| Popis: |
This paper develops a demand model for bank loans with a two-step decision process. In the first step, the agent chooses the financial institution from which she would like to borrow. In the second step, conditioned in the first decision, the agent chooses the desired amount of the loan. The model belongs to the Generalized Linear Latent Mixed Model (GLLAMM) class whose flexible structure is convenient to combine a discrete distribution (first step) with a continuous one (second step). The model is empirically estimated using a sample of data from the credit bureau system SCR from the Brazilian Central Bank. The results suggest that variables like the loan maturity and its risk classification are more relevant in the choice of the financial institution whereas variables like the total number of loans the borrower has in the SCR, value of the loans still to mature, existence of guarantees are more relevant in the amount to be borrowed. Other variables like the interest rate, years of relationship, and firm size show to be important in both steps. |
