or intramural sports, etc.) Providing care for dependents living with your (parents, children, spouse, etc.) Commuting to and from classes XT None 1c 1-5 c m 6-10 11- 1 27 31 30 21 10 14 43 38 67 17 31+ Response ^ 6 2 97 24 21 6 96 10 3 1 0 95 23 2 2 0 2 96 73 5 0 0 0 95 2Q 213Q Note. N=97, 4 did not answer Tabled 5 If you have a job, how does it affect your school work? I don't have a job My job does not interfere with my school work My job takes some time from my school work My job takes a lot of time from my school work Note. N=97, 4 did not answer 203 following? Response Percent 25.8% 26.8% 39.2% 8.2% Response Count 25 26 38 8 Tabled 6 How likely is it that the following issues would cause you to withdraw from class or from WMU? Not Likely 1 Working full-time Caring for dependents Academically unprepared Lack of finances Don't fit in Don't offer program of study that I want Very Likely 5 Response Count 42 16 16 16 7 97 51 15 18 6 6 96 42 17 20 11 7 97 39 75 16 15 13 3 12 4 16 0 96 97 48 11 13 13 12 97 Note. N=97, 4 did not answer TableGl 7 Are you a member of a social fraternity or sorority? Response Percent 96.8% 3.2% No Yes If yes, which one? Response Count 91 3 4 Note. 7V=94, 7 did not answer Tabled 8 Are you a student athlete on a team sponsored by WMU's athletics department? Response Percent 100.0% 0.0% No Yes Note. N=91, 4 did not answer 204 Response Count 97 0 Tabled 9 How supportive are your friends of your attending WMU? Response Percent 2.1% 15.5% 36.1% 46.4% Not Very Somewhat Quite a bit Extremely Response Count 2 15 35 45 Note. N=94, 7 did not answer Table G20 How supportive is your immediate family of your attending WMU? Response Percent 0.0% 6.2% 23.7% 70.1% Not Very Somewhat Quite a bit Extremely Response Count 0 6 23 68 Note. N=94, 7 did not answer Table G21 Which best represents the quality of your relationship with students at WMU? 1 Unfriendly, unsupportive, sense of alienation 2 3 4 5 Friendly, supportive, sense of belonging Note. 7V=94, 7 did not answer 205 Response Percent 0.0% 6.2% 19.6% 36.1% 38.1% Response Count 0 6 19 35 37 Table G22 Which best represents the quality of your relationships with instructors at WMU? Response Percent Response Count 0.0% 4.1% 33.0% 44.3% 18.6% 0 4 32 43 18 1 Unavailable, unhelpful, unsympathetic 2 3 4 5 Available, helpful, sympathetic Note. 7V=94, 7 did not answer Table G23 Which best represents the quality of your relationship with administrative personnel & office staff at WMU? Response Percent Response Count 0.0% 14.4% 33.0% 33.0% 19.6% 0 14 32 32 19 1 Unhelpful, Inconsiderate, rigid 2 3 4 5 Helpful, considerate, flexible Note. 7V=94, 7 did not answer Table G24 In your experience at WMU during the current school year, about how often have you done each of the following? Asked questions in class or contributed to class discussions Made a class presentation Come to class without completing Very Often 4 5 Never 1 2 0 6 32 28 30 5 21 27 26 17. 7 39 26 20 4 206 3 readings or assignments Worked with classmates outside of class to prepare class assignments 10 17 26 25 18 Tutored or taught other students (paid or voluntary) 44 21 15 12 3 Participated in a community-based project as a part of a regular course 54 21 14 5 2 Used instant messaging to work on an assignment 49 18 12 11 6 Used e-mail to communicate with an instructor Discussed grades or assignments with an instructor 0 5 20 29 42 1 15 24 30 26 Talked about career plans with an instructor or adviser 18 23 21 24 10 Discussed ideas from your readings or classes with instructors outside of class 30 26 20 10 9 Received prompt feedback (written or oral) from instructors on your performance 5 11 39 25 15 Worked harder than you thought you could to meet an instructor's standards or expectations 5 16 33 25 16 49 21 17 3 5 7 11 27 30 21 13 14 31 21 17 Worked with instructors on activities other than coursework Discussed ideas from your readings or classes with others outside of class (students, family members, coworkers, etc.) Had serious conversations with students who differ from you in terms of their religious beliefs, 207 political opinions, or personal values Had serious conversations with students of different race or ethnic background than your own Skipped class Included diverse perspectives (different races, religions, genders, political beliefs, etc.) in class discussions or writing assignments Put together ideas or concepts from different courses when completing assignments or during class discussions Note. N=96, 5 did not answer 6 12 32 22 23 15 45 22 12 2 8 15 35 25 13 2 15 31 31 17 Table G25 To what extent does WMU emphasize each of the following? Spending significant amounts of time studying Providing the support you need to help you succeed academically Encouraging contact among students from different economic, social, and racial or ethnic backgrounds Helping you cope with your non-academic responsibilities (work, family, etc.) Very Little 1 2 2 3 4 8 33 31 Very Much 5 18 3 8 32 33 14 14 18 28 24 8 17 31 29 11 4 10 20 34 19 7 2 11 33 26 19 0 2 16 37 37 Providing the support you need to thrive socially Attending campus events and activities (special speakers, cultural performances, athletic events, etc.) Using computers in academic work Note. N=92, 9 did not answer 209 Table G26 During the current school year, about how often have you done each of the following? Very Little 1 2 3 4 Very Much 5 25 19 20 15 12 Exercised or participated in physical fitness activities 13 3 31 19 25 Participated in activities to enhance your spirituality (worship, meditation, prayer, etc.) 47 13 18 2 11 23 21 20 21 14 28 27 19 Attended an art exhibit, play, dance, music, theater, or other