Stress Testing

2013 Financial Markets Conference

April 2013

An interview with David Rowe, founder and president, David M. Rowe Risk Advisory

Mark Jensen: Hi, I'm Mark Jensen from the Federal Reserve Bank of Atlanta. My guest today is David Rowe, the founder and president of David Rowe Risk Advisory and also a regular contributor to Risk magazine on their risk management column. He was a speaker this year at our Financial Markets Conference on stress testing and measuring of risk. We wanted to ask him a few questions.

One of the things that came out in David's paper was this idea of adaptation. This is often looked at as being an adaptation between the modelers and management but from a Federal Reserve kind of point of view, you also throw in this question about how financial firms adapt to regulation. You make a point that it's often the Fed's behind the curve with regards to how we set policy. How do we overcome that? How is it possible to make rules and regulations that are timely, yet ones that aren't being arbitraged away by the modelers, by those in financial institutions?

David Rowe: Well, I think Mark, in my view, we probably have gone too far in the direction of trying to control risk in specific institutions, and in the system in general, through detailed regulatory intrusions and requirements. My own view tends to be we need to get back to the point that we literally do solve too big to fail. I have this somewhat iconoclastic view that if institutions are too big to fail, well, then they're just too big and they have to be broken up. But that's a somewhat controversial view. The ultimate way you can keep institutions under control is the fear of failure. My old boss before he was at the Fed, Alan Greenspan, famously said the optimal number of bank failures is not zero. The fear of failure and the reality of failure periodically is what keeps bankers on their toes.

I agree with Andy Haldane [executive director for financial stability at the Bank of England] as he described it, we need to go back to a kind of a bargain where we create a system where banks can fail, where the fallout on society is acceptable, is minimized, and banks will be allowed to fail.

Jensen: If you allow a financial institution to fail, will there be an incentive on their part to actually construct better measures of risk?

Rowe: I absolutely think yes, and I will tell you one reason I think so. I was at Bank of America at the time of the merger with NationsBank. NationsBank previously had bought a little company called CRT, Chicago Research and Trading [Group]. The executives at CRT had a pretty lean Christmas in 1987 after the October stock market decline of 500-plus points like out of the blue. There's a saying about the business they were in, they were in the business of picking up nickels in front of a bulldozer, but you got to keep your eye on the bulldozer. In this case, the bulldozer just about ran them over and they were probably lucky to survive the year.

Well, unlike a lot of people that were prop traders in a big financial institution, where the institution is absorbing the losses and they get the bonuses if they succeed, these guys were very serious about risk management. After 1987 at CRT, I was told, they instituted what they called level one and level two stress scenarios every single day and it was four standard deviations, then eight standard deviations. They looked at those numbers and obviously not expecting they were going to lose those kind of numbers with any frequency, but looking at it from the standpoint, could we survive a shock that big?

Jensen: So in the last few years the Federal Reserve has been responsible for carrying out what's known as CCAR [Comprehensive Capital Analysis and Review], this stress test of the largest $15 billion or more assets of a bank. In that process it seems like policies become a little bit more quantitative in nature. Is that a good thing to have the regulator being involved in modeling, or is it just another area where something wrong can happen?

Rowe: I think it's probably a good thing if it's being done independently. I do worry about...and this is one of my problems with the increasing complexity of Basel III. When you impose very complicated models on the system across the board, what you're doing is you're creating a homogeneity in the system, where everyone is doing things the same way; everybody is looking at risk the same way. 

An example that I use sometimes in presentations from other disciplines is in fisheries. There's a very big farm salmon industry off the coast in Chile, in the cold waters of the southern shores of Chile. They are not endemic to that [area], they don't have wild salmon; it's purely a farm salmon business. But they had raised these by breeding and a heavy use of antibiotics to try to prevent all of the kinds of diseases that could destroy salmon. But then in one year, I think it was 2009, a special form of bacteria managed to penetrate all of these defenses. They lost 87 percent of the population in one season. Homogeneity is a very significant indicator of susceptibility to serious periodic crises. But to have the Fed do its own independent analysis I think is good because it's forcing banks then to justify what they're doing and not simply take it for granted.

