Defining and Setting Limit Structures By Dmytro Dutka, MBA, CFA (Submitted to Canadian Treasurer Magazine) Financial risk measurement has advanced significantly in recent years, and though it has complicated Treasury Management the pay off has been smarter risk taking. The pressure is on to identify, measure and to manage the multitude of Treasury-related risks. Some are “new” while others are just separated out or better appreciated in the light of new valuation models. Still, each addition requires that policies, practices, limits and limit structures be adapted. But how does a Treasurer coordinate their risk view into a relevant, logical, coherent and practical system of limits? In the academic hurry to publish and developers’ rush to popularize the most current risk valuation models into software, Treasurers have been left alone with the task of adapting limit structures. “New” statistical risk measures like liquidity risk are springing up. Non-market risks are being added. The latest is operational risk and there is talk of modeling “reputational” risk. Driving this are, a) Unprecedented “write-offs” due to unexpected FX “translation” adjustments, funding rate “movements”, and “liquidity” crises; improper hedges or even ‘three-legged stool’ off-balance sheet financing programs, and “rogue” traders b) Product innovations involving contingency assets and liabilities, c) Detailed “business unit” level profit measures (cost allocation & transfer pricing regimes for risk management) d) Valuation methods like RAROC (Risk Adjusted Return on Capital,) and e) Changing capital adequacy regimes. Singular, exclusive risk limits on, for example “liquidity”, do not logically follow from a new risk measurement. We can continue though, to limit activities or exposures that directly or indirectly affect the new risk measurements. Without a logical blueprint however, limits have been added on without consideration to the overall effectiveness of the structure. The many risks are indeed a function of a Treasurer’s role: funding, cash, interest rate, Foreign Exchange (FX) and liquidity management and sometimes financing, capital and balance sheet management. Limits act as guiding parameters and reflect the faith and confidence Treasury is given in managing certain risks. A proper limit structure defining process needs to assess the relevance and relative importance of each risk as well as Treasury’s freedom and flexibility in their management. It should be kept current via a periodic, systematic and thorough review of the business, strategies, policies, procedures and practices vis-à-vis the business environment. Then risk limits, a limited resource like capital, are assigned out on the basis of risk appetite and ability to handle exposures. Initial limit structures were simple. As the types of Treasury risks recognized multiplied, the list of limits has gotten longer. Each industry sector and even each business has its own risk nuances. But the limit structure has generally remained flat, wide and simple (Figure 1). Figure 1 Figure 1 Treasury’s Separate Limits FX Limit Funding Limit © FinLev Management Ltd. December 2002 Aged Receivables Limit Investment Product Limits ALM Gap Limits 1 Simply adding on a different limit for each type of risk involved has led to illogical and uncoordinated limit systems. There are still instances where Treasury groups follow static notional position limits (i.e. $2 million USD/CAD) even though the potential $ amount risk varies with obviously trending markets or greater volatilities. VaR is a broadly accepted “holistic” effort to bring together various market exposures into one Value at Risk number. Most Treasury managers now speak this “new math”: a statistical language of standard deviations, correlation, variance and various Greek letters. In Figure 1 VaR summarizes the FX and ALM Gap limits. Since the deficiencies of VaR were highlighted in the 1998 Long Term Capital Management debacle, Stress Testing has been added alongside individual market type or VaR limits. Stress Testing, a VaR engine outcome of potential (but highly unlikely) extreme scenarios, can have bankruptcy consequences. A separate limit for this was created. Jeremy Berkowitz (Federal Reserve Board March 20, 1999) in a paper “A Coherent Framework for Stress-Testing” suggested folding Stress Tests into the risk model (VaR), by assigning them probabilities. He argued this provided a “usable risk metric” -- a singular risk measure. The modified VaR with redefined extreme events does have a more realistic “tail”. An even more realistic result could be had by boosting the probability of the extreme events closer to that of historic occurrences. Such aggregation, however, ignores the usually different bases of time horizons between VaR and Stress Testing and it negates the original purpose of the latter. Stress Tests could separately supply limits to avoid reaching