Part 3 – Weather forecasting, black swans and the capital markets crisis
In the third installment of a series of 3 papers on central counterparties, bilateral margin agreements, OTC derivatives, weather forecasting, and a global risk research model, Bill Hodgson explains what we can learn from the weather system about global co-operation when it comes to managing financial risks.
The capital markets could learn from weather forecasting organisations around the world in a number of ways, especially in these times of extreme financial weather.
Many commentators have described the current financial crisis as a systemic problem, meaning not a failure caused by a single component, but a result of interactions within and between the entire global financial system.
With a systemic problem, how do we research the causes and cures of the current situation, and make informed decisions on how to manage the global financial system for the future?
By comparing current risk management practice with global weather forecasting we can take a look at how a complex system (the global atmosphere) is modelled, researched and predicted, and draw parallels with another complex system, the global economy and financial system.
The Big Picture
Great Britain is renowned for unpredictable weather, making the Queen’s subjects far more interested in weather predictions than some other countries who enjoy consistent heat in the summer and cold in the winter. Great Britain is a relatively small patch on the surface of the earth, so you might assume the computer systems required are correspondingly smaller than a larger country. Wrong! The UK Met Office uses super computers amongst the top 20 in the world, and the reason why? You can’t build a model of a complex system like the weather in Great Britain without modelling the entire global atmosphere so that the model is complete, without any gaps or “edges”. The mathematical models used to predict the weather use a three dimensional grid where each point is affected by the weather at the points around it. The model would fail at the edges of Great Britain if the grid of points abruptly stops somewhere in the Atlantic, meaning data would be missing for the weather formulae at the edges of the model.
In comparison the banks (or hedge funds, investment managers and so on) carry out financial predictions over their own “patch” of the markets, meaning their own trading activity. A bank needs to utilise a risk model of all trading activity within their firm, cutting across all types of trading activity, and including all customers. This risk model is a tool to enable them to create predictions about the financial conditions within their firm.
A crucial difference between the weather model and the risk model, is that the risk model must stop at the ‘edges’ of the bank, which excludes trading activity outside the bank’s boundaries and the trades the bank is not party to. This means the risk model cannot answer questions about the effects of changes in financial circumstances between multiple financial organisations, or between the bank and the wider economy, as this isn’t possible or necessary for a firm to do under the current regulatory environment.
Observation
Weather models need feeding with data from as many real world sources as possible. Weather forecasters use satellites, balloons, ships, aeroplanes and ground installations to capture data on weather conditions. The UK Met Office has the responsibility to collect this data across Great Britain, and in return exchanges data with similar organisations to achieve wide data coverage across the globe.
A bank will gather trade data from the many individual trading systems in all the markets in which they are active, so for instance, bond and equity trades, futures at exchanges, OTC derivatives contracts, loans, foreign exchange etc. The bank will also have internal controls and data quality measures to ensure that all trading activity is captured and delivered to the central risk system, and that the way the trades are delivered is consistent across systems and markets.
Figure 1 – Risk management within a single firm
Assimilation
The globe is a big place, and not everywhere has a convenient weather station to feed the weather models, for instance the middle of the Atlantic ocean and the tops of high mountains lead to gaps in data coverage. Weather experts employ numerical methods to fill these gaps so as to have a more consistent view on current weather conditions, before beginning the prediction phase.
A bank must gather market data which describe the current financial conditions affecting the firm. The data required will include yield curves, volatility surfaces, skews, exchange rates, bond prices, equity prices and so on. Risk managers also employ methods to fill gaps such as interpolating yield curve points and smoothing volatility surfaces, providing a consistent layer of data with which to model the financial circumstances within their firm.
Modelling & Prediction
Weather forecasters then apply mathematics developed over many years to predict the weather conditions across the globe, using the current state of the weather as a starting point, and applying the dynamics of weather movements to arrive at predictions over the coming week, which we then see on the nightly news.
Risk modellers combine data on current trades with market scenarios to arrive at a series of predictions on how their portfolio will behave under many market circumstances. It is this stage which makes use of Value at Risk (VaR) models to make statistical analyses of the outcome of each of these market scenarios. A bank will make many predictions about their portfolio, to explore the limits of their profits and losses in many imaginary market situations.
An interesting piece of forensic research would be to examine the market scenarios used by the major banks and see how they modelled the increase in mortgage defaults which have driven the current collapse of asset values.
What does this teach us about risk management & the current crisis?
Weather forecasters share data, methods and techniques to achieve a better global result, as there is no competitive disadvantage to doing so. On the other hand banks guard their risk management techniques with good reason – the better you can anticipate market changes, the better you can position your firm for profit.
Neither regulators nor ratings agencies have similar tools available to replicate the approach taken by banks, meaning there is limited independent validation of the risk modelling techniques used within the firms, nor the opportunity to experiment with market scenarios that aren’t in the interests of firms who carry out trading.
The epicentre of the current financial crisis involves the retail mortgage market, and it’s linkage into the core capital markets via structured OTC derivatives. The systemic effect of a global economic change leading to increase in mortgage default rates was a catalyst to disrupt an unsustainable business in these illiquid products.
