2847 WWW.GFEDUNET PART TI 41, FOUNDATIONS OF RISK MANAGEMENT. 1.1 Creating Value with Risk Management... 1.2 Market efficiency and the Capital Asset Pricing Model (CAPM). 1.2.1 Market Efficiency 1.2.2 Portfolio Theory... 123° Eq rium Theory 1.2.4 Arbitrage Pricing Theory. 1.3 Performance Measurement 13.1 Treynor Measure, the Sharp Measure, and Jensen's Alpha 1.3.3 Cross-Sectional Analysis. 1.3.4 Refinements to the Basic Return-Based Performance Models. 1.3.5 Portfolio-Based Performance Analysis 1.3.6 Returns-based style Analysis vs, Portfolio-based Style Analysis. 1.4 Financial LAL Metallgeselischatt — 1.4.2 Sumitomo Bank... 1.43 Bankruptcy OF Barings. 1.4.4 Long-Term Capital Management. 1.45 Allied ish Bank (Ala) LAG Kidder PeabodY.snnnnnsnnn LAT Bankers Trust REVERT) .. 1.4.8 Chase Manhattan and their involvement with Drysdale Securities 15 Code of Conduct... 15.1 Introductory Statement 15.2. Code of Conduct 15.3 Rules of Conduct 15.4 Applicability and Enforcement 2, QUANTITATIVE ANALYSIS 2A Basics OF Probability And Statistics. 2.4.1 Basic probability 2.2 Probability distribution. 2.1.3. Special distributions, re‘fe acer WWW.GFEDU, eu 8 HE 22. Statistle Theory. 221 Descriptive statistic 2.2.2 Sampling Theory ne ve 6 223. Estimation Theory. 2.2.4 Hypothesis Testing... 23. TheLinear Regression Model 23.1 The Two-Variable Model, 23.2. Multiple Regression Model 2.4 Monte Carlo Methods 2.4.1 What is Monte Carlo Simulation ? 25. Estmating Volatilities and Corrations 25.1 Estimating Volatility 25.2. ARCH Model. I 25.3 EWMAMod 16 25.4 .GARCH(, 1) Model. wo 25.5 Using GARCH(1,1)To Forecast Future Volatility enn a 2.6 Extreme Value Theory. 26.1 Roleof EVT. 26.2 Two broad classed of models in EVT nr 3, FINANCIAL MARKETS AND PRODUCTS. 3. The Mechanisms of Market Mar jement.. 3.1.1 Exchange and clearinghouses reduce counterparty risk mechanisms. see BB 3.1.2 Mechanism for dealing with sovereign risk exposure, 32 The Properties of Financial Products. 3.2.1 Hedging Strategies Using Futures 3.2.2. Determination of forward and futures prices 32.3 Pricing Commodity Forwards and Futures. 3.24 Commodity Spot and FULUFeS MarKetS nnn 94 325 Interest rate futures. 3.2.6 Swaps. 3.2.7 Mechanics of Options Markets 3.2.8 Properties of Stock OptionS nm 3.2.9 Option Strategies... 33. The science of term structure models. 3.3.1. Interest Rate Tree (Binomial) Model and Risk Neutral Interest Rate Tree 3.3.2 Fixedsncome Securities and BlackScholes-Merton 33.3 Bonds with Embedded OptionsS887 WWW.GFEDUNET eh Uh HO 34 EVE 342 343 344 345 346 35 a 44 Value At Risk aaa a2 443 ana ans 416 42 Credit Risk, 42a 422 43 43a 432 433 434 44 Option Pricing.. 4aa 4a2 443 Foreign exchange Different Sources of Foreign Exchange Risk Exposure Different Types of Foreign Trading Activities. Sources of Most Profits and Losses on Foreign Exchange Trading. (On-Balance-Sheet Hedging and Off-alance-Sheet Hedging With Forwards Diversification in Multicurrency Foreign Asset Liability Positions. Combination Strategy Corporate bonds.. VALUATION AND RISK MODELS Introduction to value at risk. volatlity in VaR Model VaR Methods Components of a VaR Model Risk Budget Stress Testing. ‘The Rating Agencies. Credit Risk Measurement. Fixed income securities. Bond prices, discount factors, and arbitrage Bond prices, spot and forward rate Yield to maturity One factor Measures of price sensitivity. Binomial Model. The Black Scholes-Merton Model, The Greek Letters2 RATT WWWOFEDUNET ew Oe se 1, Foundations of Risk Management + Creating value with risk management + Market efficiency, equilibrium and the Capital Asset Pricing Model (CAPM) + Performance measurement and attribution + Sharpe ratio and information ratio + Tracking error + Factor models and Arbitrage Pricing Theory + Risk management failures + Case studies + Ethies 1.1 Creating Value with Risk Management 1) Reducing the potential costs of financial distress and bankruptcy When bankruptcy and financial distress are costly, reducing risk can increase the value of the firm by reducing the present value of expected future costs of financial distress in an amount greater than the cost of the hedging strategy employed. 2) Reducing the volatility of taxable income Because of the nature of the corporate tax code, reducing the volatility of taxable income ‘can reduce a firm’s tax liability and increase firm value. 3) Reducing the weighted average cost of capital ‘The optimal amount of debt in the firm’s target capital structure can be increased by risk-reduction strategies, leading to lower funding costs and increased firm value. 4) Reducing diversifiable risk A large shareholder may be valuable to the firm so that risk-reduction strategies, which reduce the risk and required return ofa large shareholder, can increase firm value. 5) Improving management incentives Risk management can clarify the relation between managerial decisions / actions and firm value, leading to more efficient management incentive compensation schemes. 6) Reducing the probability of debt overhang 4 easeSCY WWW.GFEDUNET ew + 8 By reducing the probability that a firm will become over-leveraged, risk management can increase firm value by reducing the potential for conflicts between the interests of debt holders and the interests of equity holders and managers. 7) Reducing information asymmetries Risk-management strategies can increase firm value by reducing the problem of asymmetric information, thereby reducing the firm’s cost of capital. 1.2 Market efficiency and the Capital Asset Pricing Model (CAPM) 1.2.1 Market Efficiency ‘The CAPM depends on the assumption of market efficiency, which states that market prices reflect available information. ‘The three forms of market efficiency are weak, semistrong, and strong. * The weak form says that no excess returns can be made on a data set of historical price data. + The semistrong increases the data set to all public data. * The strong form increases the data set to all public and private data 1.2.2 Portfolio Theory 12.2.1 Risk and Return of Portfolios of Risky Assets Return + Single asset: E(R)-EPIR=1/n-ZRi + Two-asset portfoli E(Rp)=wiE(Ri)+ woE (Re) Risk + Single asset: 0? PR - E(R)}*=Vne[R; - ER)? «© Two-asset portfolio: 0} =wio} +o} +20, cov,z= wat + Wo} +2m0,610,2,SOR wwworeouNer ek Bie 1.2.22 The Efficient Frontier and the Capital Market Line Efficient Frontier Effects of correlation on diversification benefits Markowitz efficient frontier + Allrrisky assets are contained + Efficient portfolio: well-diversified or fully-diversified Optimal portfolio + The highest indifference curve that is tangent to the efficient frontier + Different investors may have different optimal portfolios Key concept: When a riskless asset is added as a portfolio asset choice, the new efficient frontier is a line beginning at the risk-free rate that is tangent to the risky-asset efficient frontier at a point representing the market portfolio. Capital Market Line ‘The capital market line (CML) expresses the expected retirn of a portfolio as a linear function of its standard deviation, the market portfolio return and standard deviation, and the risk-free rate. FR)= fye| BBO | Ou Figure 1: The CML vs. the Efficient Frontier 6 See A,2188041 WWW.GFEDUNET fea» Sot =H Capital market line Efficient Frontier Market Portfolio Risk 0» 1.2.23. The Security Market Line (SML) Key concept: ‘The CAPM equation describes the SMI, and indicates that the expected retum on any | security or portfolio is determined by its systematic risk as measured by beta. CAPM 4 total risk=systematic risk + unsystematic risk =market risk + diversifiable risk 4 Required return only depends on portfolio’s systematic risk, not its total risk 4 Security market line B(R)= Ry + A(E(Ry)-Ry) Quantity of risk: beta Price of risk: market risk premium Risk premium: beta x market risk premium 4 Beta: a measure of systematic risk COV (R.Ry) Var (Ry) CML ys. SML ‘CML: Measure of Risk: total risk (6), Application: efficient portfolio 4 SML: Measure of Risk: systematic risk (4), plication: properly valued asset eu i 7seth WWW.GFEDUNET ue ie a Assumption Of CAPM The capital asset pricing model (CAPM) requires a list of assumptions to produce its equilibrium where all investors hold the same portfolio of risky assets (i.e., the market po-tfolio) and either a long or short position in the risk-free asset. The CAPM’s assumptions listed here. 4 Investors seek to maximize the expected utility of their wealth at the end of the period, and all investors have the same investment horizon, Investors are risk averse. Investors only consider the mean and standard deviation of returns (which implicitly assumes the asset returns are normally distributed), 4 Investors can borrow and lend at the same risk-free rate. 4 Investors have the same expect: ns concerning returns, ‘There are neither taxes nor transaction costs, and assets are infinitely divisible. This is often referred to as "perfect markets.” A. weak. B. strong C. unimportant D. semistrong only Answer: B ‘The CAPM assumes that investors all hold the market portfolio because they all have the same expectations. This implies that they all have the same information, and there is no private information that influences the asset prices. ‘The Price Of Risk Using the CAPM notation, the price of the risk is the difference between the expected rate of return for the market portfolio, and the retum on the risk-free asset. ‘The Quantity of Risk ‘The quantity of risk, which is called the beta, is defined by: 8 ou: AAefetoy www oreDu NEF “ee = Be st Risk Premium and CAPM ‘The risk premium is equal to the price of the risk multiplied by the quantity of risk, and the CAPM relationship is then written as follows: BIR = Ry + BLE(Ry)~ Re] Example: In the CAPM, the beta for an asset is also referred to as its: A. price of risk B. risk premium C. quantity of risk D. information ratio Answer: C Beta is the “quantity of risk”, and the market risk premium[E(Rp-Re] is the “price of risk.” The product [E(Re-Re]*B is the risk premium of an asset. The information the CAPM. ratio is not directly associated 1.2.2.4 Nonstandard forms of capital asset pricing models CAPM is effective when it comes to describing equilibrium returns on a macro level, but it is not necessarily accurate when examining individual investors’ behavior. Alternative models to CAPM are examined by including real-world influences, such as the existence of short sales, no risk-free borrowing rate, the existence of taxes, inclusion of nonmarketable assets, investor heterogeneous expectations, and the inclusion of “price affecters.” The end result is that many of CAPM’s original conclusions still hold, even after relaxing many of the previous assumptions. Multi-period versions of CAPM include consumption-oriented CAPM, CAPM including inflation, and multi-beta CAPM. We find that growth rate of per capita consumption does affect the equilibrium returns. Also, analyzing CAPM, when including inflation, results in an equilibrium relationship for the expected return on any asset that looks to be similar to the CAPM simple form; however, market price of risk and asset risk are both modified. Lastly, it was determined that a multi-beta CAPM would allow an investor to have unique hedges against specific risks that are of most concern to the investor. Vee RTA a220877 WWW.GFEDUNET e+ i» 1.2.3 Equilibrium Theory By assuming that all investors have the same expectations concerning assets, they all then have the same return, variance and covariance values and construct the same efficient frontier of risky assets. In the presence of a risk-free asset, all choose to divide their investment between the risk-free asset and the same risky asset portfolio M (market portfolio). 1.2.4 Arbitrage Pricing Theory 1.2.4.1 Deficiencies of thie capital asset pricing model (CAPM) > Relies on several restrict assumptions. > Aone factor (meaning it considers only market risk) model. However, it appears that many factors other than market risk drive stock returns. 1.242 Assumptions of APT > The capital markets are competitive. > Investors always prefer more wealth to less wealth with certainty. > Asset returns can be expressed as a linear funetion of a set of & risk factors (or indexes). 12.43 Formula where E(R))= 4p + By + Bade +o + Oy Aw a is the expected return on an asset with zero systematic risk. 4, is the risk premium related to the jth common risk factor. », is how responsive asset # is to the jth common factor. (called factor betas or factor exposures.) 1.3 Performance Measurement 13.1. Treynor Measure, the Sharp Measure, and Jensen's Alpha Three commonly used risk/return measures are:Rb WWW.GFEDUNET ak Sk ‘Treynor measure of a portfolio [a= Sharpe measure ofa pontfolio= std f| A portfolio Sharpe ratio higher than the benchmark ratio( if statistically significant) will offer evidence of superior performance. Statistical significance can be tested using the t-statistic as follows: MER, MER, and MER, are mean excess retums of the portfolio, P, and the benchmark, B, respectively, op and o, are standard deviations of the total returns of P and B respectively. N is the number of observations. Jensen measure of a portfolio (Jensen’s alpha) = E(R,)—Ry ~[E(Ry)~ Re], Alpha (0) plays a critical role in determining portfolio performance. The performance indicated by alpha, however, could be a result of luck and not skill. We conduct a t-test under the following hypotheses: Null (Ho): True alpha is zero Alternative (Hq): True alpha is not zero Where: S.£.(a)= standard error of alpha ‘The three risk measures e different perspectives and may give different rankings of ‘two portfolios. (For example: a portfolio with low diversification may have a higher Treynor measure, a higher alpha, but a lower Sharpe measure than another portfolio.) Figure 2: Characteristics of the Sharpe, Treynor and Jenson indicators eae RA u28008 WWWGFEDUNET ew 8 A Sharpe | Toallo) Portfolio theory | Ranking portfolios with different levels of risk (sigma) theory Not very well-diversified portfolios Portfolios that constitute an individual's total personal wealth Treynor | Systematiog 8) | CAPM | Ranking portfolios with different levels of risk (beta) Well-diversified portfolios Portfolios that constitute part of an individual's personal wealth Jensen SystematiegB) | ___CAPM ‘Ranking portfolios with the same beta (beta) Exampl With respect to performance measures, the use of the standard deviation of portfolio returns is a distinguishing feature of the: A. beta measure B. Jensen measure CC. Sharpe measure D. Treynor measure Answer: C The Sharpe measure is the portfolio return minus the risk-free rate divided by the standard deviation of the return. The Treynor and Jensen measures use beta. The answer “beta measure” is a nonsensical choice for this question. 1.3.2 Extensions to Jensen's Alpha, Tracking Error, The Information Ratio, and The Sortino Ratio ‘Tracking error and the information ratio build upon Jensen's alpha. Tracking error is the standard deviation of alpha over ti TE =0(R, ~R,) . Where R, denotes the return on the bench mark portfol The information ratio is the average alpha over time divided by the tracking error. a, TE IR IR can be tested for statistical significance by using t-statistcs estimated as: Hite.2s WWWGrEDUNET le = ie IR S.ER) where S.EUR)=1, Y is the number of years of observation. 3 Bye wW a The Sortino ratio should be used for skewed data and/or there is more of a focus on the Sortino ratio: ‘The MSD, is a semi variance that only measures the variability o the portfolio’s return observations below R.,, Based on 80 monthly returns, you estimate an actively managed portfolio alpha = 1.3% and standard error of alpha= 0.13%. The portfolio manager wants to get due credit for producing positive alpha and believes that the probability of observing such a large alpha by chance is only 1%, Calculate the t-statistic, and based on the estimated t-value would you accept (or reject) the claim made by the portfolio manager. A. #10, accept B. t10,reject, C. &7, accept D. e7yeject Answer:A =_ alpha _13% 49 SE(alpha) 0.13% With 80 observations and such a large t value, you would have reject H0 (alpha=0). ‘The manager should receive credit for the statistically significant alpha. 1.3.3 Cross-Sectional Analysis Cross-sectional performance analysis classifies performance over a period of time. Cross-sectional analysis offers an easy and quick way to rank performance data; however it is subject to numerous shortcomings, as explained as follows: + There is no adjustment for non-survivorship. + There is no adjustment for size. + There is no adjustment for risk. eka aT BScf20tT WWW.GFEDUNET ea: 1.3.4 Refinements to the Basic Return-Based Performance Models + Hetroskedasticity: Correcting for non-constant error variances to draw valid t-tests inferences, + Serial correlation: Correcting for non-constant error variances to draw valid t-tests inferences, + Bayesian correction: Using prior knowledge for making ex-post adjustments in regression coefficients. + Benchmark timing: Assessing market timing skills of portfolio managers by incorporating up and down market betas. Using a benchmark timing model, we can empirically test whether there is evidence of superior market timing skills exhibited by the portfolio manager. The regression equation is: Ry = p+ Bp Ry + MLD, Ray I+ Err Where D, is a dummy variable that is assigned a value of zero for down-market and one for up-market. 