LAS VEGAS–The former head of the FDIC and the Fed’s former vice chairman for supervision shared their thoughts on what happened with the failed Silicon Valley Bank, the response by regulators, whether AI could have helped stem that crisis, the challenge in hiring people at regulatory agencies, and more.
During a discussion at the Money 20/20 Conference, Jelena McWilliams, who formerly headed the FDIC, and Randal Quarles, who was previously with the Fed, shared their insights, experiences and predictions. The session was moderated by Zach Anderson Petter, U.S. content director at Money 20/20.
Here's a look at what was discussed:
Petter: With Silicon Valley Bank’s failure, it’s been a turbulent year for banking. How many failures this year? What has this year been like for banking?
McWilliams: The system was not built to have no bank failures. That would be artificial maintenance. Sometimes, you have failures. Some are due to fraud, some to capital inadequacy and other reasons.
In almost four years at FDIC, only eight banks failed, which I think makes me the most successful FDIC chairman of all time (she said laughing). But the truth is we have had three large bank failures this year. The FDIC had built in protocols on how to take care of this, but I think they were taken by surprise by the speed and velocity, and I think those protocols were followed.
Quarles: I don’t think the failures were indicative of the health of the system, nor were they surprising. When you have rapid inflation and the Fed responds to contain by raising interest rates even further, then financial institutions that have the type of assets these institutions in particular did, you are going to have failures.
We saw this in the savings and loan crisis of the 1980s. They have very different sensitive assets so, it should not have been unexpected and that really is the principle cause of the pressure that sort of blew up in the spring.
Petter: It shouldn’t have been a surprise, but I think everyone here was surprised. Why? Were you thinking about this?
Quarles: The financial system is going to come under pressure when the Fed is responding to inflation. That was expected. We were expecting that even when I left the Fed in December of 2021. We were beginning to gear up to respond to inflation, but the speed with which things were moving in the spring was unexpected, because the depositors of these banks behaved in a way that bank depositors never have.
If you remember back to sort of the last big pressure, which was great financial crisis and the most severe bank run in U.S. history, there was one failure in September of 2008 when $16 billion left over the course of 10 days. Forty-two-billion dollars left Silicon Valley bank in a few hours on a Thursday and $100 billion was lined up to leave on Friday morning. So, that was something that I didn't expect and the bank management didn't expect and no one expected.
McWilliams: The Fed had 13 consecutive interest rate raises over a relatively short period of time. The fact that the banks either didn’t stress or didn’t stress enough the volatility of their portfolios was something that was, frankly, surprising to me. Billions were lost in deposits in a matter of hours. What was surprising to me vis a vis the FDIC is that they were so caught off guard. (Quarles’ successor at the Fed) successor said he heard about it Thursday morning. My successor said he heard about it Thursday night. It went into receivership Friday morning. (McWilliams said she had heard from some people expressing concerns as early as Monday.)
Petter: Is that a cultural thing, where middle management isn’t wanting to tell the higher-ups what is going on? How does that happen?
McWilliams: I would say from a FDIC perspective, I don’t know that it is a cultural problem, as much as a data (issue). When I was at FDIC I was surprised by how the FDIC collects data, through call report info. The big banks 5322s have about 2,400 data fields, the small banks about 1,200. It gets released in the quarterly basis profile. The second data collection is the exams, but…if you are a well-rated bank, it’s every 18 months. So, when you think about the data flow to the FDIC, it is limited, sporadic and a snapshot in time. There is no instant reporting.
Petter: Which brings me back to culture. People are shaking their heads over why this does not exist yet? As I have worked through my life in finance, I want to be a regulator one day. Regulation is the thing. Policy making is the thing that drives everything we do. But it sounds scary culturally.
Quarles: Supervisors have a very hard job. There are four thousand banks that need supervising. It's been a balance of trying to empower the supervisors in the field and then the supervised entities say, ‘Well we get entirely different answers depending on who you talking to.’ So, then the pendulum swings back too much and then the regulated entity says, ‘We can't get an answer from anyone because all 4,000 questions have come back to Washington.
And there's an environment, too, and it's perfectly understandable, in which the people who are on the ground doing the supervision and seeking to have the institution do its best job (want) to avoid political intervention in technocratic decisions. Their instinct is to say, ‘We're not going to send it up to a big political view in Washington.’ That results in things like the person at the top of the system at the Fed learning about (Silicon Valley Bank’s pending failure on a) Thursday afternoon.
Petter: Can the solution for this technologically get built inside the government?
McWilliams: We started working on this project when I was at the FDIC, but it wasn’t that the government would build it. In the government you go through this arduous process that takes a long time and the reason it ended up not being done was we started with 40 different entrants, then narrowed that to 14 and then to four. We were going to invest about $5 million into each and then see which of the projects actually worked best.
With the Administrative Procedures Act you can only test things and ask questions of nine banks (maximum). The idea was there for the FDIC to develop this. We would then insource the usage or create a private-public partnership.
Quarles: I think it was a well-designed project, and conceptually and fundamentally you can get to what you described, which is technology helping regulators with this data analysis problem. But it can’t be built in the government. We can’t pay the tech people enough in cash or equity. You can’t compete. There are limitations to the government process.
By the time a government entity has built a technology it will be fine for five years ago when the process began, and it won’t be able to be updated well enough. I think ultimately there has to be significant involvement of the private sector.
Petter: Do you think large language models and AI in general could have, one, a real impact in policy decisions, including those that are interest-rate related? What is the potential impact? Would we be making a different interest rate decision?
Quarles: AI, as it currently exists, I don’t think, would be helpful to monetary or supervisory policy. But it’s getting better all the time and a lot would depend on the data it has to draw upon. In my interactions with AI, it’s not quite ready for prime time. But it will get there. One thing it could do is that all of the experience we have could be brought to bear on a particular question.
One thing that has happened is there are fewer and fewer people who are around who lived through the S&L crisis. There are us older people who have said we have seen this before. So, some sort of technology that could ensure that experience is not lost could be useful.
McWilliams: From a bank supervisory perspective, from a loan perspective, bank examiners look at a sample, they don’t look at every loan. Based on that sample they rate a bank. So, I do think there is tremendous potential to get us to a place that maybe looks at all the loans in a portfolio. A lot of this is going to depend on the inputs into the system. I do think there is great potential.
Petter: Are the regulators able to hire the right people?
McWilliams: (The former FDIC chair said she had meetings, including in Silicon Valley, in which she sought to recruit talent to the regulator). The response I got was, ‘Why would I want to work for the FDIC? They said to me, ‘You will leave in three years and the next person will undo everything we have worked on. These are not great jobs, but they are meaningful.
Quarles: What the financial regulatory agencies need are people who understand current developments in tech and who can keep up, but they also need people who are comfortable with risk management, instead of risk elimination. Our job is to distribute risk, not eliminate risk.