Survive then thrive: determinants of success in the economics Ph.D. program

Economic Inquiry, Oct, 2007 by Wayne A. Grove, Donald H. Dutkowsky, Andrew Grodner

I. INTRODUCTION

Every spring, admission committees review a pile of applications to select candidates for their Ph.D. programs. Beyond normal attrition, Economics departments clearly have a stake in seeing a significant proportion of students finish the program. Aside from whatever resources may accrue to departments with high completion rates, they receive prestige from the placement and success of their completed Ph.D. students in academic institutions or private sector jobs. Obtaining a Ph.D. in economics requires clearing a series of distinct hurdles and constitutes a riskier venture than attending medical school, an MBA program, or law school. (1) Thus, information that helps identify success in completing the Ph.D. has significant value to doctoral admission committees, departments, and administrators.

This paper empirically investigates what determines successful completion of the Economics Ph.D. The data are retrieved from individual files of former Ph.D. students at Syracuse University (Carnegie Classification: Doctoral Research Universities II--Extensive). Our study breaks new ground in several areas. First, it represents an ex ante study of variables that determine doctoral degree completion since it uses only information known by the admission committee at the time of the selection process. (2) Second, in addition to demographic information and Graduate Record Exam (GRE) scores, we extract several important variables which have not been used in doctoral success studies. Third, we examine success for each of the distinct sequential stages of the Ph.D. program. Fourth, it provides results for a midlevel program rather than an elite one. In so doing, it focuses on the class of programs that produce the vast majority of Ph.D.s in economics. (3)

Section II introduces the models and discusses the data and variables. In Section III, we investigate the determinants of doctoral student success using logit and generalized ordered logit (GOL) estimations. Students who pass the comprehensive exams exhibit intellectual firepower (high verbal and quantitative GRE scores), have a Masters degree, and had a strong prior background in economics. But having passed the comps, completing the degree requires strong research motivation and math preparation. Research motivation, measured by whether the student mentioned in their personal statement a paper they had done, is a significant indicator for completion of the dissertation. The significant determinants of passing the comps generally become insignificant for the dissertation step. Section IV concludes the paper.

II. MODELS AND DATA

The models can all be expressed in the following form. For the ith student, i = 1, 2, ..., T, the probability of success in the jth step is P([Success.sub.i][[x.sub.i]) = F([x'.sub.i] [[beta].sub.j]), where the qualitative variable Success is a measure of success which may be binary or ordinal, F is a cumulative distribution function, x is a vector of exogenous variables, and [beta] is the parameter vector. Explicit descriptions for each of the estimated models appear in the Appendix.

The specification falls into the class of limited dependent variable models. The underlying latent dependent variable can be regarded as the cumulative number of unobservable "performance units" that determine success. If the student's units meet or exceed the standard, the student succeeds or passes; otherwise, he/she fails. Given the individual's ex ante characteristics, the student acquires these performance units in the graduate program based upon academic ability, work ethic, and behavioral characteristics.

We obtain the data by reviewing all the available files of recent Ph.D. students in the Syracuse University economics department. The sample consists entirely of students who either completed all the requirements, failed the theory or field comprehensive exam in two attempts, or left the program voluntarily. It includes no current students, even if they have finished one or more steps. As a result, the sample size is the same for all our estimations. Extracting this detailed individual information from well over 100 files yields 78 observations with data for all the outcome variables and determinants.

Summary statistics appear in Table 1. The first three rows of data consist of the Success or outcome variables, which encompass the major steps in the Ph.D. program. The variable Theory Comp equals 1 if the student passed the theory comprehensive exam, 0 otherwise. Field Comp and Completed are defined correspondingly based upon the field comprehensive exam and the dissertation. We do not distinguish between whether the student passed the Theory or Field Comp on the first or second attempt.

Students who left the program before attempting a given step receive a value of 0 for this outcome. Some students in this group decided that they do not have the performance units to reach the expected standard and chose not to try. Others who transferred before attempting the outcome (they receive a value of 1 on any previous outcomes in which they succeeded) may have the academic abilities to succeed but did not seek to obtain the necessary performance units through study. We do not distinguish between voluntary and involuntary leavers, since our goal is to evaluate what determines which students completed the degree in this program.