Risk management: the long search for real time

Copying and distributing are prohibited without permission of the publisher

06 October 2011

Real time is seen by many as the holy grail of risk management. Mythical perhaps but sought after by all for its indescribable value. With regulators putting increasing demands on banks to provide real time data and with the cost of collateral rising, the pressure is on to understand risk in a more timely manner. But what does real time mean and what are the barriers to achieving it? William Mitting found out.

On September 15 Kweku Adoboli, a 31 year old Delta One trader at UBS broke the news to his bosses that he had lost billions of dollars through unauthorised trading of ETFs. The news would send shock waves throughout the financial system and plunge a bank that was just emerging from the crisis back into turmoil.

What is perhaps most shocking about the loss is the fact that Adoboli not only apparently circumvented internal individual trader position limits, he also managed to build up a position that was not flagged for exposing the banks to such an enormous loss, allegedly due to fictitious hedges.

The incident highlights the challenges banks face in monitoring the risks taken by their trading operations. At the time of writing, details of the rogue trades were still unclear but in a world where billions of dollars can be made or lost in the blink of an eye, real time risk management is no longer a luxury but now a necessity. It is a challenge that is forcing an overhaul of how banks organise their risk systems and process data.

The root of risk

It is perhaps no coincidence that the term “risk” entered modern lexicon in Europe during the 16th and 17th century at the dawn of modern capitalism. Studies in etymology suggest that it entered the English language from Spanish or Portuguese and was used most commonly in conjunction with sailing into unchartered territories. Its origins are from Greek, where it meant “root, stone or cut of the firm land” but was soon used to refer to something that was hard to avoid at sea.

As capitalism evolved, risk no longer became something to avoid. It became something to tame, and to leverage. No longer were people set on avoiding the roots; in the modern world, we sail as close to them as possible for it is by the cut of the firm land that the best rewards are to be found. However, for all our advances, the financial crisis that has raged since 2007 is evidence that confidence in the ability to measure and manage risk was misplaced.

Real time risk management has been talked about for years and there are many area of a bank’s business where it is a reality. However, there is increasingly demand for a holistic view of total exposure to a specific risk across a bank’s total portfolio in as near to real time as possible. The complexity of such a demand means that real time does not mean instantaneous but no longer the world of T+1 that many banks operate in today.

Simon Greig, a vice president at Sapient Global Markets who runs the derivatives global platform group, says: “Real time risk management is not a new initiative but with the new regulations coming out of the US and Europe there is a push to move to more real time risk management around exotic and hard to price instruments.

“Real time in terms of complex derivatives means same day. It used to be that the risk manager would not see anything until T+1 once all the batch processing had been completed. Today, there is much more capability from applications and a greater ability of the risk manager to view the data in real time.”

Why real time?

There are three principle reasons why banks want to implement real time risk capability: minimising losses, regulation and capital efficiency. New regulation coming into force across Europe and the US will increase the burden on banks and exchanges to implement real time risk checks. At the same time, the capital is becoming more expensive due to mandated margin requirements and central clearing and so capital efficiency is being scrutinised like never before.

Mat Newman, vice president of product management at Sungard, says: “The better you can measure your risks, the more accurately you can control them and the more up to date you can be, the more efficiently you can operate within the risk appetite of your firm. If I am only able to approximate the risks I am running, the chances are that I am going to put a conservative approximation on that risk.”

While Dodd-Frank will require real time reporting in swaps from this November, the Mifir draft that was leaked this month appears to go even further with reporting required across asset classes and at trader level, rather than just member level meaning that the onus on banks is even greater.

Bob Fuller, a director at Fixnetix, says: “Up until recently, real time risk management has not been a mandatory requirement. If the regulation is passed as currently written then this will apply to all instrument types and across all venues including the prop book. Obviously, there will be a substantial increase in complexity for trading firms.

“Both the participants and the organised trading facilities, MTFs and regulated markets now have more automation to do and it has to be something that doesn’t slow down the market. You are in a situation where real time risk has to be done at the execution stage and in a short period of time.”

Slow move to real time

A recent report by Lepus found that almost all banks surveyed were taking steps towards implementing real time risk management. In addition, the study found that most banks have taken steps to overhaul their risk management systems since 2007.

Individual traders and bank divisions have had access for some time to real time risk management; the challenge for the banks is to gain a view across asset classes, which have traditionally been run in silos.

