Association for Public Policy Analysis & Management

In 2002, the Colombian government, with the support from the World Bank, introduced a subsidized loan program—Acceso con Calidad a la Educacion Superior (Access with Quality to Higher Education, or ACCES)—to increase students’ access to higher education. This program has provided college loans to nearly a half million Colombians over the past decade. The ACCES program determined prospective student’ loan eligibility based on their scores on the high school exit exam (Saber11, equivalent to SATs in the US). Between 2003 and 2008, eligibility was determined by students’ scores across different subjects (e.g., minimum eligibility in Math, Reading and Science tests separately).  That is, even if the overall Saber11 score was adequate, a low grade on one section of the exam would result in a student becoming ineligible for a loan.  Beginning in 2009, eligibility was determined by a weighted average of Saber11 scores across all subject tests, and to balance regional inequalities, different cutoffs were established for each region.

There is relatively little published research on the efficacy of the ACCES program. The one notable exception is a paper (Melguizo et Al. 2016[1]). They use a single composite ACCES test score prior to 2009 rather than use the minimum subject test scores when determining the loan eligibility of students in Discontinuity Design (RDD). In contrast, we capture the complexity of the 2003-2009 eligibility cutoffs as well as the variation in eligibility rules over time in our evaluation of the ACCES program.  This allows a cleaner depiction of the eligibility cutoff for the RDD.  Our attention to the assignment rules allows us to better evaluate the efficacy of different institutional policies on educational outcomes overall, by region, and over time. Educational outcomes considered include college/university enrollment, academic performance, as well as graduation and dropout rates. We will follow the approach of Wong (2013)[2] and Reardon & Robinson (2012)[3] both of which introduce RDDs where there are multiple dimensions in the cutoff.

From a policy perspective, we want to take advantage of the richness in variation of the assignment rules and to provide insights for policymakers on what works best in terms of the outcomes of the policy, both in terms of efficiency and in terms of the regional distribution disparities. We expect to answer which assignment rule (individual tests or aggregated scores) results in better outcomes, as well as how much inclusion costs in terms of increased risk by paying attention to the regional disparities and changes.