All Singing All Dancing Risk Management

In late September, Buy-Side Technology held a roundtable discussion in London. On the agenda were two issues - managing risk across multiple asset classes on a single, integrated risk management platform; and managing risk in ever shrinking timeframes. By Victor Anderson

Julie, is a multi-asset-class single-platform approach to risk management possible? If it's not, is it a feasible goal towards which we should aim?

Julie Johnson, Morley Fund Management:

From my experience, it does not seem to be possible at the moment. We have done a lot of work looking at the various systems that are available, and we have not yet found one that meets our requirements for multi-asset risk management. As to whether or not it is feasible, I think it has to be there. Someone somewhere is managing the total risk; whether that is the fund owner as opposed to the fund manager, someone somewhere needs to understand the total risk that is inherent in the way the money is being managed. I think we have to find a solution to this, but it is not clear that it is there at the moment.

Dario, you feel strongly about this. Is StatPro anywhere near producing an all-singing, all-dancing risk management system?

Dario Cintioli, StatPro:

With our software today we integrate all types of risk factors onto a single platform, and we provide coverage of about 200 asset classes, covering anything from a collateralised debt obligation (CDO) to an accrual bond or a complex convertible bond. It took us seven years to get where we are - it was hard work - but the real issue is to be able to follow financial innovation in a significant way and to be able to expand this multi-asset class quickly and have a system able to integrate all types of risk. In terms of support for multiple asset classes, we feel pretty confident that we are there.

Do you have buy-side clients using the full breadth of StatPro Risk across multiple asset classes?

Cintioli:

Yes. Most of our clients use us across platforms - they use us for any type of funds, from simple equity funds to the most complex instruments that use many derivatives.

Brian, you specialise in data cleansing and data interrogation technology. In terms of multi-asset class support, are you there at the moment?

Brian Sentance, Xenomorph:

Yes, I think so. One of the biggest challenges is trying to get the data model to support any of these asset classes as quickly as possible. There are a whole range of products, from something quite simple, like private equity funds, which still might be quite difficult to model from a risk perspective, through to the more complex products. One of the problems is that there is a lot of rigidity in the data models of a lot of the risk management systems that are around.

Often it requires a client to go to the risk vendor and purchase a lot of expensive consultancy to extend the product - although it would be ideal for the client to be able to extend the product themselves. In that respect we are trying to develop a data management platform that integrates complex data, such as the put/call schedules and perhaps correlation data with things like swaptions, volatility cubes or correlation matrices.

All of this data, particularly given what has happened from a regulatory point of view, might actually be quite important, and so we are trying to integrate that with all of the pricing models to give you a pricing framework for being able to support these derivatives very quickly.

What are the advantages and disadvantages of taking a single-platform approach?

Alistair Falkner, Morse:

In boutique operations where they use risk models to drive their strategies, their risk system is part and parcel of how they make their money, and so coming to them and saying 'you are going to use a standardised model' just does not wash. They have to have the tools they need to make money. If you look at it from a middle-office perspective, then perhaps one risk system for the firm makes sense, but if you are looking at it purely in terms of generating alpha in the front office, I think they will forever want the tools they want, the best tools they can get, and that probably will remain best-of-breed.

Sentance:

We have one sell-side client who has multiple risk management systems based around different asset classes. They have tried to take the approach of centralising a lot of the pricing functionality, but they eventually decided that that is too much effort. They already have the pricing models in the risk management systems themselves.

They decided they are going to use the pricing models that are actually in the incumbent risk management systems and not try to centralise those to pull everything together. One problem they have is that they are trying to set up market scenarios and apply that to all of these systems, and they all take scenarios in different formats, so even to apply a common scenario set to all of these systems is difficult.

Cintioli:

We have clients who have multiple risk management systems which they use for different needs. They use us for cross-asset coverage, but if they have a specific system for doing special strategies on risk and optimising their portfolio, they go with a best-of-breed solution which serves specifically that need of that asset manager. In my view, they are perfectly compatible.

Johnson:

I would echo that. We use multiple risk systems, and even if we do find a system that covers a multi-asset class approach, we are not going to do away with the specialised systems that we use for modelling individual asset classes because they serve a different need.

