Web Appendix for An Empirical Investigation of the Option Value of College Enrollment

The dataset used in estimation and simulation was constructed from several sources. Table A1 provides an overview of the main variables used in the analysis. The sample of individuals comes from the National Educational Longitudinal Study (NELS). The NELS is a longitudinal survey of a representative sample of U.S. 8th graders in 1988. Interviews were conducted in 1988, 1990, 1992, 1994, and 2000 and complete college transcripts were obtained for most individuals in 2000. The core schooling outcome variables, including yearly grade point average and indicators for enrollment were constructed directly from the college transcripts. The transcripts consist of course-specific records, including student ID, college IPEDS ID number, subject, month and year, credits, letter grade, and standardized numeric grade on a four-point scale. Course-level records were aggregated up to the student x college x term level to identify the primary school enrolled in, and then to the student x year level. The final transcript data contains student x year records of credits attempted, credits earned, grade point average, and several other variables. Individuals were considered enrolled during academic year t if they attempted at least six course credits (the traditional definition of part-time enrollment) at a twoor four-year college during both the Fall and Spring semesters of year t . The model describes college dropout, so I categorize people according to their number of years of continuous enrollment. Students who "stop-out," but eventually return and possibly graduate are grouped with students who dropout permanently in the same year. From the 1992 NELS surveys I utilize high school grade point average, standardized test scores, parents’ highest education level, and family income during high school. I convert NELS senior year test scores into AFQT percentile scores using the cross-walk developed by RAND researchers in M. Rebecca Kilburn, Lawrence M. Hanser, and Jacob A. Klerman (1998).