1 / 49

Market Risk Management:

Market Risk Management:. VAR Market Risk Measurement. Analytic Risk Management Tools. Risk-weighted assets (banks) 1988 Stress Testing 1992 Value-at-Risk, VaR 1993 RiskMetrics 1994 CreditMetrics 1997 Integration of credit and market 1998- Enterprisewide RM 2000-.

avani
Download Presentation

Market Risk Management:

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Market Risk Management: VAR Market Risk Measurement Bahattin Buyuksahin, Celso Brunetti

  2. Analytic Risk Management Tools • Risk-weighted assets (banks) 1988 • Stress Testing 1992 • Value-at-Risk, VaR 1993 • RiskMetrics 1994 • CreditMetrics 1997 • Integration of credit and market 1998- • Enterprisewide RM 2000- Bahattin Buyuksahin, Celso Brunetti

  3. How much can we lose? • Everything • correct, but useless answer • How much can we lose realistically? Bahattin Buyuksahin, Celso Brunetti

  4. Definition • VaR is defined as the predicted worst-case loss at a specific confidence level (e.g. 99%) over a certain period of time Bahattin Buyuksahin, Celso Brunetti

  5. Definition (Jorion) • VaR is the worst loss over a target horizon with a given level of confidence Bahattin Buyuksahin, Celso Brunetti

  6. VaR1% 1% Profit/Loss VaR Bahattin Buyuksahin, Celso Brunetti

  7. VaR 1% Meaning of VaR • A portfolio manager has a daily VaR equal $1M at 99% confidence level • This means that there is only one chance in 100 that a daily loss bigger than $1M occurs under normal market conditions Bahattin Buyuksahin, Celso Brunetti

  8. 1% of worst cases Returns year Bahattin Buyuksahin, Celso Brunetti

  9. Main Ideas • A few well known risk factors • Historical data + economic views • Diversification effects • Testability • Easy to communicate Bahattin Buyuksahin, Celso Brunetti

  10. value scenarios sensitivity Conventional Analysis $ Risk factor Bahattin Buyuksahin, Celso Brunetti

  11. price yield VaR approach $ Risk factor Bahattin Buyuksahin, Celso Brunetti

  12. Important • VaR is a necessary, but not sufficient procedure for controlling risk • It must be supplemented by limits and controls, in addition to an independent risk-management function • Sound risk-management practices Bahattin Buyuksahin, Celso Brunetti

  13. Market Risk Management Introduction to Market Risk Bahattin Buyuksahin, Celso Brunetti

  14. Value at Risk (VaR) • In theory, risk managers should consider the entire distribution of profits and losses over the specified horizon • In practice, this distribution is summarized by one number: the worst loss at a specified confidence level, such as 99% • VAR is only one of the measures that risk managers focus on • It should be complemented by stress testing, which identifies potential losses under extreme market conditions Bahattin Buyuksahin, Celso Brunetti

  15. Market Risk & Other Risks • Three main types of risks: Market Risk, Credit Risk, Operational Risk • These three types of risks are interconnected  sometimes it is difficult to distinguish among them! • Example: Corporate Bond Bond price movements may be due to: Risk-free interest rate movements (FOMC meeting)  Market Risk Likelihood of default of the company  Credit Risk Bahattin Buyuksahin, Celso Brunetti

  16. Risk Management Tools (1) • Example: five-year inverse-floater, coupon = 16% - 2LIBOR (if positive), on a notional principal of $100M, current price of the bond = $100M  very risky bond If y goes up, the coupon rate will drop and discount rate will increase  the value of the bond will decrease sharply • How can we measure the risk of this bond? • Notional Amount: $100M  this is the potential loss and does not give any indication about a realistic loss • Worst Case Scenario: interest rate rises at 8% or above  coupon = 0 The bond becomes a zero-coupon bond whose value is $68M (at 8% rate) Loss = $100M - $68M = $32M  better than Notional Amount Bahattin Buyuksahin, Celso Brunetti