catastrophic exposure to perhaps 10-year horizon events that come with little forewarning. Trading VaR limits on the other hand can be based on 1 day horizons (the time it takes to off-load a futures hedge). Even for similar horizons, one should seriously question whether Treasury should be allowed to substitute credit or concentration risk for market exposures. Most firms that do both VaR and Stress Testing do not “roll” the stress test results into VaR numbers. The two limits reflect different risks…one the kind of losses that could be sustained but should be controlled, while the latter relates to disastrous exposure. They should be used separately. A Stress Testing limit logically should have priority over related VaR limits. Other attempts to incorporate Liquidity and Credit risks into VaR (Expected Extremity Associated Risk, EXAR, Rik Sen & N.A. Hariharan 2002) have not resulted in any generally accepted “best practices” models or limits either. Just as a simple, singular, exclusive VaR limit does not follow from an aggregated valuation, a singular Stress Test limit would also be incorrect. It too would allow Treasury too much freedom over management of specific market & concentration exposures. Therefore separate, ancillary limits for market, credit and liquidity risks need to be employed. Here too, each can have their own sub limits, i.e. controlling for types of market risks depending on appetite or expertise. Detail and complexity will be limited by system and personnel competencies. Figure 2 shows an example of a coordinated limit structure. © FinLev Management Ltd. December 2002 2 Figure 2 Coordinated Limit Structure VaR Limit Stress Testing Limits Concentration Limits Regulatory, Legislative Limits Stop Loss Limits Liquidity Limit FX Limit Funding Limit ALM DV01 Limits Credit Limit Aged Receivables Limit Operational Risk System integration Investment Product Limits Staff Training & Education Not everyone has the need or desire to take on the complexity of a correlated VaR. However, simplistic limit structures should be updated to force relevance, i.e. by allowing addition (though not necessarily netting) across risks. Simple FX limits can be expressed in terms of: a) notional limits (i.e. spot exposure of $2 million USD/CAD) and b) DaR (i.e. $120,000 Dollars at Risk backed by the application of volatility figures) Apart from VaR one can choose risk measures such as Net Interest Income for accrual accounting management styles or Shortfall exposure for those focused on targeted earnings. Each has its use and potential ancillary limits, depending on the type of strategy chosen to run a firm, or the way it is valued. Unfortunately there are few formal papers written on the subject and no available benchmarking studies. Indiscriminately layered “legacy” limits, left or created without regard to a firm’s changed environment, goals, risk appetite or a Treasury group’s level of competence will result in poor management. Systematic attempts should be made to create, to prioritize and to coordinate the limits into a logical limit structure. Otherwise some limits will confuse and frustrate effective management. For example maximum DV01 limits and gap limits may seriously conflict and together may seriously overshoot Stop-loss limits making the former irrelevant, providing poor guidance. Stress Test limits should not ignore the VaR limit. Similarly, Liquidity limits should be set relative to market tolerance, credit lines in place and maximum cash use implied by maximum use of VaR limits. In short, the various limits set should be logical, coherent and coordinated. This means reviewing the various possible maximum exposures as defined by chosen scenarios (Historic, Monte Carlo or forward looking), and looking at the matrix to correct the illogical limits. No current software application attempts to automate a coordinated & customized limit structure design. © FinLev Management Ltd. December 2002 3 Limit structures cannot be standardized across disparate non -commoditized businesses or business sectors. Risk perception, too, is subjective. Where some perceive risk, others through knowledge, analysis or setting aside mitigants see little risk and therefore require different limit structures. Figure 3 Valuation or Earnings Focus Management Style (dynamic/ reactive) IT/Systems Competency (complexity) Legal & Regulatory Limit Structure Available & Validated Risk Models Type of Business (Types & No. Of Risks) Personnel Competency (complexity) Risk Appetite A limit structure redesign effort will force a healthy focus on core strategies, principles & competencies and will clarify muddy delegated risk management guidelines. Figure 3 shows some of the things to consider in creating a Limit Structure. A sample of what the redesign should incorporate includes: 1. Definitions of the types of risks Treasury faces. A list of specific, controllable related limits and sub-limits. Limits need to be justified via analysis of the appropriateness, relevance, measurement and practicality of each vis-à-vis its related risk. They should address foremost regulatory, statutory and company (Board-endorsed) policies and business strategies. GAAP disclosures may eventually require statements regarding risk management limits as these too are potential liabilities. 