35,000 principles of economics students: some lessons learned.

• Author: Elzinga, Kenneth G.

1. Introduction

   One of the most dismal assertions ever made about teaching the "dismal science" was tendered by George J. Stigler in his American Economic Association presidential address. Stigler suggested that after five years, students would retain little or nothing of what they learned in a principles of economics course (1963). If Stigler is correct, this may explain why so many economists can restrain their enthusiasm for teaching introductory courses: a nagging concern that nothing of lasting value results from their labor. Nonetheless, in U.S. colleges and universities, the principles of economics course remains the most common entry point for economic education, and the number of students majoring in economics has risen in recent years (Siegfried 2006).

   Given such enormous enrollment in principles of economics, it is not surprising that economists have been interested in evaluating this class more than any other in the economics curriculum. (1) For example, a 33-year review of the Journal of Economic Education (1975-2007) produced 72 articles on the principles course, far more than on any other class in economics. (2)

   Presumably, economists believe that learning the elements of economic analysis can be one way for students to invest in human capital during their years of college and university education. (3) Indeed, ever since Becker's seminal work (1964), teachers of economic principles have emphasized to their students that by studying economics, they are making an investment in their own human capital. (4) Because economic education is a form of capital investment, and because so many students first encounter economics in a principles course, it would be fruitful if economists better understood what accounts for different rates of capital accumulation among students. Our study is in this grain.

   Over 40 years ago, Bach and Saunders (1965) concluded that men do better (in terms of grades) in economics courses than do women. Several subsequent studies have similarly examined what difference in the educational output of the principles course is associated with various exogenous variables. However, many of the previous studies have been limited in scope because of sample size and data availability. Through our connection with the University of Virginia (UVA), we have developed an unusually rich data set of students who took principles of microeconomics (Econ 201) over a period of almost 20 years from Kenneth G. Elzinga, the senior author of this paper. (5)

2. Data Set and Questions Explored

   Elzinga's teaching career predates computerized retention of certain bits of data. Consequently, our analysis is based on data gathered from administrative records pertaining to the 15,616 students who have taken Econ 201 from Elzinga since 1989. (6) This smaller but still substantial data set has complete and comparable records for each student that contain demographic characteristics, precollege achievement measures, and grades attained while attending UVA. (See Table 1 for descriptive statistics and simple correlations between all variables and a student's grade in Econ 201.)

   Topics we explore through an analysis of this data set are conventional ones, such as male-female and racial differences in class performance, first year versus upper class differences in achievement, and high school SAT scores and advanced placement (AP) credit as predictors of success in the college-level introductory course. (7) In addition, because of the nature of our data set, we can explore relatively unexamined terrain, such as in-state, out-of-state, and legacy student differences in class performance, as well as differences in performance between students in the College of Arts and Sciences and those enrolled in the School of Engineering and Applied Sciences and the School of Architecture at UVA. In addition, we examine performance differentials between athletes and transfer students from the general student population. (8) After assessing what factors affect a student's performance in Econ 201, we endeavor to ascertain how success in an introductory microeconomics course predicts performance in more advanced college courses.

3. Econ 201: Ceteris Paribus Conditions Met and Unmet

   In order to test variables that interest us (and might be of interest to other scholars of economic education), we assume that certain variables remain the same during the time of analysis (or that any changes are of no consequence). The following factors have remained unchanged over the decades in Elzinga's Econ 201 course:

   (i) The class format has not changed in 40 years--the course has always met during the fall semester of a traditional two-semester academic calendar on Tuesdays and Thursdays in back-to-back time slots (11:00 and 12:30 in the same 500-seat auditorium). (9)

   (ii) The assigned textbook always has been the "micro-halt" of a comprehensive textbook.

   (iii) Educational technology has remained unchanged (microphone and overhead projector). (10)

   (iv) The test format has been constant. (11)

   (v) The number of credit hours has remained constant at three.