  17. Risk Management Tools (2) • Sensitivity Measure: Duration Modified Duration: 13.5 years  very high duration (as expected) Duration does not tell us whether the movement in the interest rate is likely and does not take into account of the non-linearity of the relation between interest rates and bond price • Scenario Analysis: it allows to investigate nonlinear, extreme effects but does not tell us the probabilities of these extreme events Bahattin Buyuksahin, Celso Brunetti

  18. Risk Management Tools (3) • All the above tools do not account for aggregation and therefore correlation • VaR: One number aggregates the risks across the whole portfolio, taking into account leverage and diversification, and providing a risk measure with an associated probability If the worst increase in yield at the 95% level is 1.65%, we can compute VaRas VaR = Market value × Modified duration × Worst yield increase (10.1) VAR = $100 × 13.5 × 0.0165 = $22 million The investor can now make a statement such as the worst loss at the 95% confidence level is approximately $22M Bahattin Buyuksahin, Celso Brunetti

  19. VaR: Example • Downside Risk: Maximum loss over a target horizon such that there is a low, prespecified probability that the actual loss will be greater Position: $4 B short the Yen, long the Dollar How much this position could lose over a day? Use 10-year historical data to simulate a daily return where Q0 is the current dollar value of the position For instance, for two hypothetical days S1 = 112.0 and S2 = 111.8. The simulated return is R2($) = $4,000 million × [111.8 − 112.0]/112.0 = −$7.2 M Repeating this operation over the whole sample, or 2,527 trading days, creates a time series of fictitious returns Bahattin Buyuksahin, Celso Brunetti

  20. VaR: Example • We now wish to summarize the distribution by one number Quantile • VaR: typically reported as a positive number, even if it is a loss Bahattin Buyuksahin, Celso Brunetti

  21. VaR: Example Bahattin Buyuksahin, Celso Brunetti

  22. VaR: Example • VaR is sometimes reported as deviation from the mean: VaR(c) = E(X) − Q(X, c) (2.9) • If time horizon is short, daily expected returns for financial assets are zero  the two definitions are identical • With a total of T=2,537 observations and c=95%, p=1 – c = 5% there are pT = 0.05 × 2,527 = 126 observations in the left tail. We pick from the ordered distribution the cutoff value, which is R* = $47.1 million We can now make a statement such as: The maximum loss over one day is about $47 million at the 95% confidence level  i.e. there is 5% probability that the loss will be greater than $47M Bahattin Buyuksahin, Celso Brunetti

  23. VaR: Caveats • VaR does not describe the worse loss  actual loss might be larger than VaR • VaR does not describe the losses in the left tail. VaR does not say anything about the distribution of losses in its left tail VAR fails to qualify as a subadditive risk measure, which is a very desirable properties of risk measures Subadditivity implies that the risk of a portfolio must be less than the sum of risks of the portfolio components VaR  merging portfolios may increase risk  peculiar result! Bahattin Buyuksahin, Celso Brunetti

  24. VaR: Caveats Bahattin Buyuksahin, Celso Brunetti

  25. VaR: Caveats • VAR is measured with some error The VAR number itself is subject to normal sampling variation. In our example, we used 10 years of daily data. Another sample period, or a period of different length, will lead to a different VAR number What matters is the first-order magnitude Bahattin Buyuksahin, Celso Brunetti

  26. VaR: Alternative Measures • Entire Distribution • Conditional VaR: it is an average of tail loss – i.e. computes the average of the tail  it tells us how much we can lose if we are hit beyond VaR For the Yen example, the average loss beyond VaR (= $47M) is CVaR = $74M Bahattin Buyuksahin, Celso Brunetti