The way we see it is like looking at a three-dimensional object from a number of different angles. The more different ways you can look at something, the more insight you get into it. If your risk models are telling you different things, that is when it is important to understand the underlying assumptions that are feeding into the risk models.

Sentance:

That is not a dissimilar approach from pricing model risk and having multiple vendors and cross-checking in terms of the sensitivity numbers coming out. To take the same approach with risk would also be very sensible.

Johnson:

You should have one primary system that you use for reporting because the last thing you want to do is report multiple prices to clients, or indeed multiple risk numbers for those clients who actually want risk numbers. You need one system that is your primary system; all of the others provide extra information to be used by those specialised areas that need it.

To me, the big challenge multi-asset risk models have not addressed is that they pile all the assets into one big pot and come up with a risk number (or a series of risk numbers) on the grounds that you can no longer represent risk with one number in the modern world. But then you ask, 'So what? What are you going to do about it? If you do not like the number, how do you make it different?' For that, you need to have drill down and understand how the different decisions feeding into the management of that portfolio are actually contributing to that risk number.

What is missing from multi-asset risk systems that we have seen is any kind of modelling of the hierarchical approach to fund management.

Julie, when we chatted last week you mentioned that Morley had fairly recently looked for a multi-asset risk system. What were the drivers pushing you towards that decision?

Johnson:

I think there were a number. Certainly the introduction of new and more complex instruments meant the systems we have at the moment, which are in-house developments on top of the standardised risk models, were unable to support some of the more complex derivatives. The main driver was therefore to try to find a model that could cope with the challenges presented primarily by the non-linearity or asymmetry of returns that is introduced by the presence of options. Those derivatives that have an element of optionality or credit really create the problems. The systems that we were using in the past were simple tracking error models that really could not model the asymmetry.

The other driver is the fact that the regulatory framework is changing. With UCITS III comes the opportunity to invest more heavily in more complex instruments, but also the obligation to provide risk management on a daily basis, in many cases. If you have a sophisticated fund, you have to be able to produce on a daily basis risk analysis, and our existing processes were much too cumbersome to be able to do that.

Sentance:

It was not long ago that certain asset managers were finding it difficult to report intra-day valuations, simply because it was a challenge getting hold of market data and making sure that was reliable. To head towards daily risk management, and perhaps going beyond that, is going to be an interesting challenge.

Given the increase in the number of complex instruments that traditional investment managers such as yourselves are looking to trade, do you believe that the technology is enabling the business to move in that direction, or is the technology lagging behind the business?

Johnson:

I think it is the latter, but I think that it is inevitable. Until you know what instruments are out there, you cannot anticipate what is coming. To some extent, we are always going to be playing catch up, but I do not have a problem with that. The difficulty is how long it takes to catch up; we cannot afford to wait as long as it has taken.

Sentance:

I think spreadsheets have a place as a short-term pressure-relief valve, but it still has to be managed very carefully because it is easy for it to become a monster in terms of established fund managers. Spreadsheets are still the most successful trading and risk management systems, in my view. Microsoft is doing some things about that with Excel Services and trying to run things on the server side so you can take a bit more control, but that is new, and it will probably take a little bit of time before people start using that.

Cintioli:

That is exactly the point. Not only is it difficult to give satisfying coverage to a client who uses a lot of structured and complex products today, but the real challenge is how to adapt and maintain, or improve, the coverage. My background is in investment banking, where we were creating these products, and we had to be fast to beat the competition in creating a new product. When we moved into risk management, we therefore first had to choose the right model that allowed us to be fast, which is difficult. Some models are very good, and potentially the best, but they have a lot of problems when they have to cover new asset classes, and it can be dramatically expensive in terms of time and development efforts. So, you first of all need a model that allows you to be fast.

Second, you need a process, and you may potentially need to buy additional data if it is a new asset class, which can present challenges. For example, CDO for us meant the purchase of the base correlation quoted in the market from the implicit prices of single-term CDOs. Some new prices are very easy - you can do them in a couple of days - but some others are a little bit more difficult because you need additional data.

For a technology or data vendor, one of the big challenges is knowing which direction to move and to identify from where demand is coming. Is that determined by interfacing with your clients?

Sentance:

I think there is a combination of sources in our internal process, the most obvious of which is from existing clients. We are working with them through consultancy projects and implementation projects, and through that we understand what is going on in that particular client's world.

