Exploring firm-level data we empirically study the degree of asymmetric information in the Brazilian credit market, and test whether the usual predictions of theoretical models regarding asymmetric information are observable in our data. Considering that the credit decision is based on credit risk models, we have investigated how private data, which can only be accessed by some lenders (usually incumbents) can improve the models and produce advantages for competitors granted access to said data. Our findings suggest that private information substantially increases the accuracy of credit models, enabling competitors who are granted access the opportunity to increase their potential client portfolio by more than 100%, without facing a higher default rate, which suggests that the degree of asymmetric information is highly significant. In addition, our results provide evidence that new competitors face adverse selection conditions, since riskier clients are 27% more likely to migrate from their original bank, and subsequently have a 29% higher default rate. Nevertheless, by exploring the value of private information in different clusters of clients, we have found evidence that the usual data available in the Brazilian market are useful for more easily identifying high-risk clients, but insufficient for identifying very low-risk clients. Finally, we have tested a random forest model to investigate whether using a more modern modeling approach could replace reliance on private information but have not found evidence to support this.
Complete tool for submission, evaluation and publication of proceedings for scientific events.