Michael Coleman, the founder and chief executive of Hyper Rig, says: “All the top tier banks are looking at real time, intraday risk solutions but face a significant technical problem. A lot of the current processes are built around technologies that are 10-15 years old or more. That technology is fragile, unstable and is very difficult to maintain. It is an enormous challenge to try and vamp that up into something that is more up to date.

“All the data is in different silos with different systems, the databases have different codifications. An exposure could be represented as Barcap in one silo, Bcap in another and Barclays in a third one. Simple problems like that can be very difficult to iron out when you try to consolidate information. “

Fuller agrees: “Most of the complexity is about having access to all information in real time from disparate systems in different formats all in a pre and post trade environment as well as putting a front end on that is easily understandable and controllable for risk managers/traders.”

Complex derivative and high frequency trading have added additional complexities. In today’s world of high speed electronic trading, traders can be in and out of positions within seconds building up risk and closing it down in the blink of an eye. Meanwhile, complex derivatives are hard to price and the risk they pose is often hard to compute.

Greig says: “When you talk about HFT it tends to be fairly vanilla and so the availability of data to price those instruments is the issue, it is about the timeliness and the ability to handle the millions of transactions on a daily basis. HFT has forced vendors to optimise the application capabilities.”

Achieving real time

These additional complexities and the need for real time risk evaluation have forced banks to rethink how they process risk. Newman says that there are two key elements to real time risk management: market risk and credit risk. Market risk can be hedged whereas counterparty credit risk is harder to hedge against and can’t be rapidly unwound.

“Counterparty credit risk is getting a lot of focus at the moment as to do that calculation properly, you need to have a simulation approach. Potential Future Exposure, CVA for cost of credit and also the regulatory exposure, concentration limits and cost of collateral need to be calculated,” he says.

“On the market risk side, the need for real time varies. If you are on a trading desk then you want Greeks minute by minute whereas if you are at the enterprise level most banks are running VAR calculations on a daily basis.

“What they want to do is answer more ad hoc queries rather than just a set of pre-canned reports that don’t change for three months. They want more flexibility and the ability to analyse the underlying drivers of risk,” he says.

The need to dip into the data pool and establish ad hoc risk positions has led to a sea change in how risk management systems are designed. Once banks have streamlined their operations and broken down the traditional silos, the problem then becomes how to process that data in as close to real time as possible.

Traditionally, banks have relied upon lengthy linear batch processing methods. However, recent advances in risk management have led to the implementation of extensive Monte Carlo simulations and stress testing, where a vast array of possible market movements are tested and the impact of each recorded to identify vulnerabilities.

Even taking into account the significant developments in software and processing capabilities, Monte Carlo simulations take time. They have enabled banks to get the time taken down from overnight to hours or even minutes but it is still hard to achieve the interaction required and get the calculation down to one or two seconds.

Newman says that problem has been answered with incremental Monte Carlo. Using this approach, every time you calculate the PFE or CVA for a counterparty rather than re-evaluating the entire portfolio of trades, you just calculate the incremental effect of the new trade.

“Basically, you store the valuations for the other trades away and add the new deal, integrating that into the results to get the one or two second response time but get the full effect of the Monte Carlo simulation,” he says.

Hyper Rig’s Coleman adds: “When you run Monte Carlo simulations, you are running a number of different paths. If you need to re-run the simulation a few seconds later, most of the paths will still be valid. So when you do it incrementally you are adding a number of much smaller calculations.”

Challenges remain

Even with the advances in incremental Monte Carlo simulations and the increased processing power, there are still some hurdles. “Data quality is always going to be a limiting factor,” says Newman.

“You can throw money at the technology problem but data quality is an issue that a lot of banks struggle with. You get to the point when you have this system that is producing these numbers and you assume that everything is correct but the reality is that things get booked in the wrong place or some market data is missing.

“To be in a position to achieve real time risk, to get the information when you need it, you need a system that is not just churning stuff out all the time, but can give you the information you need at that particular time and allow you to understand how much reliance can be placed on that number.”

The massive loss at UBS last month is a sign that some banks’ internal risk checks are not up to scratch. However, it is one thing stopping traders from exceeding their trading limits and another understanding the complete risk profile in real time. If banks still struggle with the former, it may be some time yet until risk is truly tamed.

Highlights from the October issue of FOW:

News

News analysis: Data gap remains in commodity speculation row