We already work with clients and prospects to gather information, but we need to move to having a more formalised user group, which is something that we are going to be doing later in the year, particularly in the area of data management for risk.

Dario, you mentioned that you are adding, on average, five new pricing functions per month. How do you determine what you add to StatPro Risk?

Cintioli:

Our data management desk follows our clients and basically inserts all of the new assets that they need; on top of this, we look at the clients' assets so that we understand what type of pay-out that asset is paying, and then we look at our available pay-outs. When we find a match, no problem; we just insert the condition and the instrument gets into the system. If there is a new pay-out that we do not cover, it is sent to a queue or to the quants. The quants essentially organise the queue based on the number of requests from clients on those functions, so the pricing functions, or new pay-outs, that have more requests from clients get higher priority in the development queue.

Falkner:

We have been talking about volatility risks. There are other types of risks with these new instruments coming in - for example, if you are investing in hedge funds, there are lock-in periods. If you have a fund of hedge funds which is locked in for a year-and-a-half, you are running some fairly big risks. I am not sure we are fully capturing in our technology the models for those to cater for the different types of risk that are involved in some of these new assets. The focus is more on investment banking and new products which are still around the pricing risks rather than some of the other risks we might be taking on.

Johnson:

That is a very important point. Traditional risk management has focused on market and volatility risk, but there are two other important aspects which I believe have been less well covered. One is liquidity risk of the kind that you are talking about. There is this feeling that it must be much more significant now than it was in the past, but it is one of those things that is very difficult to model because in extreme market events when you might want to do it, the simple ways of doing it - looking at the period to liquidate a fund, the average number of days, or whatever - go out the window. The whole situation changes, and so I think modelling liquidity risk is something that has not received enough attention.

The other area is what I would call counterparty risk as distinct from credit risk. The distinction is that credit risk is inherent in the investment that you are holding - you are getting return as a reward for taking credit risk. However, what has come in with a lot of the derivatives, especially over the counter (OTC) derivatives, is counterparty risk, which is different. Again, there are obligations under a lot of the regulations to model and manage counterparty risk, which, at the end of the day, is much more of an operational function. It is not something the fund manager typically worries about; he has bought the asset and assumes somebody else is looking after the counterparty risk.

Falkner:

I do a lot of work in the pension industry, and a particular pension fund was looking at how they could minimise their value at risk (VaR) at the bottom end. They were looking at catastrophe bonds and strategies around that, but at just one counterparty to take all that on. When that happens, that is when that counterparty will go to the courts and do everything they can not to pay that out. One counterparty is not the answer. You have to spread that around.

Sentance:

If you have one asset class for which you have many counterparties, are you able to specify different levels of counterparty risk involved in the pricing of that asset class?

Cintioli:

We do not manage counterparty risk in our software, but if I were to mention the next risk that we will have to deal with, that would be one. Our desire is to be able to measure it separately, but also to be able to integrate that type of risk inside the others, which today is possible. You have the credit default swap (CDS) curves, so all the OTC vendors have CDS curves available. You can therefore get the probability of these guys defaulting in the market price; you can even hedge that exposure.

The dream system is one that tells you, 'This is your exposure, and this is the hedge that you should do numerically if you want to remove the risk coming from this client'. Of course, when you do that, you have another OTC risk with another counterparty. At least you can measure and remove, so if you have more risk with bank X and less risk with bank Y, you can trade with bank Y a CDS on X to rebalance the situation.

That brings us on to data reliability and the transparency of the data. How important is data reliability and being able to determine, for example, risk contribution and attribution?

Sentance:

We have observed in the market a lack of transparency and understanding of data. The amount of time that people spend on manually cleansing and validating data is bad enough, even with simple products like equities. However, if you have a large data universe, it can be quite hard, particularly if you are dealing with emerging markets and trying to validate if there is any liquidity. Regulators are saying that you might have a good risk model, but they do not have any confidence in the data going in and question what is being produced at the other end. There is a great initiative in terms of automated data validation and prioritisation of what needs to be looked at, which is an audit trail of the data.

A relatively new area for us over the past couple of years has been intra-day tick data management. I never thought people would be requesting an audit trail on that kind of data because there is so much of it. People are going right through to the more complex data and the correlation, volatility services and things like that.