2. Risk models used, their assumptions, limitations and parameters need to be made explicit a) i.e. For VaR, a 1 week horizon, using Historical simulation, 5 year look back scenario, with a 95% confidence interval and correlation matrixes validated by a given firm. b) Stop Loss limits should be cumulative over time and across products/portfolios, and be related to risk appetite (itself related to management’s desired risk/return profile). It may be a notional amount, a percentage of actual or expected earnings, etc. c) For Stress Testing the scenarios need to be defined Historic, Monte Carlo or forward looking. There is no “best practices” here either. Methods depend on desired simplicity, available computer power and confidence in ability to judging the future. 3. Limits should be prioritized, with the higher level limits clearly relating to the Board’s and Executives’ interpretation of risk. Limits such as VaR, a Stop-loss, and a Stress Test are immediately meaningful to them. Other limits (such as Gap limits or DV01’s) can be sub limits of © FinLev Management Ltd. December 2002 4 4. 5. these, aiding in Treasury’s management of the higher level limits. One can get as detailed as one wishes, however the “KISS” principle (Keep It Simple Stup&#) should be kept in mind. Limit levels should be set relative to a clear risk/return tradeoff decision. Though non-financial firms with a conservative bent may wish to eliminate risk, some leeway must be given to allow flexibility in Treasury’s duties. A formal, documented process is needed to back up each of the below: a) The setting of the size of each limit. Limits can be set relative to capital, earnings, or some other measure. For VaR, one estimate is that “a one day VaR equal to about 3% of the trading capital is a pretty good sized risk in a normal environment.” (Barry Schachter, Head of Enterprise Risk Management at Caxton Corporation). Sub-limits can then be chosen as appropriate. Under VaR one can have separate FX DaR and Fixed Income DV01’s and/or product specific VaR’s. Derivatives by their nature should receive special attention. b) How far back from disaster Stress Test limits are set depends on how probable are the defined extreme event(s), and lead times (if any). c) Once limits are set procedures should provide for action when limits be breached: Obtaining limit excess approvals from Executive Management (themselves to be granted limits or authority by the Board of Directors) and, Putting into motion self-correcting specific actions that reduce, mitigate or eliminate the offending risk exposure. d) Limits can be set in stages of concern or severity. i.e. A lower funding or credit limit breach will put into motion certain remedial actions, while a higher limit breach will require more severe actions. e) Dynamic structures can be created, where limits vary with profits/earnings or changing market conditions. A limit structure’s success requires, a) Mathematical validation. Options are backtesting or Monte Carlo simulations. Will the model make sense and provide the P/L results expected? b) It to be coordinated. 6(a) above should create a matrix of results under different maximum exposure scenarios that indicate unreasonable or superfluous limits for correction. c) Validation of a given level of competency in managers for a given level of complexity. d) A system of periodic, preferably independent and automated monitoring and controls to be implemented. © FinLev Management Ltd. December 2002 5 Process of a Coordinated Limit Structure Plan y of t or ew gula vi re t , re nt dic k e e rio ar nm Pe , m iro 5) es s env n si Bu t& Se e 4) lidat nal Va lat io & e R it s s Lim s t em Sy 1) R ev St ra iew Bu s in t egy & P es s lan 2) Def ine Treasury Functions & Risks Manage d nd p a it s ou Lim Gr 3) rit ize Prio A final word on change management: Incrementalism. Introduce change in simple, tested steps, over time with the buy-in of the Board, Executives and Treasury Management, Risk Management and IT groups. A wholesale one time change that modifies policies, practices, procedures including system IT issues begs a meltdown. Managing with a coordinated, cohesive limit structure will increase manifold the likelihood maintaining a desired risk profile and attaining a desired valuation. Dmytro Dutka is President of FinLev Management Ltd., a Financial and Treasury consulting firm located in Toronto, Canada. He can be reached at (416) 245-8579 or by email: FinLev@rogers.com See www.FinLev.com © FinLev Management Ltd. December 2002 6