1, is the difference between the up-market and down-market betas and will be positive for successful market timer. +A priori beta estimate: Developing forecasts of betas (ex-ante) for assessing ex-post performance. + Value added: Using a different approach to examine returns behavior. + Controlling for public information: Explaining variations in realized excess returns on a portfolio after adjusting for dividend yield, interest rates, and benchmark portfolio returns. + Style analysis: Estimating skill based on selection returns after adjusting for style returns, + _ Controlling for size and value: Estimating variations in realized excess returns after adjusting for size and value factors. Fama and French (1993) proposed performance analysis techniques based on the following regression, Ryyy =p + Br Rey) + By XVL, + By xMY, +, Ho o Where:2m WWWOFEDUNET “eu « ae AAE , = the return to a long portfolio (on small capitalization stocks) and short portfolio (on large capitalization stacks) AMY, = the return to long portfolio (on high book to price stocks) and short (on low book to price stocks) Sharpe's style model is a retums-based style anelysis method. The procedure uses regressions to theoretically decompose portfolios and indexes according to a set of factors, where the weights om the factors are constrained to being positive and summing to 1. it requires quadratic programming. It is simple, but it does not allow an assessment of the effect of manager's decisions. The model can be written as: Ry =F +OpaBy +--+ Dae Fe Hen Subject to: 0< bp, s1 and dy, F,= the retum on index k en = the residual of the return of the portfolio One way to interpret 2, is the weighting of class k in the portfolio. This interpretation is valid given the indicated constraint of the weights all being non-negative and summing to L Example: The main effect of hetroskedasticity and serial correlation refinements to a regression model is to achieve: 1. unbiased standard errors of estimate (SEE) I. unbiased standard errors of the estimated coefficients IIL. unbiased and valid t-tests for drawing inferences about the values of the parameters IV. none of the above A. IV only B. Land IL C. Mand It D. 1, Hand Ill Answer: D Refinements lead to producing unbiased SEE, which in tum produces unbiased standard errors of the regression coefficients and which leads to unbiased t-statisties. SWEAT ATA 528H WWW.GFEDUNET ae - Be 8 1.3.5 Portfulio-Based Performance Analysis Portfolio-based style analysis uses a multifactor model to identify the effect of manager portfolio decisions. It compares the return of the portfolio to that required given the risk exposures, It delivers more information, but it is more subjective. + Portfolio returns are composed of returns from the multiple factor returns and specific returns, Portfolio returns can be defined as: B= YX x Re Sp ‘Where: R = the total portfolio return YX Re =the common factor return s, = the specific return + Performance attribution can further be decomposed into three components: Active systematic returns, active residual returns attributed to common factors, and active residual retums attributed to portfolio specific factors. Rog = Bou ®Ra)+ (SX pun ® Ba) Span Where: Brg =the beta for the active portfolio R, =the return of the benchmark Xpaq_ the residual active portfolio exposure to factor k Ry =the retum of factor k Sra =the specific return of the active portfolio + Performance analysis focuses on value and statistical significance of the attributed retum series. Using a time series of data, we can further decompose active systematic, residual common factors, or residual specific returns. We can examine active systematic returns in terms of three components: expected active beta return, active beta surprise, and active benchmark returns. 6#8401 WWW.GFEDUNET ea + 1.3.6 Returns-based Style Analysis vs. Portfolio-based Style Analysis + The returns-based approach only analyzes the returns, and the composition of the portfolio is not directly a consideration. Thus, the analysis daes not allow for pinpointing the source of value added. + The porffolio-based approach examines each component of the portfolio to determine the type of risk and return the portfolio should have. The differences between actual and projected performance can be attributed to specific decisions of the manager in the portfolio's composition. Exampl ‘The returns-based approach and portfolio-based approach are: Returns-based approach Portfolio-based approach Top-down method ‘Top-down method ‘Top-down method Bottom-up method Bottom-up method Bottom-up method Bottom-up method Top-down method Answer: B Retums-based approach is Top-down method and Portfolio-based approach is Bottom-up method. 1.4 Financial Disasters 14.1 Metallgesellschaft ATES IEE + ASSUMES 2), JeERGIR Ae TI AR + SAMEMMTRANEY, SRALARBWASILHEN (stack and rollrolling stack) HPHAKE ERRATA: + REZEPUB Chasis risk), 15h Hi) (L5| i (backwardation to contango)28809 WWW.GFEDUNET. ele hs NE + UBER . IE Cre ARB TSHR . RELBR . ALA TENS A | + RRR . REMATCH, RACAL RN SER, . RASH HOA Blt: 1993 4 OPEC RAPA, FH 1993 FAK, BMGT AK, THEMIS, SRR OTEE A ME RE O, UETTT AL Metallgeselischaft SHIR i MURR FBS LI THAK. 1.4.2° Sumitomo Bank PAB AR Be + (BARE (comer the market), AA KASHMIRI, TAINTED A RAMA. ATARI. CITRON F MRA MUA START NE. + FIN, OTP AROMAT Laka, SOA, WALA EEA: «+ THAR: . ATT Oh FB, . MULES BEAR VaR © BERBS: . RRR LAR (LME) EPA. . AAAS, ALK ATTR, MINBAT WEA & . AA ARMAGH, REMefits Www.oFEDUNET ele Bie < L FEE: 1996 ZEA PSH DN, ENTER, THA TTR FBR. 1996.5 AV, AA BM 2000 HATHA, RHR. ATT, HERMIT ERA 1996 455 A 31 Bagi 34 $305), LME SAB OPH Hh 2712 SeTEERE 1740 Fe, HERAT HGH 26 (LRT. 143. Bankruptcy Of Barings FOIE BARB SE: + ARBRIRASRIBR, AAR FL By (YAR A NS AT BALE + ADV ELS 225 #8 ze 19000-19500 DANTE ey EPSTU TEL straddles) ESRUSTNT + THK + AB HRRCK TUE ARSE AY HES VaR + MIS BELZC ARME RT + BRU. . CAFR ERRER, OAT, ANT aM . AA ARPS, REE, ERG LUROUKE. BAHE, AR * AAS HBR EM, SEA MK, WE RIESE . BAN AO EE TA Att: 1995 4F 1/17 BRMUNAAE, FZ: 225 RA THEAE 18950, 1/20, BARAMLE—F, DANUAT AB, CAPRA BONS, 123, ABA FE 17950. AT HARES, SAFER NTR F, F 1-2 ABT BR 8 (LHD, 2230, RES ERAME, BROS. FRITH, BATSofbAT WWW.GFEDUNET ee Bie fale 14.4 Long-Term Capital Management LCM #38 5: © LTCM SePeate AR TTT THE HEE” BR, BAS AA AE fa 8 Fy SB, LAP HR TE ANT ih EE DE TT AS ly E, LA, A kh. + AEH, ERM, HRPM, BRITE OSE, WAAL, AT BAL SERA RST: + THAR: © VaR EA S28 2 Bet, + SAX VaR HEATH, + RARTEAM, + HEA: RRA, MMT REELED + BE. RAR RT Pe RN, MRAM RE, ERR WUAEASL APE 2 BEPC ES AIR BED Bite: 1998 8 A, BTA RAM TPR ROM AR, Fn LBS, AGRE ALT A, BAA, EES, #1999 E12 31 ARBAB HAESPAEEM T 3S EMS, URS SMR 90 KIRAN SAS TRAN HET te, BLE TSS AN TS ARLE TH, FOU BIR. i ARCMEIA; ASE HAS BOAT EFF kB LCM YR SUPE T, PEAS Or LK, Ca BME, BAAR PR, Waa EAL TCM KSA. L TCM Freee AE Er Rubee eTi TTAtke WWWGFEDUNET oak SU «HL CORE BM ANG ET AR BIIMRES. AS ARP SRAM IES) 9 AT BU, AMIN 150 SRP MMA PM 90%, VLA (CRUE MTH, WKS 1 7, CERT UR. 9 A 23 Fl, SeiBe AMAR ZehF, LL Merrill Lynch . JP Morgan eft 14 3 HBR PEGE RRULYSHE 8 36.25 (LRTI T L TCM90 6AtARAL, JEL ET LTCM , Jit BS TENS. 1.4.5 Allied Irish Bank (AIB) > RIE > SRAM AIS LIE ATA IOL > RYfl: 2002 4F 2 A > BEA RNICATE HEB AMT SM 7.5 (LSE A > EA‘: AB AAT FRA LAK Baltimore FAA] Alifirst ASNCAE 3B Rusnak > IA 1997 EFFING, RARE REA Se EERE REM ENC ATE EL Aa > PLPMELTSE LAUFER > GMRTRAREDIG, SEA Leeson ET EARTH 45 TRA > (ATT eA BTM SEA P,P ATE TR RE RETR > (AOR ESE EAB OA AT CEE, MARR Me DERE LES, (AMET, AERA MER’. > Ge SRT AAT EA DEAE Bly ELIE ERAT 9 AMBRE MMT Aah MBAS. > ROD OA, ARPA RANGER, ME: Se BZ BEAM. > 7, BARUNRARRT ABB! ewkanonmy, afeof WwWGREDUNET ele uk I 2 2002 4F 1 A, AIB EARMIAR AIF 2) Allfirst XY SL BE ABE eT IAT EK, (OEE HA Rusnak RMT POR REMAS LH, BT AE Merete T REM ZT E. i889) Rusnak $86 T 12 FKSETEARS A ASME Sy (A A AE EE ETE Rihiti. Bb MYT BES A eR: Rusnak FEZE 1997 RFE ORE AMER ETO Pe EA MMA TRAN, BHR itm Allfirst MIA 7.5 (ZIMA. Rusnak 4 Allfirst (CAE RINE HA. HARTA RT SE Fl. 2001 FFE A FRB Rusnak WAKA MOM KTR, WABKA AMAL MAIER AATEC BEB AAT A FSA Sy AB 3H A DS HR HE BA NIE BS LSB CRusnak 930 34£ RAs 250 TSE TG). HME AEREAT BT 3 A WSR AAA RAT APE, (E Rusnak FARE FIR MEAL) JEG A TSE, BAR TRT 1%, RATAN, TB GAA, Rusnak WREAK AK OI BMI TEZE SE, TD Alifirst A9E SARRERML, DEKE | AMT BPE Rusnak BAKER SORIA THER. WEL Rusnak i RARE 1997 ERIE T (AUF ULZR TE 5 AER AREER SEA, HEE AALTY A AR TERRA HE EEA AY BE ae HER SE BOL Be FRY, TATRA, AAA ALA Allfirse 692 Xa Sa PRT WL, MATMBRT BAR, TOLLE Ae Rusmak ii 17] HA ARAT CS ASHES Fe SETTLE. ERAT AAA EIS ERT BIA RAL AR DG BON LL. JRE AR, Rusnak MMA ATE Me A CME. BEAR, Allfirst BRET ES Sh — OY SPR NL BB SEG AUB ELAT HE SABES BIE BATH HATA Bt BL FRITS Allfirst SM SARE AL SET EL ABAK, oh AIB RATHI SEP ARE. AIB ii T ABMS REG AMA, BAL 5% Allfirst AMPA, ABTA BWA: EARLE. ZS RA, BEALSBA WWWGFEDUNET ese Siok hE RAR BEE. LEG Allfirst HTH KAA REE 2002 469 A 26, AIB Epil Allfirst HA. BRA BARAT ITER AAT EAA 7.5 (2EIC, HAT Alfirst 9 AES} AIM T Ta) L, BE T MEAG Og EALTG AE RT PZ > PEMA AED, SEAR RAT A EA BA TR RAT. A HEBK. ATONE FER 15.5%. TTAB HK. ABE, RHRER, BRERA MITE 2001 FRM - RRA LS HAY 7, 2% FREE 6.496, SAPP ARFEL MEH 10. 8967F BEB 9.9%. AAR T HLS RIED, Wm AORAL IE, FFE > RALBW AAS, BRIER MA. > BPP ACE de abt BE A Bid OY: SALADS FS LH AKAGE ML, HERR BAAR AT AUR Ot» Rt Se AS hl ATER. > ARH > Te > EMAC ES > BU AERP > PRES > RaRSAMR, BLK > ARAL, BEAR, 1 Rusnak SDAP, PR ELIE, RETR. > Allfirst (E29 AIB AUF 2A, BEATER HY ALB FS Jeet SOLAN FREER, HRM MLA Rusnak fT eM PEAS it 1.4.6 Kidder Peabody: > HRA2:f2 0 WWW GFEDUNET Bw + St HEL RHR lett HANSA MLH ABM, MEM 3.3 (237A RY: 1994 4 SEA: Orlando Joseph Jett BAR: > REL SEC ARE FHWA AT A, BER ALIBI 820 DRAKA SCAT 20 HTT, BERG ERR OMS > GBZE 1994 FAH Kidder 4% #5244247 PaineWebber, Jk PaineWebber BORA THR. > SILER a 3 1 — Th 1 fl RB > REAMAMK, WA Beta PURMAEMAARAE ESRB, ASHE SAE GI Sa HE AAR HTE RL RR, AL BEA vvvy > > 1.4.7. Bankers Trust (32 Ela FART) > 4 BRA > 7620 te OO EFEPAE, RAT (1998 FSR RIT) ZEalaT SBT AE Ay ae FT > ASTI: 1994 4 10 9 > RAMANA ESR PAIR, 1.5 ZIRT > PARE LL > SRPUTEHRAWIMEERS, SREP ERS BRAK > EP RVHMT ARETE HM BN PS > ROR ERE Hb, RTA HO BBA > RARRT (BT) REM S ESHA H—IMA A Gl (Federal Paper Board Company). #7 #5416 27)(Gibson Greetings). 4°67 AFI (Air Products and Chemical) #128 2 (Procter & Gamble)— dit, Ef] PATTI AAT IO AIT BB RE ALOT LAS SWE OST N.em WWW.GFEDUNET ee + Bue BL PUA ARTA AF MAU Be AT eI eS ARBAB EBA FERRE > in: SERA aS fk FRAT UL T RAR T HC. RAT PETES MASE A AMHR AE PE. IES, HAART 20 18% ALA, (LOH EMA. BT RARE AST 1994 HEE BAM, KGAA AER BRT ARR. > 1994 10 FA: RATIRA ALUM PRAT, ORAS 1 1 9500 TRG CK ASF] BIRT He 1994 MEERA AY 1 {Z 200 TAWA ETAR ANS BOs TSAR RAT LENA ATE AAA I rH MTTRS. > 1994 4 12 As PRAT TRIE SUSE, BK 1000 RIG, Ly AE A AALS ET EA A, MMP RR ATHLHBASMIE. Mik, (PRT RAK Ue ATES > 1996 4F 1 As PRTRRSE LTA, Hl 6700 HRT, A 1994 SUIBLAE BE eA SBP RAT SEAT AI He SS BEIM 1 1Z, 700 HTH 63% > 1996 5 A: BART SRBA RAM TMI, EAA RA 3500 FR TGA, HARTER TTA SILA ANY 7800 77 RIG ER» > Lessons to be learned > Give adequate attention to all aspects of risk > An enterprise risk management program must balance the "hard side” of risk management (including policies, limits and systems) and the soft side including people, culture and incentives). It is ironic that BT, a company that was considered by many to be a leader in risk management and in innovative derivative products, lost so much of its reputation as a result of operational risk (in this case, sales practices). > Honesty is always the best policy > Reputational. risk management suggests that in a time of crisis, management should focus on integrity and openness in dealing with customer complaints and public perception: if the allegations brought by P&G had any basis in reality, Bankers Trust did an inadequate job in BRAT TMM 2sener w fea» UE + resolving them: > Align incentives properly > Performance pressure may encourage participatio deals that ultimately backfire, especially if incentives and oversight are not aligned properly. Despite the urgency of the rush to sell a deal or gain a client, proper thought needs to be given to the long-term wisdom of all aspects of the deal. > Practice good stakeholder management > Clients are stakeholders, too! In the rush to create profits for shareholders, attention must still be given to the clients who are integral to the business. 1.4.8 Chase Manhattan and their involvement with Drysdale Securities > PR > Drysdale B— KPA EB AA] (AAAS 2000 737), Chase Manhattan (£92822 A154 $29 Drysdale iEH2 AAD) 3 (LI TEAAB ARSE A, AHR NH BY): 1976 4E BFR: Drysdale UAB, 3 LR TRR AAT, Chase Manhattan 2234) F HX EGER RR STE > 5 Bankers Trust SPR EGR A > AHA BRL HA Bs | RHA AL > ZERO, Chase Manhattan RHA RIM, TERA Bh, SAMI A HEH A ACR F AB Chase Manhattan 42 —4*P [A] Amie 1.5 Code of Conduct On February 26, 2007, GARP’s Board of Trustees unanimously adopted a Code of onduct forall GARP FRM-holders and candidates, other GARP certification and diploma holders and candidates, members of GARP’s Board of Trustees, GARP Regional Directors, GARP Committee members and GARP staff. 6 en ADAPER WWW.GFEDUNET He Bis HO 1.5.1 Introductory Statement ‘The GARP Code of Conduct (“Code”) sets forth principles of professional conduct for Global Association of Risk Protessional (“GARP”) Financial Risk Management program (FRM®) certification and other GARP certification and diploma holders and candidates, GARP’s Board of Trustees, its Regional Directors, GARP Committee Members and GARP?s staff (hereinafter collectively referred to as “GARP Members”) in support of the advancement of the financial risk management profession. These principles promote the highest levels of ethical conduct and disclosure and provide direction and support for both the individual prac The pursuit of high ethical standards goes beyond following the letter of applicable rules and regulations and behaving in accordance with the intentions of those laws and regulations, itis about pursuing a universal ethical cufture. ner and the risk management profession. All individuals, firms and associations have an ethical character. Some of the biggest risks faced by firms today do not invoive legal or compliance violations but rest on decisions involving ethical considerations and the application of appropriate standards of conduct to business decision making, ‘There is no single prescriptive ethical standard that can be globally applied. We can only expect that GARP Members will continuously consider ethical issues and adjust their conduct accordingly as they engage in their daily activities. This document makes references to professional standards and generally accepted risk management practices. Risk practitioners should understand these as concepts that reflect an evolving shared body of professional standards and practices. In considering the issues this raises ethical behavior must weigh the circumstances and the culture of the applicable global community in which the practitioner resi 1.5.2. Code of Conduct ‘The Code is comprised of the following Principles, Professional Standards and Rules of Conduct which GARP Members agree to uphold and implement. 1.5.2.1 Principles 1) Professional Integrity and Ethical Conduct. GARP Members shall act with honesty, integrity, and competence to fulfill the risk professional’s responsibilities and to uphold the reputation of the risk management profession, GARP Members must avoid disguised contrivances in assessments, measurements and processes that are intended to provide business advantage at the expense of honesty and truthfulness. a‘aei2#475 WWW.GFEDUNET eM Sie iE 2) Conflicts of Interest. GARP Members have a responsibility to promote the interests of all relevant constituencies and will not knowingly perform risk management services directly or indirectly involving an actual or potential conflict of interest unless full disclosure has been provided to all affected parties of any actual or apparent conflict of interest. Where conflicts are unavoidable GARP Members commit to their full disclosure and management 3) Confidentiality. GARP Members will take all reasonable precautionary measures to prevent intentional and unintentional disclosure of confidential information. 1.5.2.2. Professional standards 1) Fundamental Responsibilities ‘+ GARP Members must endeavor, and encourage others, to operate at the highest level of | professional skill. + GARP Members should always continue to perfect their expertise. + GARP Members have a personal ethical responsibility and cannot out-source or delegate that, responsibility to others. 2) Best Practices ‘+ GARP Members will promote and adhere to applicable ‘best practice standards’, and will censure that risk management activities performed under hissher direct supervision or ‘management satisfies these applicable standards. ‘© GARP Members recognize that risk management does not exist in a vacuum, ‘+ GARP Members commit to considering the wider impact oftheir assessments and actions on their colleagues and the wider community and environment in which they work. 3) Communication and Disclosure. GARP Members i on behalf of their firm will ensure that the communications are clear, appropriate to the circumstances and their intended audience, and satisfy applicable standards of conduct. suing any communications 15.3 Rules of Conduct 1.53.1 Professional Integrity and Ethical Conduct GARP Members: 1) Shall act professionally, ethically and with integrity in all dealings with employers, existing or potential clients, the public, and other practitioners in the financial services industry. 