“We normalize the data, not the outcome”

November 2012

Conversation with Philip Winckle, Executive Director of PECDC, on – PECDC’s formation – data sharing – implementing proper safeguards

If you do not see the video already displayed below, please select the alternative sites from the tabs here:

Here is the transcript of the video.

1. Establishing PECDC

Emmanuel Daniel (ED): What is the Pan European Credit Data Consortium (PECDC), and how did it come about?

Philip Winckle (PW): PECDC started in 2004 as a collection of five or six banks who were dissatisfied with the fact that we couldn’t get data on our own model loss-given default. Loss-given default was necessary for banks to work, and there were banks in Europe that we recognised didn’t have enough data on their own, but together, they probably did. Over the course of a year or so, we put together this not-for-profit, by-banks-for-banks consortium and started collecting the data. Once we realised what we had, it snowballed.

ED: What do you have right now in terms of data? What is the size of your database?

PW: We’ve collected data from 35 member banks, and in terms of numbers, that’s about 40,000 borrowers worth of data. When you add the size of the database, it’s monstrous. When split, it has a critical mass for all the low-default portfolio side so that banks can start to model.

ED: How robust does a loss-given default database has to be, to be useful for a Basel II modeling capability, assuming that the European banks did not have a robust enough database of their own individually at the beginning of Basel II?

PW: It depends very much on the portfolio. For example, if you take a very low-default portfolio such as shipping –shipping is clearly a global asset class because the vessels themselves move around – doesn’t really matter where they’re flagged. There are a finite number of vessels and a finite number of defaults. Let’s say that there are, over the last 10 years, 250 defaults of vessel loans in the world If we could capture 120 of those, we’ve probably got a big enough sample to model with.

ED: What assets do you look at – corporates, sovereigns?

PW: We look at corporates, sovereigns, banks, aircraft finance, shipping finance, SLE loans, real estate loans, and any other special interest areas that the banks feel need to be specially targeted.

ED: With that range of asset classes, standardisation and in creating the relevant documentation, it must have taken massive effort to get all the data streamlined at the beginning.

PW: Yes. But we didn’t start from nothing. We hired an external outsourcing company –Algorithmics – who had already run a similar database in the US. Algorithmics owned the database and had rights to that model but was kind enough to give it to us as a kickoff point and allowed us to develop it. We then put a team of experienced banks modelers together, and they poured over that model and redeveloped it. For that, we have Algorithmics to thanks as our first outsourcing data agent.

ED: How European-centric did the database need to be, in order for the data to be useful PECDC’s original bank members? Was it very different from Algorithmics’ US modeling techniques?

PW: Yes. In previous attempts to model loss-given default, they failed on the fact that it wasn’t detailed enough. Banks needed to show regulators that they had enough data to understand the inherent risks and how to build the models out of it.

ED: How was the database used by PECDC member banks?

PW: The purpose of the PECDC is to help banks model and benchmark their internal models. They’ve got to understand what they’re doing and learn from the data as well as from their own experiences. No single bank can do this on their own, so PECDC enables them to model directly off the database.

ED: Did the process show up any shortcomings in Basel II?

PW: It showed the shortcomings in banks’ own data collection processes.

ED: That’s probably a given because the business of collecting data was still underdeveloped.

PW: The banks eventually got over that initial difficulty of collecting data. Once they put it to use and started to build the models or benchmark the models out of that, they were able to find significant differences between what they’re modeling and the standard 45% for unsecured LGD set by regulators. It’s fair to say in all cases that the banks have used this data properly and been able to achieve a much more accurate measurement of their risks.

ED: Which types of banks readily became PECDC members, and which types are still resistant today?

PW: There are those which came onboard quite easily – these are banks who were quite advanced in their Basel II preparation and were looking for that final piece of evidence to convince regulators that their internal models were just right. They may have run an internal model with an average loss-given default slightly better than 45%, and we provided that extra bit. These banks were easier to convince, and in fact, were some of the founding members. The second type that was easy to convince comprised the smaller banks who realised they would never, ever have enough data on their own, and the only way to get into the game was to join such a consortium. They also realised that they could gain more experience by learning from the larger banks that were also members of PECDC.

The most difficult banks to persuade were those which are not yet ready because they possess a lot of data but it’s not centralised. They may be European bank branches of global banks which have data in different places around the world. These banks want to put it all together themselves, before joining an organisation such as ours.

ED: By 2009, it was very clear that Europe was in trouble. What was your experience during that phase, and what sort of data were you looking at by then?

PW: We were up and ready to collect data at that point and the data was starting to pour in. The interesting thing is that the data you just described is now starting to mature. A challenging aspect about loss-given default collection lies in the fact that it takes, on average, two and a half years to clear or resolve a case – either lose or get the money back. The answer is normally somewhere in between. From the 2009 peak defaults across Europe, banks have seen substantial recoveries in 2011 and even now in 2012.

2. Addressing sovereign default risk in Europe

ED: How different is that compared to sovereign default, and the fact that there’s so much fear of sovereign default in Europe itself today?

PW: I would like to say that our database is going to be able to solve the sovereign default statistical problem. We will most likely be able to garner most of the sovereign defaults in the world over the next 10 years. I would say that that is the best that anyone can do, and yet, that may not be enough to correct the model, the risk of sovereign default.

