Speaker: Alvaro Carril, Princeton University
Alvaro Carril, is a PHD student in Economics at Princeton University.
Abstract: This paper studies the impact of providing personalized information about college loans on students’ loan application behavior. We use a large-scale field experiment in Colombia, where we employ a chatbot to elicit beliefs about college loans, and to update these beliefs along different dimensions. We first document a prevalent bias where students underestimate their own eligibility to credit line options, and overestimate monthly repayment obligations on average, although there is substantial heterogeneity in the direction and magnitude of these biases. Exploiting random variation in information treatments and using an instrumental variables approach, our estimates suggest that personalized information about loan eligibility and repayment increases the probability of applying for a loan by 27% for compliers, although there is heterogeneity in the effects by students’ belief of their preferred degree’s average salary. The uptick in loan assignment is strictly due to the surge in applications, with a placebo chatbot treatment affirming the non-impact of the bot itself on loan application behavior. Of substantive policy importance, merely being assigned to the chatbot increases the probability of applying and obtaining a loan by 10% in our sample, and this increase is almost entirely driven by low-SES students considering high-earning degrees. These results underscore the prevalence of ill-informed decisions in the higher education market, and the potential of personalized information to improve the efficiency of student loan programs at scale.