2). Shall exercise reasonable judgment in the provision of risk services while maintaining independence of thought and direction. GARP Members must not 2 MR TPT MH,2 f2NTT WWW.GFEDU NET Se ae hit 3) 4) 3) 6 D 153.2 offer, solicit, or accept any gift, benefit, compensation, or consideration that could be reasonably expected to compromise their own or another's independence and objectivity. Must take reasonable precautions to ensure that the Members services are not used for improper, fraudulent or illegal purposes. Shall not knowingly misrepresent details relating to analysis, recommendations, actions, or other professional activities. Shall not engage in any professional conduct involving dishonesty or deception or engage in any act that reflects negatively on their integrity, character, trustworthiness, or professional ability or on the risk management profession, Shall not engage in any conduct or commit any act that compromises the integrity of the GARP, the (Financial Risk Manager) FRM@ designation or the integrity or validity of the examinations leading to the award of the right to use the FRM designation or any other credentials that may be offered by GARP. Shall endeavor to be mindful of cultura! differences regarding ethical behavior and customs, and to avoid any actions that are, or may have the appearance of being unethical according to local customs. If there appears to be a conflict or overlap of standards, the GARP member should always seek to apply the higher standard, Conflict of Interest GARP Members shall: »D 2 1333 Act fairly in all situations and must fully disclose any actual or potential conflict to all affected parties. Make full and fair disclosure of all matters that could reasonably be expected to impair their independence and objectivity or interfere with their respective duties to their employer, clients, and prospective clients. Confidentiality GARP Members: » 2) 1534 Shall not make use of confidential information for inappropriate purposes and unless having received prior consent shall maintain the confidentiality of their work, their employer or client. Must not use confidential information to benefit personally. Fundamental Responsibilities GARP Members shall:4:f8 60% WWW.GFEDUNET 1) Comply with all applicable laws, rules, and regulations (including this Code) governing the GARP Members’ professional activities and shall not knowingly participate or assist in any violation of such laws, rules, or regulations. 2) Have ethical responsibilities and cannot out-source or delegate those responsibilities to others. 3) Understand the needs and complexity of their employer or client, and should provide appropriate and suitable risk management services and advice. 4) Be diligent about not overstating the accuracy or certainty of results or conclusions. 5) Clearly disclose the relevant limits of their specific knowledge and expertise conceming risk assessment, industry practices and applicable laws and regulations. 1.53.5 General Accepted Practices GARP Members shall: 1) Execute all services with diligence and perform all work in a manner that is independent from interested parties. GARP Members should collect, analyze and distribute risk information with the highest level of professional objectivity. 2) Shall be familiar with current generally accepted risk management practices and shall clearly indicate any departure from their use. 3). Shall ensure that communications include factual data and do not contain false information. 4) Shall make a distinction between fact and opinion in the presentation of analysis, and recommendati 1.5.4 Applicability and Enforcement Exery GARP Member should know and abide by this Code. Local laws and regulations may also impose obligations on GARP Members. Where local requirements conflict with the Code, such requirements will have precedence. Viclation(s) of this Code by may result in, among other things, the temporary suspen or permanent removal of the GARP Member from GARPs Membership roles, and may also include temporarily or permanently removing from the violator the right to use or refer to having earned the FRM designation or any other GARP granted designation, nm following a formal determination that such a violation has occurred. 30 kes TIA28UCH WWW.GFEDUNET We = Bi HT 2, Quantitative Analysis + Probability distrib + Mean, standard devi Correlation, skewness, and kurtosis + Estimating parameters of distributions + Linear regression + Statistical inference and hypothesis testing + Estimating correlation and volatility: EWMA, GARCH models + Maximum likelihood methods + Volatility term structures * Simulation methods 2.1 Basics Of Probability And Statistics 2.1.1 Basic probability 2.1.1.1 Important Terms Random variable: uncertain quantity/number, including continuous and diserete random variable. Outcome: Realization of random variables. Event: Single outcome or a set of outcomes. Mutually exclusive events: A and B are mutual exclusives. P (ANB) = 0, A and B can not both happen at the same time Exhaustive events are those that include all possible outcomes. Probability of an event: The classical definition and Relative frequency defini 2.1.1.2. Important rules + General Rule Of Addition: P(AUB) = P(A) +P(B) — P(AMB), where P (ANB) is the joint probability of A and B. + General Rule of Multiplication: If event B is conditional on occurrence of another event (A), then P (AMB) = P (A) xP (BIA). SW KOC PENA 31aeSeHCFT WWW.GFEDUNET eae + Sik ea «Total Probability Rule: P(A) = P(AMB) + P(ANB®)= P(AIB)P(B) +P (AJB) P(B); «The Bay's Rules: POA, PALA) YPapaia,) PCA, | A) Aj, Az... An are mutually exclusive events, and these mutually exclusive consist the sample space S. + Aand B are independents>P (ANB) = P(A) xP (B)UP (A\B) =P (A). © Three events A,, A,, Ay are independent PA, 4, 71 A,)= P(A)PCA,)P(A) and each pair of these three events are independent, P(.4,1.4,)= P(A,)P(4,) for j#k. Example: ‘An investor is choosing one of twenty securities. Ten of the securities are stocks and ten are bonds. Four of the ten stocks were issued by utilities; the other six were sued by industrial firms. Two of the ten bonds were issued by utilities; the other ight were issued by industrial firms. If the investor chooses a security at random, the probability that itis a bond or a security issued by an industrial firm is: A, 0.80 B.0.70 €.0.60 D.0.50. Answer: A Let B represent the set of bonds and I the set of industrial firms. The desired probability is the probability of the union of sets B and I, P(BUI). According to the theorems of probability, P(BUI )=P(B)+P(1)-P(BN1), where P(B) is the probability that a security is a bond=P(B)=10/20, P(I) is the probability that a security was issued by an industrial firm=P(1)=14/20, and P(BNI) is the probability that a security is both a bond and issued by an industrial firm= P(BN1)=8/20. 80. (BUI )=10/20+14/20-8/20=16/20: 2 WR AT ATASo fepefy WWW.GFEDUNET ea» st ee 2.1.2 Probability distribution 2.1.2.1 Univariate probability distribution + Discrete Random Variable: Arandom variable takes a finite or countable infinite number of values. + Continuous Random Variable: A random variable takes a continuous infinite number of values. + Probability Distribution Function Of Random Variable X. ‘The probability distribution function F(x) is always called CDF: F(x) =P(X-sx) forall x in (~ P(u) Fora For a continuous random variable, the CDF is. F(x)= P(X z)=1 -P(Z Y=In@X) ~ NUyz,07) A(X) = expat var(X) = exp(2yr+20*) -exp(2u+07) Key concept ‘The normal and lognormal distributions are very similar for short horizons or low volatilities. ‘The relationship between normal and lognormal ‘A lognormal distributed bution exists for random variable Y, when Y=e*, and X is normally ‘The Jogaormal distribution is skewed to the right. The lognormal is bounded from below by zero so that prices which never take negative values. s useful for modeling asset 2.13.6. T, chi-square and F distribution T Distribution a22915 WWW.OFEDUNET se If Z, is standardized normal variable and another variable Z, follows the chi-square distribution with n degree of freedom, Z,and Z,are independent, then the statistic defined as: Zz, peed V(Z,/n) tion with n degrees of freedom Follows Student’s t dist Student's t-distribution has the following properties: 4 Itis symmetrical, so the skewness is equal to 0. A It is less peaked than a normal distribution, so it has fatter tail than normal distribution. A Asthe df (the sample size) gets larger, the shape of the t-distribution approaches anormal distriby Chi-square Distribution If 2,,Z,nZ, are independent standardized normal variables, thenZ = )°Z?is said to have z*distribution with n degrees of freedom. Where: = sample size = sample variance 1ypothesized value for the population variance. ‘The chi-square distribution has the following properties: 4 The z"distribution is skewed distribution, the degree of the skewness depending on the degree of freedom. A Assample size n increase, the distribution becomes symmetrical F-Distribution ° If z and z, are independently distributed chi-square variables with &, and k, df 38 BuROR REEDSef24TT WWW.GFEDUNET 5 ue NL respectively, the variable Follows F distribution with k,and k, df. ‘The F distribution has following properties: 4 The F distribution is skewed to the right, But as k,and_&, become large, the F istribution approaches the normal distribution. A The square of a t-distributed random variable with k df has F distribution with | and k df A There exists a relationship between the F- and chi- square distributions such that: a8 the N (the number of observations ir numerator) —> «© 2.1.3.7 The Relationship between the Binomial Distribution, Poisson distribution and Normal Distribution + Ifnis large, but neither p nor q is too close to zero (thus Poisson is not OK anymore), then the binomial distribution can be closely approximated by a normal distribution, + If n is very large and p is very small, the Poisson distribution may be used to approximate the binomial distribution with 4 = np ‘The Poisson distribution approaches the normal distribution as 2 —>median>mode Skewness<0 (left-skewed) Mean30), X is approximately nenmally distributed withthe mean and the variane| x SI ne ‘When (n>30) then YON, 2-) : i ” Key concept: With independent draws, the standard deviation of most statistics inversely related to ‘the square root of nimber of observations T. thus, more observations make for more precise estimates. 22.24 Standard error of the sample mean The standard error of the sample mean is the standard deviation of the distribution of the sample means. And itis calculated as: whereo, the population standard deviation is known. where s is he sample standard deviation and the population standard deviation in unknown, 2.2.2.5 The Sample Proportion and the Distribution of the Sample Proportion In the population, the probability of success is p and a statistic P to be the proportion of the number of success in sample of size n. When the sample size n >30, the sampling ‘and distribution of proportions is approximately normal with mez, swinceen =, 2.2.2.6 The Sample Difference and the Distribution of the Sample Difference If we have two populations, Population 1 and Population 2, we drawn m, samples of size n, from population 1 and m, samples of »,. Then we can calculate the mean #1, and standard deviation, of sampling distribution of statistic $ of population 1 Avg aT a248808 WWW.GFEDUNET EW + ie HE and the means, and standard deviation, of sampling distribution of statistic $ of population 2. Take all combination of these samples from the two population, we can get a new statistic 5, -S, which is called difference and obtain sampling distribution of the difference. When the samples we drew are independent, the mean of the distribution of the difference is Hs_s, = fs, ~ Hs, and the standard deviation of the distribution of the difference ise, ., ~ Ya", +04," . Now we assume the statistic $ is the sample mean, and the two populations have means and standard deviations given respectively as %s0} and t,,0, then the mean of the sampling distribution for the differences of sample mean is f4y,_,, = Hy, ~ Hy, = /4~ Hpand standard deviation of the sampling distribution loi, +02, -\at/n, +03 /n, 2.2.2.7 The Sample Variance and the Distribution of the Sample Variance for the differences of sample mean can is o, The population variance is: ‘Note the N is the number of the members in a population. ‘The sample variance $* is: yay or 5? (n=1)s* ‘The random variable +— has a chi-square ( 7’) distribution with n-1 degrees of o freedom, when we sample from a normal distribution, 8 Peete28 CH WWwW.GFEDUNET th de HL Example: (A)The last three year-end returns for a stock are 5, -2, and 1 percent. Using an arithmetic mean, the sample standard deviation is closest to: A. 3.51% B. 2.87% C. 1.33% D. 3.11% Answer: A The sample arithmetic mean is: X=(5-2 +13 = 133% The sampie standard deviation is; eee 9Setkts WWW.GFEDU NET eae le He Example: (2)The mean equity risk premium over a 40-year period is equal to 8.0 percent. The standard deviation of the sample is 12 percent. The standard error of the sample mean is closest to: A. 030% B. 1.90% C. 1.26% D. 8.00% Answer : B Note the of the sample here is the number of years. ox= 12/40 = 1.90% 2.2.3. Estimation Theory 22.3.1. Statistical Inference ‘Study of the relationship between a population and a sample drawn from that population. 223.2 Point Estimate Point estimate is to use a single number to estimate the population parameter. There are two Ways to estimate the parameters of the population: Moment estimate and Maximum likeihood estimate, Moment estima the paramters Use the sample’s moment to estimate the population's moment. And then calculate ‘Maxiinum likelihood estimate: Choosing values for the parameters that maximize the chance (or Vikelihood) ofthe data occurring. Example: Suppose that we sample 10 stocks at random on a certain day and find that the price of one of them declined on that day and the prices of the other nine either remained the same or increased. What is our best estimate of the probability of a price decline? maxi ize: p(\- p)’ > p=0.1 Estimating GARCH(1,1) Parameters 50 RAPE,sc WWW.GFEDUNET Ek Sse 4 The best parameters are the ones that maximize the likelihood function md wu exp) Ife Fa ay) A Taking logarithms, and maximizing 5 ¥ Lem Point estimate of the population mean; 4 The sample mean % isa good point estimator of the population mean. 4. Point estimate of the population proportion P which is the proportion of the number of success in a sample is a good point estimator of p which is the probability of success in a binomial population. 4 Point estimate of the difference between the means of two populations The difference of sample mean X,—X, is a good point estimator of the difference of population mean 14 — 44 The properties of an efficient point estimate nt, and If an estimator G is better than others, we expected it to be unbiased, effi consistent. 4 Unbiased ; the expected value of the estimator is equal to the parameter being estimated, E(@)=6 A. Efficient: An estimator is said to be efficient if Var(O.g) In Construct confidence interval for a population mean with unknown population variance When variance o is unknown, 1- @ confidence interval can be obtained: s Fthy s 4, fo and 4-0 are the critical t-value, ¥+f_4,2—= is the upper limit, ¥—6o/: isthe lower limit. = ss . F [F-t.o2sF +4, 2] is the random interval. vn Construct confidence interval for a population proportion ‘When sample of size 1 > 30, the confidence interval for p is given by Paz ft ‘When we draw samples without replacement from a confidence interval are ite population of size N, the Construct confidence interval for the difference between the means of two 2 wk Oa Oh,SBT WWWGFEDUNET ak eT populations ‘The confidence interval for the difference of the population mean is constructed: X= Xt, When the population variance is unknown, we use the sample variance to substitute the population varianceo” 2.23.4 The difference between point and confidence interval estimate When we use a single number to estimate the population parameter, this is point estimate. When we use a range of number into which the true value of population parameter is expected to lie, the range of number is an interval estimate of the parameter. Example: An analyst collects a sample of 50 P/E ratios of stocks that are representative of the market. The mean P/E of these stocks is 20 and the standard deviation is 8.5. What is the 95 percent confidence interval for the mean P/E of stocks in this market? A, 17.21 022.79 B. 17.64 0 22.36 C. 18.02 1022.98 D. 1831 021.69 Answer: B ‘The 95% confidence interval = Mean+1,96 x Standard error = Mean+ 1.96 x Standard deviation / vn=20 £1.96 x 8.5/ V350 = 17.64 1022.36 2.24 Hypothesis Testing 2.2.4.1 Basic Concepts Hypothesis Testing Procedure a 3SR88¢FT WWWOFEDUNET ty Se 8 Step 1 ‘Step 2 Step 3 State null Identify the Selecta level and L-| teststatistic || of alternative significance hypotheses Step 5 Step Donot reject Hy Take a Formulate a sample, Le} decision ule Y| decison Reject Hp Source: Douglas A. Lind, William G. Marchal and Robert D, Mason, Statistical Techniques in Business and Economics Null Hypothesis and Alternative Hypothesis ‘The null hypothesis, which is denoted as. H, is what the researcher wants to disprove. The alternative, which is denoted as, is what will be concluded if the evidence does not support the null One-tail test vs. Two-tail test You should also be very clear on the difference between a one-tailed test and a two-tailed test. We use population mean as example to illustrate what the one tail and two tail test are, The null and alternative hypothesis of two-tailed test for the population mean may be constructed as H,:4= My VS.H,:H# My The null and alternative hypotheses for a one tail hypothesis of the population mean are one of the following: Uppertail: Hy: 4S vs Hy: M> ty Lower tail: Hy:12 My vs Hy: HS}, One tail test uses one critical value. And for upper tail test, we reject the null hypothesis when the value of test statistic is greater than the critical value. For lower tail test, we statistic is less than the critical value. reject the null hypothesis when the value of te: 42g BeTT WWW OFEDUNET ee $e +e 2.2.4.2. Type | and a Type Il error and significance level ‘Wheii we test a null hypothesis, there are four possible outcomes. 4 Reject a null hypothesis when itis false. This is correct decision. 4 Reject a null hypothesis when itis true. This is called a Type f error. 4 Do not reject a null hypothesis when it is false. This is called Type II error. 