   (vi) The class has consistently been offered in the College of Arts and Sciences at UVA and always has been taught in the liberal arts grain (not as a course in business administration).

   (vii) Enrollment in Econ 201 always has been large (in the hundreds). (12)

   We are unaware of any other principles of economics course with this degree of continuity and large student population.

   Naturally, some elements of the educational mix have changed over the years. First, while the textbook has always been a comprehensive one, the assigned textbook has not always been by the same authors. (13) Second, the quality of classroom teaching may have changed over time. While Elzinga has been the course's instructor for almost 40 years, he likes to believe that he is a better teacher today than at the start of his career. (14) Third, the quality of students may have changed over time--UVA contends it is a higher caliber institution than it was 40 years ago and one that attracts better students. Fourth, the deadline for students to drop a course was changed during this time period. Prior to 1996, the undergraduate course drop date occurred six weeks after the first day of class, but beginning in the fall of 1996, the deadline to drop a course was pushed up to two weeks after the first day of class. Students now presumably have less information to predict whether they are in danger of failing the course and whether they should withdraw before receiving a low grade. Finally, while Econ 201 at UVA continues to use a large lecture format, students meet with teaching assistants (TAs) in a small discussion section format for one contact hour per week. In any given year, disparity among the quality of TA instruction could have a significant impact on student performance.

4. Predicting Econ 201 Grades: The Model

   Owing to the nonlinear, discrete nature of the dependent variable in this study (a student's grade in Econ 201 being based on a four-point plus/minus scale), we used an ordered probit regression model. Our model is similar to that of Yang and Raehsler (2005), which was used to predict what factors contribute to a student's performance in an intermediate microeconomics course at Clarion University. Since grade values are inherently ordinal in nature, they readily lend themselves to this type of analysis.

   A general ordered probit model uses observed points in a data set to predict values of an unobserved (latent) variable, [y.sup.*], such that

   [y.sup.*] = x[beta] + [epsilon], (1)

   where x are explanatory variable matrices, [beta] are parameter matrices, and [epsilon] are independent, identically distributed error terms. In our study, the latent variable is a measure of educational output corresponding to a choice problem on behalf of Elzinga's students, while the descriptive characteristics and prior performance measures capture what affects this choice. Since the latent variable measures educational output and has unknown units, the cutpoints of the model will not be readily interpretable.

   Our most general regression uses all demographic characteristics and precollege achievement measures as explanatory variables and is defined by

   [Y.sub.i] = [[beta].sub.1] FIRST [YEAR.sub.i] + [[beta].sub.2] SECOND [YEAR.sub.i] + [[beta].sub.3] THIRD [YEAR.sub.i] + [[beta].sub.4] [MALE.sub.i] + [[beta].sub.5] [BLACK.sub.i] + [[beta].sub.6] [ASIAN.sub.i] + [[beta].sub.7] NATIVE [AMERICAN.sub.i] + [[beta].sub.8] [HISPANIC.sub.i] + [[beta].sub.9] [INTERNATIONAL.sub.i] + [[beta].sub.10] [ENGINEERING.sub.i] + [[beta].sub.11] [ARCHITECTURE.sub.i] + [[beta].sub.12] IN-[STATE.sub.i] + [[beta].sub.13] [LEGACY.sub.i] + [[beta].sub.14] HIGH PROFILE [ATHLETE.sub.i] + [[beta].sub.15] [ATHLETE.sub.i] + [[beta].sub.16] CC [TRANSFER.sub.i] + [[beta].sub.17] [TRANSFER.sub.i] + [[beta].sub.18] SAT [VERBAL.sub.i] + [[beta].sub.19] SAT [MATH.sub.i] + [[beta].sub.20] AP [CREDITS.sub.i] + [[beta].sub.21] [PRE-DROP.sub.i] + [[epsilon].sub.i]. (2)

   In order to account for individual- and time-specific unobservable heterogeneity, we define our error structure...