performance Tried to better understand someone else's views by imagining how an issue looks from his or her perspective Learned something that changed the way you understand an issue or concept 1 Note. N=9l, 10 did not answer Table G27 Overall, how would you evaluate the quality of academic advising you have received at WMU? Response Percent 8.6% 23.7% 46.2% 21.5% Poor Fair Good Excellent Note. N=93, 8 did not answer 210 Response Count 8 22 43 20 Table G28 How would you evaluate your entire educational experience at WMU? Response Percent 1.1% 20.7% 57.6% 20.7% Poor Fair Good Excellent Response Count 1 19 53 19 Note. N-92, 9 did not answer Table G29 If you could start over again, would you still attend WMU? Response Percent 3.3% 12.0% 53.3% 31.5% Definitely no Probably no Probably yes Definitely yes Response Count 3 11 49 29 Note. N=92, 9 did not answer Table G30 Would you recommend WMU to a friend or family member? Response Percent 4.3% 95.7% No Yes Note. N=92, 9 did not answer 211 Response Count 4 88 Table G31 To what extent do you agree or disagree with the following statements about your high school? Strongly disagree ° 2 4 4 29 My high school teachers made extra efforts to help students 2 7 27 Teachers at my high school understood and met the needs of each student . -- . Teachers at my high school were fair to students 4 4 38 27 13 31 24 At my high school, students were interested in learning new things 8 28 Students at my high school had fun but also worked hard on their studies _ My high school teachers were patient when a student had trouble learning Students at my high school understood why they were in school Students at my high school worked hard to complete their school assignments 1Q agree 5 Know Rating . Average 31 15 5 3.59 32 16 4 3.63 , - _„ 12 3 3.46 11 4 5 2.54 34 13 1 4 2.65 oo 34 14 1 3 268 ^ 33 10 2 3 Strongly 4 c . Don't T, 4 2 .55 Getting good grades in high school was important to me 0 1 10 14 59 4 4.56 I pushed myself in high school to do better academically 1 3 15 24 41 4 4.20 212 In high school, I believed that I could be successful 0 4 4 18 When I was in high school I believed that going to college was important to my future My high school teachers believed that I would graduate from high school 57 5 4.54 72 5 4.83 69 9 4.84 My high school teachers believed that I would succeed in college 1 2 2 9 64 10 4.71 My high school teachers had high expectations of me in class 2 4 1 15 59 7 4.54 I had a high school teacher who was a positive role model for me 1 5 8 15 54 5 4.40 4 2 10 34 32 6 4 07 3 5 2 3 32 19 6 3.72 15 24 14 9 3.37 At my high school teachers or counselors encouraged students to think about their future At my high school teachers or counselors helped students plan for future classes and for future jobs At my high school teachers or counselors helped students with personal problems Students at my high school could _.! , j _, . .cx i get help and advice from teachers or counselors When I was in high school I . ,j • T J i received the assistance I needed to go to college My high school prepared me well for my future 22 _ „ „ - „ 3 27 28 1r, 5 19 , 6 - ,. 3.65 , 6 . 4 ,, 16 „, 26 _-, 31 , 5 _ „„ 3.87 7 14 18 30 15 4 3.38 c Note. iV=88, 13 did not answer 213 Table G32 To what extent do you agree or disagree with the following statements regarding the Kalamazoo Promise? Strongly Disagree 1 2 3 4 Strongly Agree 5 Teachers and/or school staff at my high school spoke to me about the Kalamazoo Promise 4 6 17 28 32 My parents/guardians have spoken with me about the Kalamazoo Promise 2 9 17 15 44 21 22 30 15 28 My parents/guardians encouraged me to work harder in school because of the Promise The Kalamazoo Promise gave me more flexibility about which college or university I may choose to attend 12 12 20 The Promise hasn't really made a difference to my educational goals or plans 29 17 15 I changed my career goals because of the Kalamazoo Promise 43 15 15 I worked harder in high school because I knew that the Promise would pay for college I was confident before the Promise that I could afford to go to college, using financial aid, scholarships, and/or my family's resources 21 12 24 14 16 11 13 20 18 24 I wasn't sure that I could afford college before the Promise. I didn't know if I would be able to get the scholarships, financial aid, or loans that I would need 26 21 15 13 12 I still am not sure if I can afford college, because I am not eligible for 100% of tuition from the Promise 65 I wanted to go to college even before the announcement of the Kalamazoo Promise in November 2005 Note. N=&7, 14 did not answer 214 18 69 Table G33 To what extent do you agree or disagree with the following statements regarding changes in your high school after the announcement of the Kalamazoo Promise? Str0ngly Disagree & 2 3 4 Stranelv 7°nglJ Agree 5 23 14 28 15 6 16 12 33 19 6 16 22 27 14 5 27 24 22 11 2 5 3 18 26 33 23 26 32 4 1 Students became better behaved and were getting into less trouble 29 24 23 7 3 More information was provided about higher education opportunities 6 7 25 33 13 My peers were more motivated to succeed in school 9 15 30 27 5 I talked about college more often with peers 8 8 24 33 12 The quality of student academic performance improved 10 1937 16 4 „ „ , _ . . 7 25 31 16 My attendance in high school improved school My school started offering more college prep courses I enrolled in more college prep courses Teachers expected that more students would go to college The amount of homework increased More support from community • *• -A A * *, A 4. orgamzations was provided to students and families Note. N=86, 15 did not answer , 6 215 1/r
© Copyright 2026 Paperzz