Jensen: There has been a lot of debate on whether or not the Federal Reserve should be more transparent about what their actual models are. Some of us take a view, well, is there a teachable moment where we report our results and we explain them, and then try to provide a little bit of education to those institutions, or is it just going to be gamed away?

Rowe: I think it's probably going to be gamed away. I mean part of the problem in the subprime mortgage market is as soon as the rating agencies reveal the process they were using to rate these tranches, immediately people started figuring out how to structure their portfolios to game their way around and get the rating they wanted without actually having the quality in the portfolio that was appropriate. Yes, it's a very tricky question, but I'm inclined to think that a certain amount of confidentiality about the specific way you're doing it at least, you can talk philosophically, but to get the parameters down now you run the risk that people are just going to be sure they can put things right across the margin where it falls into the lower-risk category.

Jensen: One of your takeaways from the financial crisis has been, you have to have structural imagination. From a stress-testing standpoint as a macroprudential regulator, there has to be a macro stress scenario. One thing that you also point out is that there's a lot of limitations to a model if you don't have the data. How do we provide a stressful scenario? Say, for instance, an interest rate hike, when we live through a period of four or five years where interest rates have been zero.

Rowe: Well, it differs in certain cases. I think actually simulating in the current context an interest rate shock is not all that hard. We know where interest rates have been historically and we know that we're well below any kind of historical norm. You can make a reasonable argument that if you had zero expected inflation that probably equilibrium interest rates are probably 3 percent. But if you got some inflation then, that pushes it up to five or six, maybe something like that. One of my worries is that if we put rates back to those levels, not just overnight but if they got there fairly quickly, let's say like as occurred in 1994 when the Fed began to raise rates pretty rapidly, we will see some problems come out of the woodwork, because I'm sure there are a fair number of institutions, both financial institutions and others, that have tried to take advantage of this prolonged period of near zero interest rates and very low interest rates to try to somehow arbitrage that and lever up because of the attraction of the low borrowing rates. And that's potentially going to create some serious problems.

Jensen: You weren't able to really talk about this during your talk, but you do in your paper for the conference, this idea of third-party vendors. You particularly talk about it in terms of a data issue. Because of the requirements that CCAR has placed on these financial institutions they've had to build up bulk data collection so that they get enough granular enough data to be able to answer the type of questions that are being asked of them in their projected losses. But it's also led to a number of financial firms utilizing third-party vendor software to be able to actually do these projected loss forecasts and estimates. How is that going to play out?

Rowe: Somehow this question of, "Do we buy software or do we build our own software?" becomes almost like a religious conviction. We are a build-it-yourself house or we always buy something else because it's cheaper. Neither of those views is very credible because software in any given application goes through a life cycle.

What's happened in the last 20 years is the industry has come to a consensus about how to do this. Vendors have worked on the technology and the best way of implementing it. Now a good example is under Basel II, there's this middle-level, standardized approach, which is more than mark-to-mark plus add-ons, gives you some offsets but it's not full simulation. I always thought that was kind of a useless zero member set. Because it really doesn't capture the full nuance of the dynamics of your portfolio with each counterparty, and yet it's a lot more complicated than mark-to-mark plus add-ons. Whereas today with a vendor, there are numerous vendor solutions that will give you the full simulation result, let you do netting the way it actually would be calculated scenario by scenario within each path that you simulate. I think what happens is generally when software becomes commoditized, basically, it starts to become cheaper. You don't want to be caught up with an in-house, unique system that has to be maintained, that's quite expensive, when other people are in fact essentially operating with far more efficient software.

Jensen: Well thank you, I'm Mark Jenson with David Rowe for the Financial Markets Conference of 2013. Thank you very much.

Rowe: OK, thank you, Mark.