4 Do not reject a null hypothesis when it is true. This is correct decision. The Type 1 error, denoted by the Greek letter alpha, is often called significant level of the test. If we set a to be 5%, it means that the probability that we reject the null hypothesis when the null hypothesis is true is 5%. The Type Il error, denoted by There is trade off between these two type errors. The only way to reduce both two type errors is to increase the sample size n, Decision True Condition Ho is true Ho is false Incorrect Decision ‘Type Il Error Do not reject Hp | Correct Deci Incorrect Decision Correct Decision Reject Ho | Type I Error Power of the test Significance level, @ =P(Iype 1] = 1~P(Iype I! Error) Error) 2.243. The p-value Compared to critical value method, calculating the probability value (p-value) of the observed test statistic method can provide us further information about how strong the evidence given by the sample shows we should accept or reject the null hypothesis. The p-value is the observed significance level of the test statistic. The p-value is the lowest level at which #1, can be rejected. The smaller the p-value, the stronger is the evidence against the null hypothesis, 2.2.44 Test Statistic and Critical Value and Decision Rule sample statistic-hypothesized value Test statistio=7P Standard error of the sample statistic Critical value Ue OB THA. 5sSGC WWW GFEDUNET 4 The distribution of test statistic (z,t, 72, F) A Significance level (a) 4 One-tailed or two-tailed test Decision Rule + Reject H, if test statistic}>critical value A Fail to reject 1, if test statistic} Autocorrelation Answer: A Stratification is not related to regression analysis. Choices B, C, and D describe tations that can produce inaccurate descriptions of the relationship between the independent and dependent variables. Multicollinearity occurs when the independent variables are themselves correlated, Heteroscedasticity occurs when the variances are different across observations, and autocorrelation occurs when successive observations are influenced by the proceeding-observations. Example: 2. A linear regression function assumes that the equation is linear ia: A. both the variables and the coefficients B. the coefficients but not necessarily the variables C. the variables but not necessarily the coefficients D. neither the variables nor the coefficients Answer: A The term linear applies to both the coefficients and the variables. aba aA sST Www.cFEDUNET Example: Assume a skewness of 0.3 and a kurtosis of 4. For a sample of 150 observations compute the Jarque-Bera value. As B.8.5 C9 D.95 Answer: B 23.2. Multiple Regression Model Generalizing the Two-Variable population regression function, we can write the multiple peplaoneresion function a: =H, 5M, 46 6, j=1...n are called partial regression coefficient. Simple form Y= Xb+e Fe K LY b= Orb BY = (6 826,)" 1X, x, 1X, Xe Xx, IE Xn oe) Assumptions of the multiple linear regression models: AL-The regression model is linear in the parameters and that it is correctly specified. Al.e,~N(,07) x, is uncorrelated with the disturbance term ¢, . AS.E(e,)=0 ‘Ad. var(«,) = 07 . (Homoscedastic) 66 ee 9 a,stk WWW.GFEDUNET ee RE AS.cov(s,,8,)=0 147 A6.¢,~ N(0,07) ATNo multicollinearity ‘A3, A4, AS are the Gauss Markov condition. ‘The collinearity of two variables means that one variable can be expressed linear function of another. In the case of perfect collinearity, ® can not be estimated. In the case of near collinearity, cov(b,,d;) i%j can be very high. 23.2.1 Test of Significance of Regression Coefficients The case testing hypoth: variables mode! about individual partial regression coefficient is as two The case testing the joint hypothesis: the null hypothesis H,:6,=0,=..-0. The F-statistic can be calculated as follow: m BBLEL orp p-9 ‘The degree of freedom ESS is p, the degree of freedom of RSS is n-p-1. ‘The a significance level, if F >¥,(p,n—p—l) then reject H,. Relationship between F and R?: Red d-R)Mn-k) Where: n= the sample size; k = the number of coefficients including the intercept Adjusted, RF =1 (m2 2.3.2.2. Predicted value for dependent variable Predicted values can be calculated as:228815 WWW.GFEDUNET. WM = Be where: 23.2.3. Restricted least squares and unrestricted least squares A restricted least squires regression imposes a value on one or more coefficients with the ‘goal of analyzing if the restriction is significant. There is an implied restriction in each of the two variable regressions: + Baciy * lockup), + Desyenoee * (Xperience), In essence, each of the two-variable regressions is a restricted regression where the coefficient on the omitted variable is restricted to zero. To help illustrate the concept, the more elaborate subscripts have been used in these expressions. Using the i notation, the first specification that only includes “lockup” is restricting b., rene the unrestricted multivariable regression, both Byeuy and ba, are allowed assuming the values that minimize the RSS. There is a much more latitude in restricting a given coefficient than just restricting it to zero. In the three-variable model, the independent variable can be transformed by creating a new variable, say, ¥=¥,-ax X, and then regressing ¥/ on the other independent variable, X,,. The R° from that regression is called a restricted or R?. For comparison, the unrestricted from the specification that includes both independent variables is given the notation R?, . Following is the F-statistic that can test if the restriction is significant or not: (RL-R)/m (RD Mn- The symbol “m” refers to the number of rest would be equal to one. 6 Swe ATTAefor WWWGFEDUNET pe = ese IH Examph Which of the following situations is NOT possible from the results of a multiple regression analysis with more than 50 observations? R Adjusted R? A 11% 69% B. 83% 86% C. 54% 12% D. 10% 2% Answer: B Adjusted R? must be less than or equal to R?, However, if R? is low enough and the number of independent variables is large, adjusted R’ may be negative. Es HiME oSof WWW.GFEDUNET eh + ik «i Example: (1) An analyst is using a statistical package to perform a linear regression between the risk and return from securities in an emerging market country. The original data and intermediate statistics are shown on the right. The value of coefficient of determination for this regression is CLOSEST to: A. 0.043 B. 0.084 C. 0.916 D. 0.957 Risk %(X,) — retun% (¥,) Ml 32 1s 35 22 4 36 45 43 48 5.1 5a 67 52 as Yew, -¥y =249, (4, -Pyy, =908 Sy -Fy = sor: SG Fy = ssr=3312 Answer: C Coefficient of determination = SS R/ SST = 3.312 /3.617 916. 2.4 Monte Carlo Methods 2.4.1 What is Monte Carlo Simulation ? Monte Carlo was derived from the name of a famous casino established in 1862 in the south of France (actually, in Monaco).Monte Carlo Simulation is a statistical simulation method, 0 WRT HA2teks WWW.GFEDUNET uh se 24.1.1 Simulating a price path Geometri¢ Brownian Motiox (GBM) Model {as, where dz is a random variable distributed normally with mean zero and variance dt The parameters 4,and ©; represent the instantaneous drift and volatility at time [as, =5(uar+oeVan), Where® is now a standard normal random variable ‘The process of simulation: Assumption T=100 days, 42=0, 6 =0.1, S,=100_ the result is foltowing, 2.4.1.2. Simulating a price path TABLE 12-1 ‘Step? | Previous Price Random Increment Current Price Sires Variable © as Sus 7 160.00 0.199) 0.199) 100.20 2 100.20 1.665 1.668 101.87 3 101.87 0.445 -0.453 101.42 4 101.41 -0.667 -0.676 100.75 100 92.47 -1.153 “1153 91.32 Monte Carlo Methods: lognormal property of stock price SWS PHBA nS827 WWW GFEDUNET Bae aa a anew 24.1.3 Creating Random Numbers > Inverse transform method > X-U[G1] x=) y=") ntoom + uml sou 2.4.2 The Bootstrap Sw 8 30 ‘An alternative to generating random numbers from a hypothetical distribution is to Sample from historical data. Sy = SC 4+ Rog) Advantage of bootstrap: Can include fat tails, jumps, or any departure from the normal distribution, 2 SURAT HT OAFLAT WwW.GFEDUNET Be «ok «Mt Account for correlations across series because one draw consists of the simultaneous returns for N series, such as stock, bonds, and currency prices Limitation of bootstrap: For small sample sizes, may be a poor approximation of the actual one Relies heavily on the assumption that returns are independent. 2.4.3 Computing VAR 1. Choose a stochastic process and parameters. 2. Generate a pseudo-sequence of variables &» >.» &,, fom which prices are 5, computed 28.5.5, 3. Calculate the value of the asset (or portfolio) F ,,,=/F, under this particular sequence of prices at the target horizon. 4, Repeat steps 2 and 3 as many times as necessary, say K=10,000 5. Creates a distribution of valuGs>--» (F)-OlF.0) [VARGT)- 2.4.3.1 Risk management and pricing methods Under the risk-neutraf valuation method, Monte Carlo simulation consists of the following steps: “=r 1. Choose a process with a drift equal to the risk-free rate, 2. Simulate prices to the horizon S, 3. Calculate the payoff of the derivative at maturity T F(S,) 4, Repeat these as often as needed. 5, Discount at the risk-free rate and averaging across all experiments, then obtain the current value of the derivative : 2.4.4 Acceleration Methods Antithetic variable (X}{§2%) technique: Consists of changing the sign of all the random saffiples This method is appropriate when the original distribution is symmetricBHO WWW.GFEDUNET ee + ik a Control variate technique: Importance sampling technique Which attempts to sample along the paths that are most important to the problem at hand. The idea is that if our goal is to measure a tail quantile accurately, there is no point in doing simulations that will generate observations in the center of the distribution. Strat ied sampling technique Simulate correlated random variables using Cholesky Factorization To account for correlations between variables, we start with a set of indefiendent variables, which then fire transformed into the . In a two variable setting, we construct “1 my eon 24.5 Deterministic simulation Monte Carlo simulation methods generate independent, pseudorandom points that attempt to “fill” an N-dimensional space. Quasi-Monte Carlo is to use a deterministic scheme that is constructed to provide a more consistent fill 0 the N-space. Comparison of distributions. | ____Pstuderandom sequnes on. os: on. 6 2.5 Estmating Volatilities and Corrrlations ‘The Purpose of Learning This Chapter: 4 Awe 9 TIA.2808 WwW.GPEDUNET eu» ik Explaining how historical data can be used to produce estimates of the current and future levels of volatilities and correlations, For the calculation of VAR using model-building approach Current levels of volatilities and correlations. For the valuation of derivatives (e.g., options) Future levels of volatilities and correlations. Three approaches we must master after this reading ARCH (autoregressive conditional heteroscedasticity) EWMA (exponentially weighted moving average) GARCH (generalized autoregressive conditional heteroscedasticity) 2.5.1 Estimating Volatility Define o,, as the volatility of a market variable on day n, as estimated at the end of day nl. o? as the variance rate. Define S, as the value of the market variable at the end of day i. Define 1, as the continuously compounded return during day i (between the end of day For the purpose of monitoring daily volatility, we give the following changes: iis assumed to be zero. m-1 is replaced by m. Then we can get a simple formula for the variaince rate Weighting Schemes The above formula gives equal weight to 12 4,207 p,....-52%y. Our objective is to estimate the current level of volatility, so we give more weight to recent data, ERA Pe HA 15224 F WWW.GFEDUNET ek 8 ~L (a, j) 2.5.2 ARCH Model ‘Adding a long-run average variance rate and be given a weight Where V; is the long-run variance rate and ¥ is the weight assigned to Vi Defining w= pV, then the model can be written 2.5.3. EWMA Model In an exponentially weighted moving average model, the weights assigned to the a decline exponentially as we move back through time. The estimate, o,, of the volatility for day n (made at the end of day 1-1) is calculated from 6, the estimate that was made at the end of day n-2 of the volatility for day n-1) and u,.. (the most recent daily percentage change). Why decrease exponentially? Example: Suppose that & is 0.90, the volatility estimated for a market variable for day rel is 1% per day, and during day n-1 the market vari This means thato2, = 0.01 =0.0001 and u2, = 0.02? 0.90,0001 + 0.1%0,0004 = 0.00013 6“sagem WWW.GFEDUNET at Relatively less data needs to be stored, We need only remember the current estimate of the variance rate and the most recent observation on the value of the market Variable. ‘Tracks volatility changes. The value of 4 governs how responsive the estimate of the daily volatility is to the most recent daily percentage change. RiskMetrics uses 4 = 0.94 for daily volat ty forecasting. 2.5.4 GARCHU, 1) Model InGARCH (I, 1), 6? is calculated from a long-run average variance rate, V,, as well as from a, ,and 1,.. . The equation for GARCH (1, 1) is EWMA model is a particular case of GARCH(1, 1), where y=0, a 2B X ‘The “(1,1)” in GARCH(1,1) indicates that ois based on the most recent observation of u? and the most recent estimate of the variance rate. Setting c= 7¥,, the GARCH(1, 1) model can also be written + Bor. lea w+ out ek a A ncf8 477 WWW.GFEDUNET eu 8 + Ne Example: ‘Suppose that a GARCH(I, 1) model is estimated from daily data as +0.8602, 2 = 0.000002 4.0.13; this corresponds to « =0.13, = 0.86, and « = 0.000002Because y=1-a~fit follows that y= 0.01. Because o = yV, ,it follows that V;, = 0.0002 In other words, the long-run average variance per day implied by the model is 0.0002. This corresponds to a volatility of {0.0002 = 0.014, or 1.4%, per day. Suppose that the estimate of volatility on day n-1 is 1.6% per day, and that on day ‘n-l the market variable decreased by 1%. Then 02 = 0,000002 + 0.13 x 0.01? + 0.86% 0.016? = 0.00023516 the new estimate of the volatility is therefore ~{0.00023516 = 0.0153, per day. Maximum likelihood methods Choosing values for the parameters that maximize the chance (or likelihood) of the data occurring, Example: ‘Suppose that we sample 10 stocks at random on a certain day and find that the price of one of them same or increased, What is our best estimate of the probability of a price decline? : maximize: p(1- p)’ > p=0.1 8 ouSef bket WWW.GFEDUNET “ie «Bik = 2.5.5 Using GARCH(1,1)To Forecast Future Volatility o, = (a= BY, tant, + Bon, oy -¥, = au, ~Vi)+ BOM) Four Vi, = Ath s Vi) + BF) gf Fl i) = Fis Blo -Vi1=(a+ B) Ky {using this equation repeatedly Elon, ~¥ I= (a+ B) (on -V,) ol 1=V, ax Bye -V)| Exampl In the yen-dollar exchange rate example considered earliera+ f= 0.9602 and V, =0,00004422. Suppose that our estimate of the current variance rate per day is 0.00006. In 10 days the expected variance rate is 0.00004422 + 0.9602" (0.00006 -0,00004422} = 0.00005473 The expected volatility per day is 0.74%, still well above the long-term Volatility of 0.665% per day. However, the expected variance rate in 100 days is 0,00004422 + 0.9602!” (0.00006 — 0.00004422) = 0.00004449 and the expected volatility per day is 0.667%, very close to the long-term volatility. Estimating Correlations Pa [Sun For EWMA model 200%.) I= A), Yo For GARCH(I,1) mode! ew HB %0 2.6.2.2 Peaks-Over-Threshold (POT) Models ‘The models are in the more modem group. These are based on a mathematical result hat extreme realizations above a high (or below a Jow) threshold are described by a particular distribution, the generalized Pareto distribution (GPD). The GPD has three parameters, uis the threshold, & isa shape parameter that determines the fatness of the tail and B is an additional scaling parameter. Values of E>0, indicating a fat tail, are assumed of this process, While @ normal distribution exhibits 2 continuously decreasing curve from the peak to the tail, the GPD exhibits a curve that dips below the normal distribution prior to the tail. It Swen DRA aa ES g:00H WWW.GFEDUNET “ie = then moves above the normal distribution until it reaches the extreme tail. The GPD then provides a linear approximation of the tail, which more closely matches empirical data than the normal distribution. With knowledge of distributional properties of the tail, VAR can be computed at high confidence levels, regardless of the distribution describing the data, 2 eke21H Ww oreDUNET ee Bie HL 3, Financial Markets and Products + Clearing house mechanisms, structural hubs, exchanges + Netting, collateral and downgrade triggers + Futures, forwards, swaps, and options + Derivatives on fixed ~ income securities, interest rates, foreign exchange, equities, and commodities * Measuring portfolio exposures + American options, effects of dividends, early exercise + Trading strategies with derivatives + Minimum variance hedge ratio + Cheapest to deliver bond, conversion factors * Commodity derivatives, cost of carry, lease rate, convenience yield + Basis risk + Foreign exchange risk + Corporate bonds + Debt equity swaps, loan sales 3.1 The Mechanisms of Market Management 3.1.1 Exchange and clearinghouses reduce counterparty risk mechanisms + Pay ipant standards Set standards for membership and approve new members. ‘+ Margins set requirements that reflect the volatility of the traded instrument. and ‘monitor the creditworthiness of participators. + Netting Setting off payments in one direction against those in the opposite directions. + Collateralization When each participant’s credit exceeds a certain given limit, in order to advance business, some forms of reassurances are needed. + Marking to market Using market price at the close of the business day or intraday ek OT ATO N, ®12077 WWW.GFEDUNET eee HL to determine the value of each instrument participant hold, to decide whether margins and collateral calls are needed + Downgrade triggers Occurs when the credit of one party is downgrade beyond to a certain point, that party must post collateral in order to continue trading. 