The Effect of Class Size on Student Performance and Retention at Binghamton University Jack Keil Peter J. Partell Motivation for Study Preliminary Audit by NYS Comptroller’s Office recommended SUNY Campuses increase class sizes (“faculty productivity”) Majority of research on subject conducted at elementary & secondary levels Gauge effect of auditor’s recommendation at Binghamton University Hypothesis 1: Student Performance There is an inverse relationship between class size and student performance. Binghamton University - Institutional Setting 9,300 Undergraduates 2,700 Graduate Students Highly Selective Outstanding retention and graduation rates Often cited among top universities in the U.S. Colleges Within the University Decker School of Nursing Harpur College of Arts & Sciences Management Watson School of Engineering Education & Human Development (no freshmen) Data Sample: Performance Freshmen entering Fall 1996: • First-time • Full-time • Degree-Seeking Fall 1996 & Spring 1997 Semester Courses One record per student per course Dependent Variable: Grade in Course End-of-Semester Course Grade Collapse +’s and -’s • Convenience • No difference Assign Numeric Values (0-4) Distribution of Grades in Sample A’s = 33% B’s = 39% C’s = 20% D’s = 4% F’s = 4% (n=9345) Explanatory Variable: Class Size End-of-Semester Enrollments Cross-listed Courses Average = 142 Linear versus Non-Linear Specification Possible Form of Relationship Between Class Size and Grades + Negative Linear Grade - - Class Size + Possible Form of Relationship Between Class Size and Grades + Negative Log Grade - - Class Size + Characteristics of Negative Log Grade Negative Log 90 80 70 69 10 20 60 70 Class Size Ordered Logit Regression Categorical Dependent Variable More than 2 categories Categories are ordered n Chi-Squared Significance Variable Constant Class Size EOP Student Management Student Nursing Student Engineering/Computer Science Student Age of Student Female Student Unknown Ethnicity Black Hispanic Asian Non-Resident Alien TOEFL Test High School Rank SAT Verbal SAT Math Science or Math Course Subsection Subsection*Class Size Subsection (Science/Math) Sci./Math Subsection*Class Size Model II (Linear) Model IV (Log) 9,345 2,030 .0000 9,345 2,129 .0000 Coefficient -2.9800*** -.0019*** .0887 -.0452 -.1414 -.5369*** .1296*** .1890*** -.1412 -.3686*** -.3057*** -.3048*** -.4084 1.1593*** .0206*** .0022*** .0037*** -1.1525*** -.7556*** .0004 -.1224 -.0014*** Std. Error .7311 .0002 .0982 .0675 .1249 .0693 .0371 .0436 .1058 .1004 .0914 .0577 .2758 .1415 .0019 .0003 .0003 .0576 .0905 .0006 .1428 .0005 Coefficient -1.8019** -.3221*** .0168 -.0411 -.1539 -.5249*** .1232*** .1792*** -.1431 -.3882*** -.3164*** -.3115*** -.4071 1.1770*** .0207*** .0022*** .0037*** -1.0771*** -1.3707*** .1842*** 1.2803** -.3058*** Std. Error .7401 .0246 .0988 .0679 .1258 .0689 .0372 .0437 .1061 .1006 .0917 .0577 .2770 .1416 .0019 .0003 .0003 .0570 .2997 .0634 .5097 .0950 Effect of Class Size on Freshman Grades 0.6 Predicted Probability 0.5 0.4 0.3 0.2 0.1 0 5 20 35 50 65 80 95 110 125 140 155 170 185 200 215 230 245 260 275 290 Class Size F D C B A Additional Variables of Interest Variable Significant? EOP Students SAT Verbal SAT Math High School Rank Science and Math n y y y y Direction + + + - Additional Variables (cont’d) Variable Significant? Direction Subsection Subsection*Class Size y y + Subsection (Sci./Math) Sub. (Sci./Math)* Class Size y + y - Hypothesis 2: Student Retention Students with high average class sizes are less likely to be retained. Data Sample: Retention Freshmen entering Fall 1996: • First-time • Full-time • Degree-Seeking Fall 1996 & Spring 1997 Semesters Dependent Variable: Retention Check Fall 1997 Enrollment • If yes, Retained = 1 • If no, Retained = 0 Distribution: • Not Retained = 9% • Retained = 91% Explanatory Variable: Average Class Size Student’s Average Class Size Across Fall 1996 & Spring 1997 Semesters Average = 99 Linear versus Non-Linear Specification Linear Model Log Model N Chi-Squared Significance Variable Coefficient 1221 87.94 .0001 n Chi-Squared Significance Prob. Level Variable Constant Average Class Size (Natural Log) EOP Student Management Student Nursing Student Engineering/Computer Science Student Age of Student Female Student Unknown Ethnicity Black Hispanic Asian Non-Resident Alien First-Year GPA (Natural Log) High School Rank SAT Verbal SAT Math % Science and Math Credits Constant Average Class Size 2.8268 -.0093 .5102 .0153 EOP Student Management Student Nursing Student Engineering/Computer Science Student Age of Student Female Student Unknown Ethnicity Black Hispanic Asian Non-Resident Alien First-Year GPA (Natural Log) High School Rank SAT Verbal SAT Math % Science and Math Credits -.0053 .6430 1.1079 -.2090 .9935 .1634 .2916 .5976 .1053 -.3736 .2817 .9981 -.5455 .0035 -.7111 2.8506 -.0150 -.0018 -.0031 .8580 .6226 .1271 .6361 .2289 .2388 .9911 .5278 .0001 .2431 .2713 .1186 .2060 1221 87.10 .0001 Coefficient Prob. Level 5.6162 -.8214 .2380 .0271 -.0342 .6546 1.1767 -.1674 .9580 .1559 .2613 .6706 .1102 -.3866 .2700 .9738 -.5554 -.0050 -.7029 2.8609 -.0151 -.0019 -.0031 .7248 .6057 .1139 .6474 .2351 .2299 .9869 .5321 .0001 .2393 .2609 .1190 .2727 Effect of Class Size on Freshman Retention 1 Predicted Probability 0.9 0.8 0.7 0.6 0.5 5 20 35 50 65 80 95 110 125 140 155 170 185 200 215 230 245 260 275 290 Class Size Additional Variables of Interest Variable Significant? EOP Students SAT Verbal SAT Math High School Rank % Sci./Math First-Year GPA n n n n n y Direction + Summary of Findings on Performance Negative relationship between class size and student performance (grades) Probability of getting an A decreases at a decreasing rate Summary of Findings on Retention Negative relationship between average class size and freshman retention Probability of being retained decreases at an increasing rate Contributions of Study Analyzes effects of class size at a post-secondary institution Tests the effect of class size on variable other than student performance: retention Large n analysis Future Research Add data from multiple institutions • Current results apply to • Binghamton University only Add faculty variables, e.g., Rank Test sensitivity of results by using mid-semester enrollments