  27. VaR: Alternative Measures • Standard Deviation • Advantage: takes into account all observation • Disadvantage: Symmetrical, it cannot distinguish between large losses or gains, computing VAR requires distributional assumption. • Where α is z value, σ is the volatility of the rate of return, μis the expected value and W is the amount invested. Bahattin Buyuksahin, Celso Brunetti

  28. Semi Standard Deviation • Semi-Standard Deviation • Considers only losses. • Advantage: accounts for asymmetries in the distribution Bahattin Buyuksahin, Celso Brunetti

  29. drawdown • Drawdown: decline from peak over a fixed time interval. • Of course maximum drawdown is the decline from peak to though. • If returns are not independent from period to period or if the portfolio is actively managed. • Disadvantage: It is backward looking. Time interval inconsistencies. Bahattin Buyuksahin, Celso Brunetti

  30. Cash Flow at Risk • Cash Flow at Risk (CFAT) measures the worst shortfall in cash flows due to unfavorable movements in market risk factors CF = Q × (P – C) where Q is quantity, P is unit revenue (price), C is unit cost • Example: value of a farmer’s harvest, say corn At the beginning of the year, costs are fixed and do not contribute to risk The price of corn and the size of the harvest in the fall are unknown. Price movements are driven by supply shocks  weather If there is a drought during the summer, quantities will fall and prices will increase, and conversely if there is an exceptionally abundant harvest Negative correlation between Q and P  total revenues will fluctuate less Such relationships need to be factored into the risk measurement system because they will affect the hedging program VaR 1) model the relationship between Q, P and risk factors 2) Simulation  VaR Bahattin Buyuksahin, Celso Brunetti

  31. VaR Parameters: c The higher the confidence level c, the greater the VAR measure  c: 95%, 99%, 99.9%, 99.99% Each value will produce an increasingly larger, but less likely, loss As c increases, the number of observations below VaR shrinks  it is problematic to compute quantiles accurately With 1,000 observations and c = 99.9%, the quantile has only 1 observation  VaR is computed based only on 1 observation Solution to the problem: Consistency of c across time and firm, choose c according to the objective If VaR is used for Capital Adequacy: how much capital to be set aside to avoid bankruptcy  c should be very high Backtesting: checking that the frequency of losses exceeding VaR is in line with p = 1 - c Bahattin Buyuksahin, Celso Brunetti

  32. VaR Parameters: Horizon • The longer the horizon T, the greater the VAR measure To compute VaR for longer horizons from a daily VaR, we need to assume: Rt ~ i.i.d. (,2) and () is stable – i.e. does not change for longer horizons (Normal) VaR(T days) = VaR(1 day) × √T (10.10) Key concept: VaR can be extended from a one-day horizon to T days by multiplication by the square root of time. This adjustment is valid with i.i.d. returns that have a normal distribution Bahattin Buyuksahin, Celso Brunetti

  33. VaR Parameters: Horizon • The choice of the horizon also depends on the characteristics of the portfolio: if the positions change quickly  VaR with short horizon • The choice of the horizon depends on the use of VAR: if the purpose is to provide an accurate benchmark measure of downside risk, the horizon should be short; if VaR is used for Capital Adequacy, then a long horizon is advisable • In practice, the horizon cannot be less than the frequency of reporting of profits and losses: Daily for banks, trading desks; corporations use a longer interval (ranging from daily to monthly). This interval is the minimum horizon for VaR • Backtesting: power of the statistical tests increases with the number of observations, it is advisable to have a horizon as short as possible Bahattin Buyuksahin, Celso Brunetti

  34. Basel Rules The Basel market risk charge requires VAR to be computed with the following parameters: 10 trading days horizon (or two calendar weeks) 99% confidence interval A year of observations (historical data) and updated at least once a quarter Under the internal models approach (IMA), the market risk charge (MRC) is k: supervisor determined multiplier; SRC: specific risk charge Bahattin Buyuksahin, Celso Brunetti