3.1.2 Mechanism for dealing with sovereign risk exposure + Debt-for-equity swaps For example, The investment bank had loans to the foreign sponsoring country’s government at $100, and the bank sell those loans on the secondary market to a market maker at $92, then a corporation buy the loan from this market maker for $94 to finance its investment in the sponsoring country, and the purposed government set convert exchange rate lower than the free market. in the end, three parties except the original bank share the discount 8% from face value accepted by the bank. + Multiyear restructuring of loans (MYRAS) If a country is unable to keep its payment 9 on a loan current and an FI choose to maintain the loan, some amount an FI reed to concede or give up to the borrower. several factors do matter in the loan rescheduling process, such as fee, interest rat, grace period, maturity, option and guarantee features, the net cost or degree of concessionality is equal to Concessionality =[prenent value of original loan|-[ prenent value of restructured loan] The lower the present value of the restructured loan, the greater the cost of loan restructuring. * Sale of LDC loans on the secondary market Benefit: removal loans from the balance sheet and free resource rest of balance sheet can endure the cost provide a tax write-off for the lender disadvantage: the tax adjusted difference. + Bond-for loan swaps (Brady bonds) Transform an LDC loan into a highly marketable and liquid instrument ~ bond. 3 SW ELPA,22754 WWW.GFEDUNET. le = St = tL 3.2 The Properties of Financial Products 3.2.1 Hedging Strategies Using Futures 32.1.1 Short hedge and long hedge + Short hedge occurs when the hedger shorts (sets) a futures contract to hedge against a price decrease in the existing long position, A short hedge is appropriate when you ave a long position and expect prices to decline. + Long hedge occurs when the hedger buys a futures contract to hedge against a price increase in the existing short position. A long hedge is appropriate when you have a short position and expect prices to rise. 32.12 Basis Basis sisk is the risk that a difference may occur between the spot price of a hedged asset, and the futures price of the contract used to implement the hedge. Basis risk is zero only when there is a perfect match between the hedged asset and the contraet’s underlying in terms of maturity and asset type. Basis = spot price of asset being hedged — futures price of contract used in hedge Variance of the basis, OS, FO) = OMS, ) +E (0) 20S, oF" (9) where p is the correlation coefficient between the futures and spot price series. Classical ofthe effectiveness of hedging a spot position with futures contracts o(basis) a(S) 3.2.1.3 Sources of basis risk ‘Three sources of basis risk are: Interruption in the convergence of the futures and spot prices Changes in the cost of carry Imperfect matching between the cash asset and the hedge asset, wk 35‘S88 HT WWW.GREDUNET Ak Se ’ 3.2.14 Optimal Hedge Ratio A hedge ratio is the ratio of the size of the futures position relative to the spot position. ‘The optimal hedge ratio, which minimizes the variance of the combined hedge position, is defined as follows: , Which is also the beta of spot prices with respect of futures contract prices since Coy, Coy, 5 and CONSE Op Tsp Op Where: Ps,p =the correlation between the spot prices and the futures prices =the standard deviation of the spot price oy =the standard deviation of the futures price 3.2.1.5. Hedging with Stock Index Futures When hedging of equity portfolios using futures contracts on stock indices, the number of ‘contracts requires to completely hedge an equity position is: portfolio value value of futures contract Pratt { portfoliovalue ‘ Sutures pricex contract multiplier } roi of tga 32.1.6 Adjusting the Portfolio Beta B is our portfolio beta f° is target beta after we implement the strategy with index futures, P is portfolio value, A is the value of the underlying asset. The appropriate number of futures is: Namberofcontracts= (f° — A) Example: An equity portfolio is worth $100 million with the benchmark of the Dow Jones Industriat Average. The Dow is currently at 10,060, and the corresponding portfolio beta is 1.2. The futures multiplies for the Dow is 10. Which of the following is the closest to the number of contracts needed to double the portfolio beta? % Eiki ATA.FU fS WWW.GREDUNET eu ae we A. 1,100 B. 1,168 Cc. 1188 D. 1,200 Answer: D Ct. FE 2C000)1,200 3.2.1.7 Rolling a Hedge Forward ‘When the hedging horizon is longer that the maturity of the futures, the hedge must be rolled forward to retain the hedge. This exposes the hedger to rollover risk, the basis risk when the hedge is re-established. 3.2.2. Determination of forward and futures prices 3.2.2.1. Investment and consumption assets ‘An investment asset is an asset that is held for the purpose of investing, A.consumption asset is an asset that is held for the purpose of consumption. 3.2.22. Short selling and short squeeze Short sales are orders to sell securities that the seller dose not own. The short seller may be forced to close their position if the broker runs out of securities to borrow. This is known as a short squeeze and the seller will need to close their short position irumediately. 3.2.2.3 Future contracts and forward contracts Future contracts and forward contracts are similar in that both: * Can be either deliverable or cash settlement contracts. + Are priced to have zero value at the time an investor enters into the contract. + Future contracts differ from forward contracts in the following ways: ‘+ Future contracts trade on organized exchanges. Forwards are private contracts and do not trade on an exchange. + Future contracts are highly standardized. Forwards are customized contracts satisfying the needs of the parties involved. ob eT. 872807 WWWGFEDUNET ek ue si + A single clearinghouse is the counterparty to all futures contracts. Forwards are contracts with the originating counterparty. * The government regulates futures markets. Forward contracts are usually not regulated. 3.224 Cost-of-Carry Model ‘The cost-of-carry model is used to price forward and futures contracts. It states that the total cost of carrying the underlying asset to expiration must be the futures price. Any other price results in arbitrage opportunity The futures price or cost-of-carry model is easily accommodated for interim cash flows from the underlying asset. Forward price: Fos," F,=(S,-De™ —_Lrepresents the present value of the cash flows over T years F sero , oS. 4 represents the continuously compounded dividend yield 3.2.2.5 Value of a Forward Contract The value of the long contract on an asset with no cash Nows is computed as, ~ Ke"; with cash flows (with present value I) it isS,—/~Ke"7; and with a cont dividend yield of g, itisS,e — Ke” | The curent value of an outstanding forward comract can_be found. by entering an offsetting forward position and discounting the net cash flow at expiration. By definition, the value of a futures contract is zero at inception. Abas hRAcfu WWWGFEDUNET fee eu 3.2.3 Pricing Commodity Forwards and Futures 323.1 Basic Equilibrium Formula for Pricing Commodity Forwards and Futures ‘The time T forward price is discounted at the risk-free rate back to time 0, this is the present value of commodity received at time T, Commodity and financial forward contracts are similar in some regards. The price of a commodity forward must be based upon expectations, but there are several factors to consider, such as storage cost, lease rate and convenience yield. So the commodity forward price today is defined as a biased estimate of the expected spot commodity price at time T as follows: F,; = E(S,)e""" .The bias is due to the risk premium on the commodity, r—ar 323.2 Lease Rate ‘The lease rates are defined as the amount of interest that a leader of commodity requires. In other words, it is the return that makes an investor willing to buy and then lend a commodity. If an active lease market exists for a commodity, a commodity lender can eam the lease rate by buying a commodity and immediately selling it in -the forward market. 3.23.3 Calculating Lease Rate for Commodity Forwards and Futures The commodity forward price for time T with an active lease market is expressed as: =S.e-ar Rap = S04 3.2.3.4 Determine the Contango or Backwardation of Forward Market Using Lease Rate ‘A market is said to be in contango when the futures price trades at a premium relative to the spot price while a market is said to be in backwardation (or inverted) when forward prices trade at a discount relative to spot prices. The commodity market is in contango with an upward-sloping forward curve when the lease rate is less than the risk-free rate. ‘The market is in backwardation with a downward-sloping forward curve when the lease rate is greater than the risk-free rate. Key concept; Markets are in contango if spot prices are lower than forward prices. Markets are in backwardation iffspot prices are higher than forward prices. Backwardation occurs wien there is higher than forward prices. Backwardation occurs when there is high current demand for the commodity, which implied high convenience yields.SBC WWWGFEDUNET eu Se 8 3.2.3.5 Calculate the Forward Price of Commodity with Storage Costs A commodity owner will only store the commodity if the forward pric equal to the spot price plus the future storage costs as follows: F,, >S,e" + 4(0,T), is greater than or where (0,T) represents the future value of storage costs for one unit of the commodity from time 0 to T If storage costs are paid continuously and are proportional to the value of the commodity, the no-arbitrage forward price becomes F,; = S,e""4" A=continuous anmual storage cost proportional to the vale of the commodity 3.2.3.6 The No-Arbitrage Bounds for the Forward Price of Commodity when Considering a Convenience Yield If the owners of the commodity need the commodity for their business, holding physical inventory of the commodity creates value. Holding an excess amount of a commodity for a non-monetary retum is referred to as convenience yield. The forward price including a convenience yield is calculated as F,;2S,e"**" , where ¢ is the continuously compounded convenience yield, proportional to the value of the commodity. Because a convenience yield can not be eamed by the average investor who does not have a business reason for holding the commodity, the forward price including a convenience yield is able to be created in a range of no-arbitrage prices as follows: sel chy, Sel Combinations of costs and benefits, be sure to increase the exponent for costs and reduce for benefits, the price of commodity forward is: For eer ‘Example: Consider a 6-month futures contract on the S&P 500, and suppose the current value of the index is 1330. Suppose the dividend yield is 1.5% annually for the stocks underlying the index, and that the continuously compounded risk-free interest rate is 5.5% annually. What is the cost of carry for this futures contract? A. 4.0% B. 4.0% C. 2.0% D. -2.0% 90 SURG TATRAfet) WWW GFEDUNET, Bub 8b at ‘Answer: A 3.2.3.7 The Factors that Impact the Pricing of Gold, Com, Natural Gas and Oil Futures Gold, com, natural gas, and oil are all examples of commodities with Characteristics that differ with respect to storage costs, the ability to store, production costs, and seasonal demand. These unique differences influence the commodity forward prices and the shape of the forward curves. Gold forward price Gold can ea a retum by being loaned out, when a positive lease rate is present, the synthetic gold is preferred to physically holding the gold. PV of gold production= »”, Fa, — x(t," ‘at Where mount of ounces of gold we expect to extract, with an extraction cost of x(,) Corn forward Com is produced at one be consumed when 1¢ of the year, but consumed throughout the year. In order to is not being produced, com must be stored. So the storage cost must be concemed. Between the harvests, the forward price of corn rises to reward storage, and it falls at each harvest. Natural gas forward Natural gas is another market in which seasonality and storage costs are important. ‘Natural gas has several characteristics. First, gas is costly to transport internationally, so prices and forward curves vary regionally. Second, once a given well has began production, gas is costly to store. Third, demand is highly seasonal, with peak demand arising from heating in winter months. Gas has constant supply and seasonal demand, Oil forward Oil is easier to store than gas. Thus, seasonal in the price of crude oil are relatively unimportant. The long run forward price is less volatility than the short run forward price. increase since supply is fixed. In the long run, both supply and demand have time to adjust to price changes with result that price movements expect the price attenuated, In the short run, an increase in demand will cause a Example:S877 WWW.GREDUNET “lk + ik Which of the following best describes what we would normally expect to see in a seasonal agricultural market like wheat? Assume “the harvest” is normal and not unusually big or unusually small. Now consider the following statements about the market. 1. Prices fall at the harvest and rise after the harvest. Il. Prices are constant on average across the year regardless of seasonality. IIL. Prices rise at the harvest and fall afterwards TV. The market is in contango when the harvest comes in, V. The market is in backwardation when the harvest comes in. VI. If the market goes into contango, it is most likely to do so right before a hnew harvest. VIL. If the market goes into backwardation, it is most likely to do so right before a new harvest. Now choose the letter that best describes which of the above statements is true, A. Land IV are the only true statements B. 1, IV, and VI are the only true statements C. III, V, and VII are the only true statements, D. 1,1V, and VII are the only true statements Answer: D Explanation: The new harvest ‘resets’ the storage market. For a while, consumption and produetion occur directly from the new harvest, and prices are low. Prices begin to rise as storage begins to occur. As the next harvest approaches, inventory may get tight, sending the market into backwardation, 3.2.38 Calculate a Commodity Spread Some commodity is the inputs in the creation of other commodities, which gives rise to commodity spreads. For example Soybeans can-be crushed to produce soybean meal and soybean oil. A trader with a position in soybeans, for example long position, and an ‘opposite position, short position in equivalent quantities of soybean meal and soybean oil has a crush spread. Similarly, crude oil is refined to make petroleum product such as gasoline, kerosene, and heating oil. The split of oil into these different components can be complemented by a 2 SR HHAS877 WWW.GFEDUNET uh Sk 8 process known as “cracking”, the difference in price between crude oil and equivalent amounts of heating oil and gasoline is called the crack spread. For example if3 gallons of crude oil can be split into 2 gallons of gasoline and 1 gallon of heating oil. We will speak of “3-2-1”. The refiner could use a futures crack spread to lock in both the cost of oil and the output prices. But because there are other inputs to production and it is possible to produce other outputs, so the crack spread is not prefect hedge. 3.23.9 The Differential between a Strip Hedge and a Stack Hedge We engage in a strip hedge when we hedge a stream of obligations by offsetting each individual obligation with a futures contract matching the maturity and quantity of the obligation. With a stack hedge, we enter into futures contracts with a single maturity, with the number of contracts selected so that changes in the present value of the future obligations are offset by changes in the value of this stack of futures contracts. There are two reasons using a stack hedge. First there is often more trading volume and liquidity in near-term contracts. With many commodities, bid-ask spreads widen with maturity. Thus a stack hedge may have fower transaction costs than a strip hedge. Second, the manager may wish to speculate on the shape of the forward curve. You might decide that the forward curve looks unusually steep in the early months. If you undertake a stack hedge and the forward curve then flattens, you will locked in all your oil at the relatively cheap near-term price, and implicitly made gains from not having locked in the relatively high strip prices. However, ifthe curve becomes steeper, itis possible to lose Example Imagine a stack-and-roll hedge of monthly commodity deliveries that you continue for the next five years. Assume the hedge ratio is adjusted to take into effect the mistiming of cash flows but is not adjusted for the basis risk of the hedge. In which of the following situations is your calendar basis risk likely to be greatest? A, Stack and roll in the front month in oil futures B, Stack and roll in the 12-month contract in natural gas futures C. Stack and roll in the 3-year contract in gold futures . Al four situations will have the same basis risk Answer: A The Explanatio term structure is highly volatile at the short end, making a ARE LIRA, 3S84 07T WWW.GPEDUNET pik UE front-month stack-and roll hedge heavily exposed to basis fluctuations. In natural gas, much of the movement occurs at the front end, as well, so the 12-month contract won't move as much. In gold, the term structure rarely moves much at all and won't begin to compare with oil and gas. 3.2.4 Commodity Spot and Futures Markets 3.2.4.1. Major risks in commodity spot transaction 4 price risk 4 transportation risk A delivery risk A credit risk 3.2.4.2. Similarities and differences between the fundamental types of transactions Spot trading 4 commercial contract 4. flexible convenants 4 juridical commitments of the buyer and seller until execution of the contract, 4 long transaction 4 illiquid and discontinuous market > allow the transfer of goods in conditions suiting the demand Forward contracts 4 bilateral agreement 4 flexible convenants 4 replace spot transactions on many occasions 4 form of contracting totally appropriate for commodities A. credit risk fully present 4 flexibility regarding the optimal transfer of goods Futures contracts 4. standardized instrument A necessity of a physical delivery or termination of the posit ion before maturity 4 buyer and seller only refer to the clearing house 4 SW eH.Sete WWWGFEDUNET eu = Bide 3 4 central clearing mechanism generating market prices 4 price transparency 4 liquidity 4 low transaction cost 3.2.4.3. Spot and forward freight markets The Baltic International Freight Future Exchange offered freight futures contracts. it exhibit that forward market leads the spot market, and can be used as price discovery vehicles. Liquidity may be measured by the sizeof the trade it takes to move the market. Market depth may be measured by the time it takes for an order of a standard size to be executed. 