The assignment of public credit guarantees is usually based on creditworthiness only. However, many financially sound firms may not need such support. The paper compares, in the 2011-15 period, the criterion used by the Italian Guarantee Fund with an alternative benchmark that, using machine learning methods, takes into account the probability that a given firm is credit constrained as well.
The use of machine learning methods could strengthen the positive impact of the public guarantee on the amount of loans granted to credit constrained firms, without increasing firms' default rate.
Published in: Journal of Economic Behavior & Organization, v. 198, pp. 434-475