  35. Basel Rules: Specific Risk Rule • The specific risk charge is designed to provide a buffer against losses due to idiosyncractic factors Example: Corporate Bond by Ford Motor: credit rating BBB The market risk component should capture the effect of movements in yields for an index of BBB-rated corporate bonds SRC should capture the effect of credit downgrades for Ford SRC can be computed using VaR of portfolios that contain specific risk Bahattin Buyuksahin, Celso Brunetti

  36. Basel Rules: VaR • The Basel Committee allows the 10-day VAR to be obtained from an extrapolation of 1-day VAR figures. Thus VAR is really VARt (10, 99%) = √10 × VARt (1, 99%) The 10-day period corresponds to the time required for corrective action by bank regulators should an institution start to run into trouble The 99% confidence level corresponds to a low probability of bank failure due to market risk Even so, one occurrence every 100 periods implies a high frequency of failure  there are 52/2 = 26 two-week periods in one year  one failure should be expected to happen every 100/26 = 3.8 years, which is still much too frequent This explains why the Basel Committee has applied a multiplier factor, k ≥ 3, to guarantee further safety Bahattin Buyuksahin, Celso Brunetti

  37. VaR: 3 steps • From market data, construct the distribution of risk factors (e.g., normal, empirical, or other) • Collect the portfolio positions and map them to the risk factors • Based on a VaR method (normal, historical, Monte Carlo), construct the portfolio VaR Bahattin Buyuksahin, Celso Brunetti

  38. VaR: 3 stepsStep 1 The risk factors represent a subset of all market variables that adequately represent the risks of the current portfolio Simplification: There are literally tens of thousands of securities available, but a much more restricted set of useful risk factors choose market factors that are adequate for the portfolio Example 1: fixed-income portfolio  one bond market risk factor may be enough to represent the risk of the whole portfolio Example 2: option portfolio  volatilities should be added as risk factors In general, the more complex the strategies, the greater the number of risk factors that should be used Bahattin Buyuksahin, Celso Brunetti

  39. VaR: 3 stepsStep 2 Portfolio positions – assumption: all positions are constant over the horizon (this is a simplification) The true risk can be greater or lower than the VaR measure It can be greater if VaR is based on close-to-close positions if traders take more risks during the day, the true risk will be greater than indicated by VaR Conversely, the true risk can be lower if management enforces loss limits limits the risk that traders can take if losses develop Bahattin Buyuksahin, Celso Brunetti

  40. VaR: 3 stepsStep 3 The choice of the VaR method depends on the nature of the portfolio Example 1: for a fixed-income portfolio a linear method may be enough Example 2: portfolio containing options we need to include nonlinear effects Risk management is as much an art as a science  Risk managers need to make reasonable approximations to come up with a cost-efficient measure of risk Traders could be induced to find “holes” in the risk management system A VaR system alone will not provide effective protection against market risk It needs to be used in combination with limits on notionals and on exposures and, in addition, should be supplemented by stress tests Bahattin Buyuksahin, Celso Brunetti

  41. FRM-99, Question 89 • What is the correct interpretation of a $3 overnight VaR figure with 99% confidence level? • A. expect to lose at most $3 in 1 out of next 100 days • B.expect to lose at least $3 in 95 out of next 100 days • C. expect to lose at least $3 in 1 out of next 100 days • D. expect to lose at most $6 in 2 out of next 100 days Bahattin Buyuksahin, Celso Brunetti

  42. Stress Testing VaR does not account for extreme losses  stress testing: identifies situations that could create extraordinary losses  Identify areas of potential vulnerability The goal of stress testing is to identify areas of potential vulnerability Stress testing is a key risk management process that includes (1) scenario analysis; (2) stressing models, volatilities, and correlations; and (3) policy response development Scenario Analysis: Moving key variables one at a time; using historical scenarios; Creating prospective scenarios Event Risk: Changes in government; changes in economic policies; coups, civil wars, invasions; currency devaluations Bahattin Buyuksahin, Celso Brunetti