ED: Is there a temptation to enter the analysis business? Are you already in the analysis business?

PW: It is tempting. PECDC members meet twice yearly, and we have a good talk about these things. Some members would like us to give them a little more service and a little more help and wouldn’t mind paying. However, there are also members who are less willing to allow us to compete with their own efforts. The general fear is that someone comes out – whether it’s a rating agency or our own organization – and states a specific truth, saying, for example, “The reality is 23% loss-given default for large corporates”. Put any number on the table, and most member banks will regard it wrong or misleading.

ED: How does Basel III add a new dimension – a dimension of liquidity, dimension of value at risk?

PW: For PECDC, Basel III, probably in the credit space, is adding extra incentive to get the risk weight of assets down. A lot of banks, particularly in Europe, have found themselves caught with undercapitalisation.

ED: The definition of capital is still being worked out by regulators. How do you normalise what you’re putting together?

PW: What we normalise is the data. We don’t normalize the outcome. As of recently, there is a great demand amongst banks themselves to know what is a fair number. When they calculate risk weighted assets and look at the risks in their portfolios, the boards of directors at these banks want to know that they’re measuring correctly because, after all, we’ve had such a misunderstanding of what the bar measurement really meant.

Banks’ boards naturally demand more understanding of risk. Part of that understanding is to figure out if their calculations of risk on their respective balance sheets are correct or not. The only way to do that is to benchmark and look at how other banks are doing it.

ED: That’s a benchmarking type of exercise, which technically, you should be in a position to provide.

PW: Yes. That we do provide. There was actually demand from banks, from chief risk officers who might be able to collate all Pillar 3 reports from banks around the world, and still know nothing.

ED: A rating agency would listen to this conversation and say, “We look at the same data, so why are you recreating what we already do?”

PW: Rating agencies are, by definition, rating a sub-portfolio of reality. They rate those counterparties who want to be risk-rated and are prepared to pay for it. After all, they’re the ones paying the rating agencies. What we’ve learned from the data is that the counterparties who are rated by a rating agency perform differently in both loss-given default and default frequency.

3. Extending PECDC’s presence beyond Europe

ED: This whole concept of data sharing; is it well developed among banks? How are you extending your proposition outside of Europe right now?

PW: This isn’t a normal thing for banks to do, to share data in this sort of large corporate and specialised lending space. We keep reminding banks that most are already sharing portfolio defaults and non-performance data by retail borrowers. This includes sharing an individual’s personal data. Banks have already crossed that threshold. It’s not that difficult once we explain our position and the banks understand the safeguards in place. It’s not difficult to convince that banks can share data on an anonymous basis.

ED: Data is shared on an anonymous basis?

PW: Absolutely.

ED: How are PECDC’s expansion efforts, into both EU and non-European territories, going?

PW: We have two member banks in Norway, which is not part of EU, but definitely part of Europe. We do have participation from Australian and South African banks and they’ve been great supporters. We’ve been able to bring some Canadian banks on board and are starting to see some US banks join the consortium. US’ banks’ motivation for being part of PECDC is partly because of their interests in Europe, and PECDC is seen as a European specialist. But it is partly also because there is nothing in the US to rival this.

ED: You are one of the founding members of PECDC. Where does the motivation come from? What do you enjoy doing the most?

PW: My real motivation is to help the banking community learn from its mistakes. I started my banking career in Australia in 1979, when the property crisis hit in the 1980s. I then worked at a Swedish bank which experienced the Swedish crisis of 1992-93, which almost brought the country to its knees. I arrived at Singapore, and similarly, had to deal with the Asian financial crisis.

When I went back to Europe, I was there in time for the Baltic crisis and the sub-prime crisis. I’ve seen so many crises, I can pinpoint certain factors in those crises which are common. One of them is lending – lending foreign currencies to people who really can’t afford it or shouldn’t be able to handle that risk. Another factor is the inability to differentiate between different types of borrowers and different types of collateral. It’s that latter that led me to want to keep a written record for banks that they can add to year after year.

My dream is that in 50 years’ time, we will have enough data for really robust bank models. There doesn’t always have to be losses, if banks take the proper safeguards and really understand their own risks.

ED: You’ve seen all of these crises. Interestingly, they were always foreign-exchange related. This comes at a time when foreign exchange risk is a temptation because it’s cheaper. Do you see the world going back to that stage with QE3, for example?

PW: The Greek debt crisis and the sovereign debt crisis across Southern Europe is yet another example of the foreign exchange crisis. Instead of borrowing in low-priced Swiss Francs, for example, they joined a common community and borrowed in low-priced Euros. I must admit, when I saw the Euro Zone forming and the common currency, I should have been able to say, “My goodness, this is going lead to fixed exchange rates for economies that can’t afford them.” It’s effectively countries living beyond their means, in a bigger context.

ED: A lot of that is underlined by property as an asset. Would it be different if it was security as an asset?

PW: Security itself is a proxy for something underlying it. A piece of real estate is a collection of bundled cash flows going into the future. That’s my analytical viewpoint. When I see a piece of real estate, I see rental flows stretching into the future. When I see a bond, I see bond coupon payments stretching into the future; it’s the same with an asset-backed security. So no, I think it’s all the same.

Download full Length Video (MP4)
Download full Length Audio (MP3)


You must be logged in to post a comment.