3.2.5 Interest rate futures 3.2.5.1 Day count conventions Day count conventions play a role when computing the interest that accrues on a fixed income security. The bond buyer must pay any accrued interest eared through the settlement date. n Accrued interest = MP", n= number of days from last coupon to the settlement date m = number of days in coupon period. Three commonly used day count conventions US Treasury bonds use actual / actual. US corporate bonds and municipal bonds used are 30 / 360. US money market instruments (Treasury bills) use actual / 360. 3.2.5.2 Cheapest-to-Deliver Bond and Conversion Factor U.S Treasury Bond (I-Bond) Futures Contract Conversion Factor ‘The conversion factor defines the price received by the short position of the contract Specifically, the cash received by the short position is computed as follows: euko rime 9sPRA WWW.GFEDUNET Ge + oe = HL Cash received = (QFPxCF)+ AI Where: QFP=quoted futures price cl “ max(Xe” ~S,,0) 108 ewe Ee HeS8t0F WWW.GFEDUNET elk Bk Call options cannot be worth more than the underlying security, and put options cannot be worth more than the strike price. When the stock does not pay a dividend, European call options cannot be worth less than the difference between the current stock price and the present value of the strike price. European put options cannot be worth less than the difference between the present value of the strike price and the current stock price. Put-call parity is a no-arbitrage relationship for European-style options with the same characteristics. It states that a portfolio consisting of a call option and a zero-coupon bond with a face value equal to the strike must have the same value as a portfolio consisting of the corresponding put option and the stock: c+ Xe” = p45, Key concept: ‘A long position in an asser is cutie to a long position in a European call with a short position in an otherwise identical put, combined with a risk-free position. f Cz ex max(S,- Xe"",0) It is never optimal to exercise an American call option on non-dividend-paying stock Prior to expiration should never be exercised early. If the asset pays income; early exercise may occur, with a probability that inereases with the size of income payment. ~~ : ‘American put options on non-dividend-paying stocks can be optimally exercised prior to expiration if the put is sufficiently in-the-money. P p> max(S,—Xe",0) Key concept: ‘An American put option on a non-dividend-paying stock (or asset with no income) may be exercised early. If the’ asset pays income, the possibility of early exercise decreases with the size of the income payments, Call options are always worth more than corresponding put options prior to expiration when both are at-the-money. SWAT PT HA 109sef8 2605 WWW.GFEDUNET ome NL 3.2.8.3 Computing Option Value Using Put-Call Parity Put-cal! parity is no-arbitrage relationship for European-style options with the same characteristics. It states that a portfolio consisting of a call option and a zero-coupon bond with a face value equal to the strike must have the same value as a portfolio consisting of the corresponding put option and the stock: e+ Xe” = p+, European and American Option Putcall parity only holds for European options. For American options we have an inequality: S,-X $C-P D+Xe"-S, All else equal, the payment of a dividend will reduce the lower pricing bound for a call option. The payment of a dividend will increase the lower pricing bound for a put option. With dividends, the put~call parity becomes p+5S,=c+D+Xe7 With dividend, relationship between American call and put options is modified as follows: S,-X-DSC-PsS,-Xe" Example ‘An American investor holds a portfolio of French stocks. The market value of the portfolio is €10 million, with a beta of 1.35 relative to the CAC index. In November, the spot value of the CAC index is 4,750. The exchange rate is USD 1.25/€. The dividend yield, euro interest rates, and dollar interest rates are all equal to 4%, Which of the following option strategies would be most appropriate to protect the portfolio against a decline of the euro that week? March Euro options (all prices in US dollars peré) Strike Call euro Put euro 10 seule 6 #0 A280K WWW.GFEDU NET eae Sse HEL 125 0018 0.022 Buy calls with a premium of USD 180,000 Buy puts with a premium of USD 220,000 Sell calls with a premium of USD 180,000 Sell puts with a premium of USD 220,000 pop > Answer: B A. Incorrect. This would not protect against a decline in the euro and would rather provide upside in case of appreciation of the euro. B. Correct. This would protect against a decline in the euro and the premium would be USD.022 x€10 million = USD220, 000. C. Incorrect. This would not protect against a decline in the euro and would rather make the Investor subject to (theoretically) unlimited losses (writing naked calls); the amount of premium is also incorrect and should be USD.018 x €10 million = USDI80, 000. D. Incovrect. This would not protect against @ decline in the euro and would rather protect against a decline in the US dollar; the amount of premium is also incorrect, and should be USD .022 x €10 million = USD 220,000. Seeceeeeceeceeee 3.2.9 Option Strategies 3.2.9.1 Covered Call or Protective Put Strategy Stock options can be combined with their underlying stock to generate various payoff profiles. When an at-the-money long put position is combined with the underlying stock, we have created a protective put strategy. Protective Put Strategy SwkOn ATMA u22188077 WWW.GFEDUNET Pur Option Another common strategy is to sell a call option on a stock that is owned by the option writer. This is called a covered call position, 3.2.9.2 Spread Strategy Spread strategies combine options in the same option class to generate various payoff profiles. Spread strategies include bull, bear, butterfly, and calendar spreads. ‘The buyer of a bull call spread expects the stock price to rise and the purchased call to finish in-the- money. However, the buyer does not believe that the price of the stock will rise above the exercise price for the out-of-the-money written call. Itcan be created by * buying a call option on a stock with a certain strike price and + selling a call option on the same stock with a higher strike price. Both options have the same expiration date Bull spreads can also be created by buying a put with a low strike price and selling a put with a high strike price. Buill spread created using call options m2 WA APTN A,F897 WWW.GFEDUNET. ok Bok tt Bull spread created using put options Payoff from a bull spread Stock price Payoff from | Payoff from Total payoff range Jong call option | short call option S,2K; S,-K, K,-K, K,K Sr-k 0 Sr~K similar to a straddle except that the A strangle (or bottom vertical combination) i option purchased is slightly out-of-the-money, so it is cheaper to implement than the straddle. Astrangle 19Sf WWW.GFEDUNET dee ie Mf Stock Price Payoff from a Strangle Range of Payoff from Payoff from Total stock price eal put payell HSK 0 Ki <5 y= NM) sew 37fee WWW.GFEDUNET eu ae im Sumietive stand normal 4 The Bootstrap A altemative to generating random numbers from a hypothetical distribution is to Sample from historical data. §,,, = 5,(1+ Ry) Advantage of bootstrap: Can include fat tails, jumps, or any departure from the normal distribution 's of the simultaneous Account for correlations across series because one draw consi retums for N series, such as stock, bonds, and currency prices. Limitation of bootstrap: For small sample sizes, may be a poor approximation of the actual one Relies heavily on the assumption that returns are independent. 4.1.4.8 Comparison of these methods ‘The delta-normal method 4» Advantages of the delta-normal method are that it is fast and easy to use, and enables analysis. 4 Disadvantages include a higher proportion of distributions with fat tails and an inability to account for nonlinear relationships. Historical approach + Advantages: Easy to use if enough data exists, full valuation of a portfolio is based on actual prices. Historical simulation also avoids model risk and uses correlations provided by historical data. 4 Disadvantages often a from the fact that sufficient historical data frequently 138techs WWW.GFEDUNET le Bit does not exist, and since this historical data assumes only one path of events, fixed volatility and correlation estimates are the result, Monte Carlo simulation method Monte Carlo situation uses a great number of scenarios with values randomly drawn from the output of a stochastic process. 4 Advantages: The most powerful model and can account for both linear and nonlinear risks. It can include variations in risk and correlations and can provide a nearly untimited number of scenarios. 4 Disadvantages: Expensive, complicated, and subject to model risk and sampling variation, wa nea Sao rors we a Re ae ws ca we sett x wit a onset ite aH te wat reels aR mt erm Es 2 a aR Ante nae memeowe ARES aR ieee * a * ewe epee z * ornate em #8 we vane war vee eset mae WAH WR URANO RO eur 4.1.5 Risk Budget 4.1.5.1 Definition In the world of risk management, these same three elements of control—planning, budgeting, and monitoring. Risk budgeting is a top-down process that involves choosing and managing exposures to risk, 4.1.5.2 Use VaR to Design Better Investment Guidelines and for the Investment Process VaR techniques can help move away form the ad hoc nature and overemphasis on SuKOn OR HBA 13921a WWW.GFEDUNET poe + ie + notionals and sensitivities that characterize the guidelines many managers now use. Such guidelines are cumbersome and ineffective in that they focus on individual positions and can be easily circumvented, ‘VaR is useful for the investment process. When a trader has a choice between two new positions for a portfolio, the trader can compare the marginal VaRs to make the selection. Wher deciding whether to increase one existing position over another, the trader can compare the return-to-VaR ratios and increase the position in the one with the higher ratio. 4.1.5.3. Budget Risk across Asset Classes Budgeting risk across asset classes means selecting assets whose combined VaRs are less than the total allowed. The budgeting process would examine the contribution each position makes to the portfolio VaR. 4.1.54. Budget Risk across Active Managers For allocating across active managers, if the tracking errors of the managers are independent of each other, it can be shown that the optimal allocation is achieved with the following formula: I x(portfolio's tracking error vlatliy) op g TR, x(manager’s tracking error volatility) weight of portfolio managed by manager given group of active managers, the weights may not sum to one. The remainder of the ‘weight can be allocated to the benchmark, which has no tracking error. 4.1.5.5. Performance measurement tools Performance measurement tools 4 Tool #1—The Green Zone 4 Tool #2—Attribution of Retums 4 Tool #3—The Sharpe and Information Ratios 4 Tool #4—Aipha versus the Benchmark 4 Tool #5—Alpha versus the Peer Group 4 To determine whether a manager generates consistent excess risk~sdjusted performance vis a vis a benchmark. 4 To determine whether a manager generates superior risk-adjusted performance vis a vis the peer group. 140 we EDT HESef 4KFT WWW.GFEDUNET ak Sis 4 To determine whether the returns achieved are sufficient to compensate for the risk assumed in cost/benefit terms. 4 To provide a basis for identifying those managers whose processes generate high-quality excess risk-adjusted returns. We believe that consistently superior risk-adjusted performance results suggest that a manager’s processes, and the resulting performance, can be replicated in the future, making the returns high-quality. 4.1.5.6 Risk Budgeting for Pension Funds and Investment Managers Using VaR ‘VaR Definition: “One year 84% confidence VaR” and “Tracking Error” What are the key market risk for pension funds ? 4 Defined benefit plans: Surplus risk : the chance that the assets in portfolio might underperform the pension liabilities it owes to its staff, causing the sponsor of the pension plan to have to unexpectedly contribute funds to make up the shortfall ‘Tracking error to planwide asset allocation 4 Defined contribution plans/money purchase schemes Inappropriate asset atfocation Rogue manager What are the key market risks for an asset management firm? A Variable fee income A Customer satisfaction; product integrity What is “risk budgeting” for an investor? For simply , “tisk budgeting” is the process of allocating an allowable measure of potential loss to different aspects of the investment process, monitoring whether those pieces of the investment process have exceeded their measure, taking corrective action (if deemed necessary) when a measure is exceeded, and using the risk measurement process to evaluate risk-adjusted return. The first step of risk budgeting is to determing which parts of the investment process need monitoring in this fashion, and set a risk tolerance level. 4. What to monitor ? Surplus at risk (SAR): the amount by which the pension’ policy asset allocation ewe 1 OH m1Sf 807 WWW.GFEDUNET ‘esk + iok = mE might underperform its pension liabilities, over a given time horizon(ie, one year) at a given confidence interval (ie, 95%). Implementation risk or tactical asset allocation risk : the degree to which the plan’s tactical asset allocation might underperform,its strategic asset allocation over, for example, one year at 95% confidence. Active risk ,planwide: the amount by which the actual assets in which the plan has invested, across all its portfolios ,covid underperform its tactical asset allocation, usually in on year at 84%confidence(this is the same as tracking error ,for te whole plan). ‘Active risk per manager: the amount by which a given manager might underperform their benchmark within the tactical asset allocation, usually at one year,84% confidence (eg, tracking error), 4. Setting risk tolerance thresholds Once an entity has decided which measures to monitor, the next step is to establish risk tolerance levels. This is more difficult: if the level is set too tightly, it climinates the chance of outperformance and may make the fund performan less well than imore aggressive peer funds. If set too loosely, large-losses will be possible before the measure ever gives a signal that a corrective action is needed. 4 Trade-offs and peer comparisons 4 A reminder of how this is different from asset allocation Risk budgeting has two important qualities which differentiate it from asset allocation: downstreaming and dynamic triggers, 4 Maintaining a quanlity VaR mearue 4 Taking corrective action Surplus at risk: 1. Elevate the information to the board; 2. If'decision makers agree that the change which has triggered the risk threshold is persistent and important, shift the strategic asset allocation to reduce the potential threat to surplus. 3. If the change is thought to be technical or temporary, take no action. If the change persists beyond a given time period, the item should be raised again. Implementation risk 1. Elevate the information to the investment committee. we eas#280 WWW.GFEDUNET ee ge 2. If decision makers agree that the gap between the tactical and strategic asset allocations has widened beyond their comfort zone, bring the tactical asset allocation closer to the strategic. 3. Ifthe change is thought to be technical or temporary, take no action but continue to monitor its severity and revisit if it does not self correct. Active risk ,plan wide 1. Elevate the information to the investment committee or a responsible individual 2. Determine the drivers of the off-benchmark risk. Active risk, per manager 1. Elevate the informatic individual . to the investment committee or a responsible 2. Determine the drivers of the off-benchmark risk. 3. Understand manager's beliefs and intentions, 4, Decide whether the historical dataset which gave rise to the VaR signal should be ignored in favor of a forecast; the investment committee could decide to accept the portfolio and raise the Active Risk badger or employ other measures ‘to monitor the off-benchmark strategy. 5. The benchmark could be changed to one which better accommodates the manager’s strategy. 6. If the off-benchmark risk is thought to be unacceptable, the investment committee could: 7. request a shift in the manager’s portfolio to reduce active risk to an acceptable level; 8. replace the manager; 9. put in place an overlay to reduce the off-benchmark position without disturbing the manager's portfolio. A Evaluating risk-adjusted performance Risk budgeting versus asset allocation 4 Leverage : VaR and risk sensitivities are usefisl measures in monitoring leverage. 