  43. Example: Turmoil in Argentina • Up to 2001, the Argentine peso was fixed to the U.S. dollar at a one-to-one • The Argentinean Government had promised to defend the currency at all costs • Argentina Experienced the worst economic crisis in decades  excessive borrowing • In December 2001, Argentina announced it would stop paying interest on its $135 billion foreign debt  This was the largest sovereign default recorded so far • January 6 devaluation: the exchange rate promptly moved from 1 peso to more than 3 pesos per dollar • Such moves could have been factored into risk management systems by scenario analysis • What was totally unexpected, however, was the government’s announcement that it would treat bank loans and deposits differentially  Dollar-denominated bank deposits were converted to devalued pesos, but dollar-denominated bank loans were converted to pesos at a one-to-one rate • Banking system technically insolvent  loans (bank assets) overnight became less valuable than deposits (bank liabilities) Nobody predicted this move • In 2005, the Argentinean government proposed to pay back about 30% of the face value of its debt. This recovery rate was very low by historical standards Bahattin Buyuksahin, Celso Brunetti

  44. Liquidity Risk • Component of market risk • Liquidity risk consists of both asset liquidity risk and funding liquidity risk • Asset liquidity risk, also called market/product liquidity risk, arises when transactions cannot be conducted at quoted market prices due to the size of the required trade relative to normal trading lots • Funding liquidity risk, also called cash-flow risk, arises when the institution cannot meet payment obligations Bahattin Buyuksahin, Celso Brunetti

  45. Asset liquidity risk • Example 1: A bank has a large position in IBM shares The bank would like to sell the position - IBM close price $79.36 - The market is illiquid At what price would the bank be able to sell the large IBM position? • Example 2: A bank has a large portfolio The bank would like to assess the risk of the portfolio The market is illiquid How should the bank compute volatilities and correlations? Bahattin Buyuksahin, Celso Brunetti

  46. Asset liquidity risk • Aggregate versus stock specific liquidity Aggregate liquidity is market-wide, fluctuates over time, and has different impacts on different stocks (October 1987, August 1998 – LTCM) Stock specific liquidity is a property of a stock • Definition: Liquidity is the ability to trade quickly any amount at the market price with no additional cost Bahattin Buyuksahin, Celso Brunetti

  47. Asset liquidity risk: Stylized facts • Expected returns increase with illiquidity (liquidity premium): - Amihud (2002), Amihud & Mendelson (1986) • Volatility increases with illiquidity: Pastor & Stambaugh (2003), Papanicolaou & Sircar (1998), Schoenbucher & Wilmott (2000) • Correlation increases with illiquidity: Bhansali & Wise (2001) • Sharpe Ratios increase with illiquidity: Lo, Petrov & Wierbicki (2003) Bahattin Buyuksahin, Celso Brunetti

  48. Liquid Assets • Tightness, which is a measure of the divergence between actual transaction prices and quoted mid-market prices • Depth, which is a measure of the volume of trades that can be made without affecting prices too much (e.g., the bid/offer prices) and is in contrast to thinness • Resiliency, which is a measure of the speed at which price fluctuations from trades are dissipated • In contrast, illiquid markets are those where transactions can quickly affect prices Bahattin Buyuksahin, Celso Brunetti

  49. Liquidity Risk • When the market is illiquid  flight to quality: there is a shift in demand away from low-grade securities toward high-grade securities • Flight to quality may even affect government bonds: The yield spread can widen between off-the-run securities and corresponding on-the-run securities On the-run securities are those issued most recently and hence are more active and liquid. Other securities are called off-the-run The latest-issued 30-year U.S. Treasury bond This bond is called on-the-run until another 30-year bond is issued, at which time it becomes off-the-run Because these securities are very similar in terms of market and credit risk, the yield spread is a measure of the liquidity premium Bahattin Buyuksahin, Celso Brunetti

More Related