We have clients who want to see audit trails of that sort of data; they want automated notification to other systems so that this volatility surface is updated and other systems are aware of it. I think it is a big issue, and it is almost a bit of a wasted resource. In risk management, you tend to have people who have good abilities, but if they are cleaning up data because they are the only people who really understand what is needed, that is a wasted resource.

Julie, is this an issue within your firm?

Johnson:

Yes, I think that is right. The best value-added of the risk analysts and risk managers we employ is in deciding what we do about the output from the risk models, which is why it is very important to have more than just a headline number. You have to be able to understand the contributions to that number, because then you have the 'So what? What are we going to do about it?' If we do not like this number, how do we change it? Who do we need to be talking to? In a multi-asset class situation, what are the contributions and, importantly, where are they offsetting one another?

Sentance:

That depends on your view of risk and at what level you are aggregating.

Falkner:

We see it at a very low level. We work with people who have trouble with their calendars, so they get the interest rate curves wrong because they have interpolated an interest rate off a day when the market was closed; it is as simple as that. People still have a lot of problems around that. We then have the asset allocation-type issues where we are doing hedges for things, and someone else within the fund is doing something diametrically opposed to the overall scheme-level hedge that they have on the overall funds. If you are running a pension fund with multiple funds, someone in the fund can be running a hedge which is diametrically opposed to a scheme-level hedge across all the funds, so it works at many levels. People are trying to get all that consistency working together.

I'd like a snapshot about what the future of risk management holds. Perhaps you could also touch on regulation?

Cintioli:

Our immediate efforts are in improvements like extending the credit model to asset-backed securities and commercial mortgage-backed securities. We are also making improvements in a number of things, such as pricing and extending coverage. In terms of new risks, we are looking at counterparty risk and I would love to have a solution for liquidity risk, but to be frank. I do not see a modelling, quantitative solution that you can generalise today.

Johnson:

Personally, I like the direction in which regulation is going because we have moved from the situation where derivatives are bad - 'They are risky, so don't touch them' to 'Yes, you can use derivatives, but you've got to have a risk management process in place; you have to demonstrate that you understand what you are doing and the risks involved in that'. I think that is a positive step.

The beauty of UCITS III is that institutional funds, pension funds, and life funds can buy UCITS III funds, and there is no look-through because they are regulated funds. If a life fund wants to take on a more sophisticated investment policy that might be banned as a direct investment, they can do it in a UCITS III fund because the regulators are aware that the risk management processes have to be sophisticated.

Cintioli:

I think UCITS III is good. We are in a situation at the moment where there is a lack of confidence, and having UCITS III in place in a way gives confidence. One of the problems, in my opinion, is that these money market funds were not classified as sophisticated. They were using a complex ABS, but they were not classified as sophisticated. In terms of UCITS III, I see more direction from the regulators about what is classified as unsophisticated as opposed to sophisticated.

Johnson:

Of course, there is an incentive for fund managers to avoid classifying funds as sophisticated if they possibly can because the requirements are so much more onerous. Perhaps that is a problem that they are trying to address in Luxembourg through the new guidelines.

Going back to the evolution of the process and where it is going, I totally agree with Dario. I think counterparty risk is something that can be managed, and therefore should be the focus of a lot of the attention. Liquidity risk remains out there as an unanswered question. Even if you could measure it, what could you do about it? That is not to say that it is not a good thing to look at because at least you know you are taking the risk.

Sentance:

There will certainly be somebody who invents a liquidity derivative. That is going to be the next thing.

Cintioli:

Now we have mortality swaps. Who would have thought five or 10 years ago that we would trade mortality swaps? Maybe we will have liquidity swaps or similar products.

Johnson:

I would say the biggest challenge to be addressed is that the risk models focus not just on modelling the new instruments that are coming out, but on modelling the fund management process. We need to get not just a headline number; we need to actually delve down into it and say 'OK, how do we adapt our process to the would-be implications of that?' I want to see the whole investment process so that we can understand where the risks are coming from, not just 'We can model this or that instrument'.

Sentance:

I do not think I have a great deal to add to the comments on UCITS. Jumping a bit further into the future, I think we are heading towards real-time risk management across all of the asset classes. Perhaps that is a scary topic, but even some of the algorithmic trading vendors are talking about real-time value at risk compliance, which is going to have a load of technical problems. ><