Working Paper: NBER ID: w22725
Authors: Evan Riehl; Juan E. Saavedra; Miguel Urquiola
Abstract: This paper explores the implications of measuring college productivity in two different dimensions: earning and learning. We compute system-wide measures using administrative data from the country of Colombia that link social security records to students’ performance on a national college graduation exam. In each case we can control for individuals’ college entrance exam scores in an approach akin to teacher value added models. We present three main findings: 1) colleges’ earning and learning productivities are far from perfectly correlated, with private institutions receiving relatively higher rankings under earning measures than under learning measures; 2) earning measures are significantly more correlated with student socioeconomic status than learning measures; and 3) in terms of rankings, earning measures tend to favor colleges with engineering and business majors, while colleges offering programs in the arts and sciences fare better under learning measures.
Keywords: college productivity; earning; learning; Colombia
JEL Codes: I23; J24; J44
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Colleges' earning productivity (D29) | Colleges' learning productivity (D29) |
| Socioeconomic status (I24) | Earning measures (J31) |
| Colleges' earning measures (D29) | External factors (parental connections) (J12) |
| Colleges' learning measures (I23) | Contribution to human capital (J24) |
| Mix of majors offered (M39) | Rankings of colleges based on earning measures (A14) |
| Mix of majors offered (M39) | Rankings of colleges based on learning measures (I23) |