Almost a month ago now I attended the wonderful 3rd Bristol Workshop in Banking and Financial Intermediation at the University of Bristol. I was there to act as discussant of a paper titled Economics of voluntary information sharing, presented by Jason Sturgess from Queen Mary. Other names in the workshop included Marie Hoerova, Jean-Charles Rochet, o Javier Suarez.

The goal of the paper is to understand why banks might want to share information about their borrowers with other banks. In some countries, the sharing is mandatory—in Spain, for instance, banks send the information of their borrowers, such as loan amounts, identify, and repayments, to the credit register. This has allowed wonderful empirical banking papers in the recent years (AER, Econometrica, JPE).

But in other countries, such sharing is voluntary. And why would a bank share information with other banks about its own borrowers? Because it would get more information in return. In particular, the moment you join the credit bureau, you have access to the information of all the members. This might be very useful especially if you are a small bank.

Let’s say you are a small bank. New borrowers knock on your door to ask for credit. What do you know about them? Very little, since they are new to you. You have no prior relationship, and hence you cannot see how much they’ve been earning, whether they have paid the previous credit on time, etc. For all you know, they can be someone rejected from their previous bank for not paying. You are faced, hence, with adverse selection. If you could identify which borrowers have previously been rejected from those borrowers that are simply looking for new credit, life would be easier. And that’s what the credit bureau provides.

This is, in theory, less of an issue for a big bank that has a high market share. The reason is that this bank already knows many borrowers, and hence whenever a new borrowers shows up they are less likely to be a rejected one. And, by sharing information, the big dominating bank would be losing substantial information advantages.

All these stories can be empirically assessed with the data that Jose Liberti, Jason Sturgess, and Andrew Sutherland employ in the paper. They have the information contained in the credit bureau. As such, their information is restricted to current members—not all banks have joined it—but these members provided all the past credit history, not only from the moment they joined. Hence, they can study why banks decide to join earlier or later, what happens to credit supply after a lender joins, etc.

The paper confirms many of the previous intuitions—which are formalised in several models in the 90s and early 2000s. Given the granularity of the data, they are able to control for many potentially confounding factors. In the jargon of the sector, they can use plenty of fixed effects. This has been the trend in the last decade or so in empirical banking, as we saw the explosion of papers using national credit registers (I got some as well). What this paper brings to the table is the notion of voluntary information sharing, which I found very novel and extremely interesting.

As I final note, you can see below a picture of the place where we had dinner the day before the workshop.