4 Derivatives, foreign exchange: Risk budgeting versus standard deviation Risk budgeting versus beta M3P88 WWW.GFEDUNET ek uk Beta expresses the relative volatility between a portfolio and “the market”, usually an equity benchmark, ‘VaR differs mainly in that it instead computes tracking error, which is related to beta but not identical. ‘Tracking error measures the degree of potential underperformance between a portfolio and its specific benchmark, while beta is often assumed to use a single benchmark to represent the market. Tracking error is extensible across equity, fixed income, balanced, domestic and international portfolios, where there is no single benchmark for the market. Risk budgeting versus duration Duration sensitivity shows how sensitive a fixed income portfolio is to a 0.01% interest rates across the yield curve. rat ‘VaR blends duration information with the volatilities of the various segments of the market and various countries, and their correlation to one another, to weight the potential loss in a portfolio where non-parallel shifts may occur. Controlling liquidity, credit, concentration risk ‘Using back testing to calibrate the VaR model 4.1.6 Stress Testing 4.1.6.1 Role of Stress testing Stress testing focuses on the infrequent but large scale events that occurring in the left tail Of the return distribution. The use of stress testing addressed the shortcoming in VaR. It is apparent that stress testing should be used as a complement to VaR measures, rather than as a substitute 4.1.6.2. Primary Approaches to Stress Testing Stress-testing is a key risk management process, which includes * scenario analysis + stressing models, volatilities and correlations, + developing policy responses 4.1.63. Scenario Analysis Unidimensional Scenario Analysis Unidimensional scenario analysis identifies key risk factors, shocks the factor by a lange 4 tbe resoneff WWW.GFEDUNET Sh + BR HE amount, and measures the impact on portfolio value (revalue the portfolio) ‘Key concept; Unidimensional scenario analysis does not consider correlation across. multiple risk factors. Multidimensional Scenario Analysis Multidimensional analysis incorporates correlation across risk factors, but increases the complexity of the analysis. It can take two general forms: historical or prospective. ‘Prospective Scenarios Prospective scenarios analysis is either factor-pushing or conditional. Prospective scenarios are hypothetical based on reasonable and relevant scenarios that could generate large losses. Factor push shifts each variable in the direction that would adversely impact the portfolio. Conditional Scenario The conditional scenario method incorporates correlations across a subset of key risk factors correlations may not hold during a hectic time period. i Historical Scenarios It is backward looking, while the prospective approach is forward looking. Historical scenarios will examine previous market data to infer the joint movernent of key financial variables during times of market stress. The obvious limitation is the limited number and unique features of each event. Worst Case Scenario Measare as an Extension to VaR ‘The Worst Case Scenario (WCS) focus on the distribution of the loss during the worst trading period (one day or two weeks) over a given horizon (100 days or one year). WCS extends VaR risk measurement estimating the extent of the loss given an unfavorable event Example Which of the following is true about stress testing? SKE RTA 1458ACHT WWW.GFEDUNET ae Bi = ‘A. It is used to evaluate the potential impact on portfolio values of unlikely, although plausible, events or movements in a set of financial variables. B. It is a risk-management tool that directly compares predicted results to observed actual results. Predicted values are also compared with historical data. C. Both ‘a’ and ‘b’ above are true D. None of the above is true Answer: A A. Correct. It describes ‘stress testing’. B, Incorrect. It is not about ‘stress testing”, C. Incorrect. As *b’ is incorrect. D. Incorrect. As ‘a’ is correct. 4.1.6.4 Large Stress Losses Despite best efforts, the magnitude of loss from stress testing may still be unacceptably large. The possible responses: 4 Buy protection through insurance contracts, credit default swaps, and other derivatives. A Modify portfolio to decrease exposure or ersify. A Alter business strategy; restructure business lines or product mix. + Develop contingency plan in case of trigger event. ss cure alternative funding in liquidity stress. 4.2 Credit Risk 4.2.1 The Rating Agencies 42.1.1 Agencies around the world Rating agencies specialize in evaluating the creditworthiness of debt securities issued by corporate, financial, structured finance, municipal, and sovereign obligors, and by M6 ey HA,te WWW.GFEDUNET ek se evaluating the general creditworthiness of the issuers themselves. In other words, what is, the probability of repayment? Ratings agencies have assumed enormous importance in the management of credit risk For borrowers credit ratings are critical because they affect their access to markets and the cost of their borrowings. Increasingly regulators have designated in the United States as nationally recognized statistical rating organizations (NRSROs) in evaluating the quality of loan and investment portfolios and the equity capital needed to support the risk in these portfolios. In the Basel II Agreement, finalized in June 2004, the Basel Committee of the Bank for International Settlements (BIS), whose membership consists of the major central banks in the world, has also raised the profile of the agencies, described as extemal credit assesament institutions (ECAIs), again highlighting their role in providing the basis for capital adequacy calculations. The three major US. rating agencies are Moody’s Investors Service, Standard and Poor’s (S&P), and Fitch Ratings. Over time the agencies innovated with their business model and in the 1970s switched their revenue base from subscription-paying investors to fee-paying issuers. They also greatly broadened their services to cover the entire spectrum of instruments and obligors, including asset-backed securities, commercial paper, municipal bonds, counterparty risk, the claims-paying ability of insurance companies, and credit risks of all kinds. The importance of the ratings agencies is reflected in the rapid increase in the number of ratings that they provide. 4.2.1.2 Ratings performance Ratings are used to communicate opinions about the creditworthiness of issuers and obligations. The rating itself contains a lot of pieces of information, including information about the probability of default and the loss severity in the event of default Moody’s indicates that they are looking at a variety of horizons depending on the maturity of the instruments being rated and the nature of the issuer. S&P explains that in rating an industrial bond it focuses on the following areas _ Business risk _ Industry characteristics _ Competitive positioning _ Management SSeKATT ATA ur2288071 WWW.GFEDUNET 4.2.1.3 Ratings and reguiations The refationship between regulators and ratings agencies is deep and often ambiguous. The regulators are attracted to the high quality, the independence, and very widespread acceptance of the rating agency opinions on credit quality. On the other hand, they are concerned about putting so much reliance on the agencies over whose activities they have no control. For the agencies, the use of their opinions by the regulators is an important validation of their work. Also, because the regulators only accept the ratings of a few of the agencies (the NRSROs in the United States and the so-called ECAls under the Basel Il Capital Requirements Directive), this creates an important competitive advantage. 42.1.4 Emerging trends ‘The more power and influence the rating agencies have acquired the more controversial they have become. By and large market participants, both issuers and investors, are ied with the products produced by the agencies. They understand clearly that agency opinions are not intended to be the sole source of information to be used in making credit decisions. They expect there will be transparency in the process so that it will be clear how ratings are derived and what considerations might lead to their being changed. They also expect greater forensic activity from the agencies, to dig into nonpublic information and be more active in questioning key aspects of an issuer’s performance. Market participants themselves use financial data and securities prices to track issuer performance but they expect that the agencies will base their views on fundamental credit measures to give a more stable view of intrinsic financial capacity. ‘The rating agencies are once again in the spotlight and facing heavy criticism for the enabling role they have played in the growth of the structured finance market (Lagard 2007). As a result of the current problems, the agencies have been actively recalibrating the default and toss assumptions in their rating models as well as their correlation assumptions for different asset classes in CDOs. The agencies are defending themselves by insisting that they have provided transparency in their rating approaches. 4.2.2 Credit Risk Measurement Default Risk 42.2.1 Credit Rating External Rating Credit rating given by those rating agencies usually has two meanings. us sabe btWen WWWGFEDUNET Ye Bie «MEL 4. First, it represents the agency's opinion about the creditworthiness of an obligator with respect to a particular debt security or other financial obligation. 4 Also, it indicates an issuer’s general creditworthiness, Generally speaking, the ratings are divided into two main categories, investment grade and speculative grade. 4 Investment grade means that the firm has adequate repayment capacity with high quality and strong financial status, 4 Speculative grade implies high risks imbedded in the debt securities as the issuer might default. Description ‘S&P Moody's Investment Grades AAA Aaa AA Aa AA BBB Baa Speculative Grade BB Ba BB ccc Caa cc Ca cc Cc D Rating Process The rating agencies usually use industry and firm-specific inputs in their rating process. The industry specific factors relate to the industry characteristics and competitive factors, and the firm specific factors are about the financial conditions of the company. ‘The procedures are as follows: 4 Meet with the management of the firm and review qualitative and quantitative factors and compares the company’s performance with that ofits peers 4 Convene a committee meeting discussing the lead analyst's recommendation before voting a M9222 WWW.GFEDU.NET ee «Bie 4. Notify the issuer of the rating and major considerations supporting it Other important knowledge: 4 All ratings are monitored on an ongoing basis, regular meetings with issuer's ‘management are organized, and any new information is under surveillance. 4 “Outlook” concept is used; positive outlook means that there is some potential upside conditional to the realization of current assumptions. 4 Credit ratings do not constitute any recommendation to purchase, sell, or hold any type of the security. The Link Between Ratings and Default There are ‘three main conclusions about the relationship between ratings and the probabilities of default. 4 Ratings and default rates are inversely related on the whole but no precisely , ic. better ratings mean low default rates 4 There is a substantial jump in default probability at the delineation of investment and speculative grades A There is a striking difference in default patterns between investment-grade and speculation-grade categories. Factors Impacting The External Ratings Several factors, such as time horizon, economic cycle, industry and geography are considered to impact the extemal ratings. 4 Time horizon ‘The creditworthiness of an issuer or the long-term debt security should be assessed over a long period so that the different impact in business-risk sensitivity from the economic cycles could be eliminated. Usually, the rating agencies try to igate the effect of cycles on ratings by incorporating, the effect of an “average cycle” in the scenarios, Thus, the final rating will be less volatile and less sensitive to expected changes in the business cycle. However, if it is convinced that a worsening of economic conditions both at the firm level and the macro level is persistent, the rating will be downgraded, Empirical studies show that the level of consensus among rating agencies is much lower for financial institutions than it is for corporate, and for a given rating category, banks tend to show higher default rates than corporate. This phenomenon is attributed to the information transparency of the different sectors. 1s28807 WWW.GFEDUNET uy 8 HAE 4 Beonomic Cycle ‘Transition matrices also appear to be dependent on the economic eycle as downgrade and default probabilities increase significantly during recessions. For investment-grade counterparties, migration volatil periods than during recession. s much lower during growth 4 Industry and Geography Homogeneity Studies on the impact of geography lead to two different conclusions, one is that there ‘would be differences in performance between US firms and non-US firms, as rating is originated in the United States and the rating history outside the U.S. is much shorter. The level of consensus among ratings agencies is much lower for financial institutions than it is for corporations, the rationale for such difference is often linked with the opacity of financial institutions. ‘The study also shows, first geographic homogeneity is convincing. And for a given rating category, banks tend to show higher default rates than corporations. Impact Of Rating Changes On Bond and Stock Prices ‘Most studies support the idea that a downgrade (upgrade) is likely to have a negative (positive) impact on bond prices. Several possible reasons are considered to explain this phenomenon. A The inverse relationship between default probability and rating, a downgrade(upgrade) is likely to have a negative(positive) impact on bond prices. 4. Information content of ratings is not fully anticipated and previously incorporated in asset prices 4 Ratings may influence the supply of and demand for securities and therefore trigger price changes A Rating triggers, ic. bond covenants based on the rating of a bond issue, For instance, the step-up bonds whose coupons increase when the issuer downgraded below a predefined threshold may lead to vicious-circle effect which means recently downgraded firms more likely to be downgrade again because downgraded company often suffer higher funding cost. The impact of rating change on stock prices is less obvious. Followings are the conclusions of the empirical studies. 4 Downgrades due to different reasons may lead to different effects. ‘+ The downgrades associated with increase in leverage might result in a Ist‘WHO WWW.GFEDUNET “eh = Bide AE wealth transfer from bondholders to shareholders and the equity price may erease. ‘+ While the downgrades linked to deteriorating financial prospects is bad news for both bondholders and shareholders and may bring about the decline in prices. 4 The value of equity may fall when rating is downgraded, for the probability of default increases and some of the value is transferred to third parties, which triggers the bankruptcy costs. 4 Downgrades affect stock prices significantly but upgrades do not. One possible explanation is that firm’s managers tend to divulge good news and retain bad news so that an upgrade is more likely to be expected than a downgrade. Another alternative is asymmetric utility functions with downside risk priced more dearly than upside potential Internal Rating ‘The Through-the-cycle and at-the-point approaches to Score a Company ‘The through-the-cycle and at-the-point approaches are the two ways to rate a company. ‘Their main differences lie in the time horizon used in the rating. 4 The through-the-cycle approach captures the creditworthiness of a firm over a longer time horizon, including the impact of normal cycles of the economy. A. through-the-cycle assessment therefore embeds scenarios about the economy as, well as business and financial factors 4 The at-the-point approach assesses the credit quality of a firm over the coming months (generally one year), KMV Credit Monitors EDF use this approach. Due to the different time horizon, the ratings by through-the-cycle approach are much ‘more stable than those by the at-the-point approach. How Internal Ratings Models My Create The Procyclicality Effect Procyclicality indicates that linking capital requirements to probability of default (PDs) may thereby reinforcing credit and economic cycles. Atthe-point measures of short-term PDs tend to underestimate risk during growth periods and overestimate is in recession. Therefore, expected loss will be very volatile due to the high volatility of PDs calculated using at-the-point methods. During recession, short-term PDs will increase sharply and the bank will have to reduce significantly its loan exposures in order to maintain stable expected losses. If such types of PD measure are used by a majority of banks, then firms will face liquidity shortages it sake ye.272407 WWW.GFEDUNET eae ee +N resulted from credit rationing. As a result, real economic cycles may be amplified. 