This paper addresses the importance of credit constraints explaining the gap on college enrollment between students coming from rich and poor families. Measuring the effect of credit constraints on college enrollment is a complex task due to the unobserved nature of credit constraints and the existence of other variables that affect college enrollment that are unobserved as well. This paper exploits a natural experiment that produces variation on credit constraints directly, analyzing two programs that give college tuition loans to students who score above a given cut-off in the national college admission test. This enables a regression discontinuity design that addresses the problems of unobserved omitted variables and self-selection. Moreover, the paper uses a rich and detailed data set from a nationwide admission process that does not rely on unobserved subjective variables to select students, eliminating potential biases from the supply side. With this exogenous variation on loan access, I estimate the causal effect of credit constraints on college enrollment. The college enrollment rate increases significantly for students that are eligible for loans, and is statistically the same for all income groups after the elimination of credit constrains.