Firms had allowed themselves to move from products with a healthy secondary market into a game of musical chairs with assets that couldn’t be off-loaded in a time of crisis. Only when firms had difficulty placing new issues into the market did they begin to question the quality of those assets, relying on the pipeline of deals to generate profits.
The world puts it’s trust in national regulators and commercial ratings agencies and assumes financial disasters are therefore unlikely to happen, or at least limited in scale. The current crisis has shown how ineffective this approach has become, the world financial system is now too complex for a single national body to understand, model or control.
Systemic research?
If we take the weather modelling analogy and think again about global risk management, is it necessary to conceive of a global risk management system into which we place all the world’s capital markets trade activity and market parameters, so that researchers can begin to try and understand the capital markets as a complete system? We don’t model the weather by picking out the data state by state or country by country, we model the entire globe and then pick out the results from the region of interest. How would we go about creating such a system?
Scope: The data observations need to include enough financial organisations to ensure the majority of the world financial risks are included, some analysis would be needed to define a criteria for inclusion, but it would be fair to assume that the top 100 banks, and top 100 non-bank financial institutions (such as hedge funds and investment managers) would be a proxy for the wider market. It would also be necessary to include the activity of the worlds central banks and their programme of debt issuance, given they are the foundation on which the world capital market is built.
Data gathering: Given that all major firms in the capital markets have already invested in risk management technology, you would assume that there are pools of data within each firm, which if merged would give a wide view of trading activity in the market. Firms will already have done the work to normalise the data making contribution to a global risk system easier than building such a thing from scratch.
Figure 2 – Collecting data already present within firms risk systems
Market scenario modelling: Lets assume that this global risk system is a research tool, and is intended to go beyond the interest of the markets own risk management efforts. The market scenarios to be modelled need to include a wide range of imaginable stress scenarios, plus a quantity of unimaginable “Black Swan” scenarios. For those that haven’t come across Nassim Nicholas Taleb, he is the author of “The Black Swan: The Impact of the Highly Improbable”. The book explores the idea that we humans are poor at understanding randomness, the occurrence of rare events, and the enormous consequences of these occurring. Of course including market scenarios in this global model, which no-one can imagine but are low probability is like expecting the unexpected!
Ownership: If such a system existed, the power that would give the operator would potentially enable the prediction of the success and failure of major firms in the capital markets. Such information could not possibly be allowed to reach the public domain in a direct sense, for fear it would bankrupt any firm subject to such disclosure. This system would need to be an independently owned and discreet organisation, producing results and analysis that could be used by governments and regulators to manage the global financial system for years to come. A nomination for the owner of this system would be the oldest international financial system, the Bank for International Settlements in Basle. Their mandate as “a centre for economic and monetary research” makes them a candidate and aligns them with the need for this research tool. Joseph Stiglitz in a speech at Oxford University in the UK suggested the only organisation suited to this task would be the United Nations.
Cost: A really big weather forecasting computer costs about $200 million, one of these is probably enough to run such a system, given this project isn’t safety critical. Building the system including all the data feeds might cost $5m per firm with 200 firms involved giving a budget of roughly $1bn, plus annual operating costs. The current estimate of the amount of money spent to bail out the global financial system so far seems to be around the $1 trillion to $5 trillion mark in the US alone therefore funding shouldn’t be an issue, especially if global governments and the firms themselves share the cost. The world has spent $3.8bn on the Large Hadron Collider to explore the extremely small, so a $1bn+ investment on exploring the extremely large global financial system seems good value given the possibility it will allow us to look further ahead and begin to understand the complexity in the financial system we now have.
Complex products: Having an independent body carrying out risk management research would address the point above about complex products being difficult to price or include in a CCP environment. Adding such a product into this global system would itself be a test of the reasonableness of trading such a thing.
The way forward?
Many commentators put risk management at the heart of the current crisis, either blaming it for missing the obvious, or being fundamentally flawed. Questions have been raised about the purpose of ratings agencies and whether their insight into complex products is the independent reference point required. Unfortunately there isn’t any other silver bullet to help us understand how the financial system functions, and given the silo approach of regulators, CCPs and specialised exchange traded markets we don’t have much chance to replicate the current disaster in laboratory conditions to begin to find a new way to understand the world financial system.
The resources required to create such a financial laboratory are easily within reach, and the human assets to make use of such a tool walking the streets right now. If we agree that the world financial system is as complex as weather (or even more so), and as fundamental as energy and food, who would argue not to begin a project as important as preserving the financial stability of the human race?
Follow up points:
Adsatis can provide advice on trading strategies, structured products and risk management covering topics such as:
• Distressed portfolio management
• Fixed income & OTC derivative products
• Risk management and stress testing
• Central Counter Parties
• Exposure Management
Post scripts
The following article discusses the idea of an attempt to model the world economic system:
http://www.edge.org/3rd_culture/brown08/brown08_index.html
This article by Nassim Taleb explains why the random events which cause such a crisis cannot be successfully modelled:
http://www.edge.org/3rd_culture/taleb08/taleb08_index.html
Contact Bill Hodgson
Part 2 – Comparing the ISDA bilateral exposure management model with a CCP
Part 1 – A central counterparty for OTC credit derivatives – are we over estimating the importance?