42.2.2 Loan Individual Loan Credit Risk ‘The Contractually Promised Gross Return on a Loan ‘The three major types of‘loans issued by U.S. commercial banks are commercial and industrial (C&1), real estate, and individual. The return on a loan to a financial institution isa function of: 4 Interest rate on the loan. 4 Loan-related fees. 4. Credit-risk premium on the loan, 4 Any collateral backing the loan. balances and reserve Other terms not relating to the price (particularly compens: requirements). Compensating balances represent the proportion of a loan that borrowers must keep on deposit at the financial institution that made the loan. During the time the loan is outstanding, the firm is not permitted to make any withdrawals from the compensating balance amount. Compensating balances enhance retums to financial institutions beyond the explicit (stated) rate on the loan. Higher fees and collateral can also be used to compensate financial institutions for lending risk in addition to explicit rates, risk premiums, of restrictions on the amount of available credit. The contractually promised gross return on a loan, k, for each dollar of the loan amount is f+(L+m) 1[ol-R))" computed as follows: 1+k=1+ Where: #= loan origination fee base lending rate m= risk premium b= compensating balance requirement R= reserve requirement eure Ere IP HN 133ef 2k WWWGFEDUNET fea ioe HO Let's look at an example to see how noninterest rate terms can change the return om a loan. Example: Assume that a bank makes a $5 million, 1-year, spot C&I loan, The loan rate set by the bank equals some base lending rate plus an appropriate risk premium. Assuming that the base lending rate is 10 percent and the risk premium is 4 percent, the stated loan rate is set at L + m= 14 percent. Suppose al that the following additional charges apply to this loan: + Loan origination fee is 1/89, to the borrower. + Compensating balance requirement is 5%, noninterest-bearing. + Reserve requirement is 15%, paid by bank. This reserve is required by the Federal Reserve on all demand deposits, including compensating balances. Answer: ‘The gross rate of return for this loan can be calculated as follows: 0.00125+(0.10+004) 01412 1-[0.05(3-0.15)] 0.9575 14k =11475,k= 14.75% I+k=1+ ‘Note that this is greater than the simple promised interest return of 14 percent (= L + m). In today's market environment, the borrower's expense from fees and compensating balances are being driven down by competition between financial institutions, making these factors less important in calculation the total cost of a loan. The risk premium, m, tends to be the dominating force behind the promised yield once the base rate has been set. The base lending rate, L, in this example typically reflects the bank's marginal cost of funds. Rates such as the London Interbank Offered Rate (LIBOR), the rate on commercial paper, the federal funds rate, or the prime lending rate are often used as a base lending rate. LIBOR is the rate for interbank dollar loans in the offshore or Eurodollar market of a given maturity (e.g., 3-month LIBOR). The prime lending rate or simply the prime rate is the base lending rate periodically set by banks. The prime rate is commonly used to price longer-term foans, whereas the federal funds rate is used as the base rate for shorter-term loans. 1st uke HEN22824077 WWWGFEDUNET eu Se OL The Relationship between the Promised Return and the Expected Return on a Loan ‘The promised yield on the loan is the agreed rate at which the loan is made. The actual, or realized, yield to the financial institution will differ from the promised yield if the borrower is unwilling or unable to comply with any of the terms of the loan, most importantly the payments, ‘Thus, the expected return [E(F)] per dollar on money loaned is E(r) = p(1 + k), where p is the probability that the loan will be repaid. Note that k and p are not independent. If the premium, m, and other loan-related charges are set too high, the probability of loan repayment, p, declines. ‘The Probal lity of Default under a Linear Probability Model Using’ selected economic and financial information about borrowers, credit-scoring models can be used to calculate the probability of default for individual borrowers or to segregate borrowers into default risk categories. With credit-scoring models, it is critical to identify objective economic and financial measures of risk for any given class of borrower. For consumer debt, a credit-scoring model will likely include income, assets, age, occupation, and location. Corporate-debt applications of credit-scoring models employ financial ratios. Linear probability and logit models project a value for the expected probability of default ifa loan is made. A linear probability model uses linear regression with financial ratios to explain historic repayment patterns. The relative importance of the ratios, as indicated by the regression results, is used to forecast repayment probabilities, p, for new toans. For example, suppose that there are two observed characteristics of the borrower that Fixed costs explain past default behavior: fixed costs %( Ft ) and leverage ‘sales ah ). Now, suppose that the linear probability model for the probability of sets default for borrower i is represented by the equation (Z;) Z=0.45(FC)+0.20D/A)) (DI A= rated as: - Band ises So, if borrower i's fixed costs equal 35 percent of sales and its leverage ratio is 25 percent, its probability of default can be estimated as: (0.45x0.35)+(0.2%0.25)=20.75% One problem with this procedure is that it can yield probabilities greater than one. The logit model overcomes this problem by restricting the estimated range of default APHASIA 155‘fe bef WWW.GFEDUNET eu we probabilities such that (0 Z, <1). Linear Discriminant Analysis Models In contrast to linear probability and logit models, which project the expected probability of default for a borrower, discriminant models assign borrowers to high or low default-risk classes, depending on their financial characteristics. Altman's model for U.S. manufacturing firms is a prime example of the application of a linear-discriminant modell. Altman's model expresses the measure of the default-risk class of a borrower, Z, in terms of several financial ratios. The higher the value of Z, the lower the borrower's default-risk class. Airman's discriminant function is: Z=12 X,+14 Xz43.3 Xs40.6 Xet 1.0 Xs Where: X,= working capitalfotal assets ratio X2 = retained earningy/total assets ratio X5= earnings before interest and taxes/total assets ratio X4= market value of equity/book value of long-term debt ratio Xs = sales/total assets ratio According to Altman's credit-scoring model, a firm with a z-score less than 1.81 should be classified as a high-risk borrower. Note: Z ih this context is a measure of ability to repay, but in other applications in this topic, Z refers toa default probability. Drawbacks of using discriminant-analysis models to make credit-risk assessments include the following: 4 Two-state world. The models allow for only two outcomes--default or no default. ‘This is not reflective of reality where there can be different degrees of default. 4 Stability of factor weightings. There is no economic reason to believe that the weightings assigned to the model's variables will remain constant over time. 4 Nonquantifiable variables. The models do not include important factors that are difficult to quantify. A borrower's reputation and the stage of the business cycle are examples of nonquantiflable variables. 4 Data. There is no existing database on centralized business loans. 136 Wk AE THA‘sefeaits WWW.GFEDUNET Se Probability of Default and Marginal Default Probability ‘Newer credit-risk models use financial theory and financial-market data to make inferences about default probabilities on loans and debt instruments. As such, these models are most relevant to the credit-risk assessment of large corporate borrowers. The term structure approach to credit-risk measurements makes use of the spread between riskless government-debt securities and risky corporate-debt securities. Zero-coupon corporate debt and Treasury strips are often used with this method. Zero-coupon instruments, or zeros, are fixed-income securities that have no coupon payments, Zeros are issued at a significant discount to their par values and pay off par at maturity. ‘The simplest way to apply default risk measurement using the term structure of interest rates is to consider 1-year loans and bonds. Generally, a financial institution requires an expected return on a 1-year corporate debt issue that is at least equal to that of a 1-year Treasury bond (T-bond). The relationship used to determine the probability of default on the L-year corporate bond is p(I + k) = 1 + i, where p equals the probability that the principal and interest of the corporate debt will be repaid. The promised returns for the risky I-year corporate debt and the riskless I-year T-bonds are 1 + k and | + respectively. Given the observable T-bond and corporate bond rates, the implied probability of repayment can be calculated as p: a and the probability of default is 1 - p. The probability of default in any given year is referred to as the marginal default probability Fxample: Calculate the implied probability of default if the 1-year T-bill rate is 9.0 percent, and the rate on 1-year zero-coupon corporate bonds is 15.5 percent. Answer: The probability of default is: eek arta isS897 WWWGFEDUNET eu se Thus far we have assumed that the financial institution receives nothing if default occurs. To calculate the required risk premium when a total loss is not expected to accompany default, let_y be the percentage of principal and interest that is collectible in the event of default, where 7> 0. In this case, the risk premium to be applied to the 1-year corporate bond (loan) is: Ii he eeltieee Y+P- py (+i) Using the same information as the previous problem, if the lender can still collect 80 percent of the promised amount in the case of default, the required risk premium would be 1.24 percent: 1.09 18+ 0.9437 —(0.8%0.9437) (1.09) =0.0124.0r 1.24% ‘The Cumulative Default Probability and Marginal Default Probability To derive default probabilities for bonds that mature in more than ome year, itis necessary to know the borrower's marginal default probability. The cumulative default probability is the probability that a borrower will default over a multiyear period, as shown in the following: C, = 1 =P, P2X---Pa) Where: C,= cumulative probability of default P, = marginal probability of no default in yearn Notice that for the first year, the cumulative and marginal default probabilities are equal. To illustrate this concept, let's compute the cumulative default probability for a 2-year period. Example: Given: P; = 0.06 = marginal probability of default in year | = 0.08 = marginal probability of default in year 2 Answer: ‘The cumulative probability that default will occur over the 2-year period is G, =1e(p, xP,)=1-(0.94«0.92)=13.52% ‘This means that there is a 13.52 percent probability that the borrower will default 138 eee TPTefUENE WWW.GFEDU NET 5 + BU HAE not default at any time over the 2-year period. Compute a Marginal Default Probability Using the Term Structure Approach Recall that previously the yield curve was used to derive the marginal default probabi for a -year bond. The yield curve can also be used to compute the marginal default probability for future L-year periods using forward interest rates, which may be derived from the yield curve. A forward rate is the 1-period rate of interest expected on a bond issued at some time in the future. Assuming no arbitrage opportunities (no riskless profit), 1-year forward rates can be determined as follows: Where: i, P =G+iP0+6) (1+ in)? = the retum from holding a 2-year bond for two years 1 +i,= the return on a I-year bond 1 + fi= the expected I-year rate in one year (i.e., the 1-year forward rate) 139‘Example: Assuming 1-year and 2-year Treasury rates of 11 percent and 12 percent and the L-year and 2-year corporate bond rates are 16.5 and 1? percent, respectively, find the marginal probability of default for the corporate bond in its second year. Answer: The 1-year forward Treasury rate can be determined as follows: (ge (itl) The same procedure can be followed to derive forward rates from the corporate-bond yield curve. Given a 1-year corporate rate (ki) of 16.5 percent and a 2-year corporate rate (ks) of 17 percent, the I-year corporate forward rate (c)) can be determined as follows: : (4k, 1.1301 or f, =13% (-a)= (Hk) Using the forward rates derived earlier and rearranging the formula: p.(l+q)=(+h) ‘The probability of repayment in year 2 is determined to be: P,(l+q)=(I+A) ‘Therefore, the expected probability of defauit in year2 (the year 2.merginal default probability) is: 1p, =1-0.9617 =0.0383, 0F 3.83% Now, the probability of repayment on I-year bonds in ane year (p,) can be determined from the 1-year forward rates on T-bonds and corporate bonds as follows: A criticism of the term structure approach to deriving credit risk is that a yield curve for comporate bonds may only be available for large corporate borrowers. Critique the Mortality Rate Approach to Deriving Credit Risk The mortality rate for a bond or loan is based on historic default behavior. Si marginal default probability discussed earlier, the marginal. mortal to the rate is the probability of a bond or loan defaulting in any given year. The marginal mortality rate (MMR) for a grade AAA bond (loan) is computed as: 160 fewBk WWW.GFEDUNET * cas total valueof grade AAA bondsdefaultingin year tof ~ total valueof gradeA AAbondsoutstandirg in year tof issue, adjustedfor defaults,callssinkingfund redemption, and prior- yearmaturities MMR ‘The mortality rate approach to estimating defavit probabilities is criticized because it estimates future default behavior based on historic values. This means that the estimates will be dependent on the period used to derive marginal mortality rates. RAROC APPROACH TO DERIVING CREDIT RISK Almost all large U.S. banks use the risk-adjusted return on capital (RAROC) model to evaluate and price credit risk. Traditionally, financial institution managers evaluated loans using return on assets (ROA), which measures loan income against assets loaned. The rationale for using RAROC is that instead of measuring loan income against assets Joaned, net loan income should be measured against a measure of asset (loan) risk. This is expressed as: t-yearincomeona loan RAROC= Coen {oan (asset)riskor risk capital With the RAROC approach, a benchmark RAROC is set, and loan decisions are made relative to the benchmark. Also, the RAROC of an existing loan can be evaluated against the benchmark RAROC. If it falls below the benchmark, the loan's terms can be adjusted to make it acceptable. The denominator in the RAROC equation is often expressed as: Dx i | GR Where: AL=risk exposure(in dollars) D, =loan duration AL: ize of loan =maximumchangein loan credit premium 14R ’ « S AR Let’s take a closer look at “credit shock” expression, “> in the denominator of the + RAROC equation. Here, AR~ max[A(R- Ro)>0], where A(R, —R,) is the change in the spread between/-rated corporate bond Ids (Ri) and the matched duration government bond yields, Ro, over the past year. R is the prevailing average yield on i-rated corporate bonds. tA. 161SEAT WWWGFEDUNET a = Ee Diseuss How Migration Analysis Is Used To Measure Credit Risk ‘The following are two widely used models for measuring credit risk concentration: 4 Migration analysis. 4 External limits on the maximum value of loans that may be granted to individual borrowers or sectors. Migration analysis uses a loan migration matrix, also called a transition matrix, which provides probabilities that the credit quality of a loan will migrate from one quality class to aniother quality class over a period of time, usually one year. For example, a cell in a migration matrix may contain the probability that a loan that began the year with an AAA rating will have an AA rating by the end of the year. The probabilities contained in a migration matrix, referred to as transition probabilities, are based on historic averages of credit quality migration. Migration matrices provide useful information about the behavior of pools of loans over time and, as such, are often used as a benchmark against which new loan pool behavior can be measured. Compute a Concentration Limit for a Given Borrower Another method that managers use to measure and control concentration risk is to set external maximum limits on the loan amounts that can be granted to an individual borrower or business sector. The concentration limit is the maximum permitted loan amount to any individual borrower in a given sector, expressed as a percentage of capital as follows: Concentration limit = maximum loss as a percent of capital x loss rate Example: Assume a bank wants to limit its losses in a particular sector to 5 percent of its capital and that the loss rate for this sector is 60 percent. Calculate the concentration limit. ‘Answer: Concentration limit =0,05% (1/0.6)=8.33% 4.22.3 Loan Portfolio Credit Risk The Inputs Required to Compute the Expected Return and Variance of a Loan or Bond Portfolio Aggregate portfolio concentration risk can be measured and controtfed using modern portfolio theory (MPT). MPT requires expected return and variance for each asset in the 182 Suen pemafeo WWWGFEDUNET uk Bide portfolio, as well as the covariance between individual asset returns. Given the expected return for the loans and/or bonds in a portfolio, the portfolio's expected return can be calculated as follows: R-DxR Where : R, = the expected return on the asset portfolio, p R,= the expected return on the ith asset in the portfolio, _X,= the weight of asset i in the portfolio (the desired concentration) Portfolio risk expressed as variance is directly related to the degree of co-movement of the retums of the individual portfolio assets. As correlations between asset returns become smaller, portfolio risk decreases. Thus, by combining assets with low or negative correlations, it is possible to reduce portfolio risk. To avoid confusion in the following discussion, returns correlations and default correlations may be thought of in the same vein when evaluating loan portfolios. ‘The central theme behind MPT is that it is possible to construct a portfolio that has the highest expected retum at any given level of risk. Portfolios that, meet these conditions are referred to as efficient portfolios. To illustrate this idea, assume that a portfolio of loans is combined using the value weighting of the individual loans such that the Portfolios have the lowest possible risk at every level of expected retum. A plot of these portfolios will resemble Figure 1. In MPT, the curve starting at "B' and extending upward through "C" is referred to as the efficient frontier. Illustrate the Concept of Diversification within a Modern Portfolio Theory (MPT) Framework In Figure 1, the advantages of diversification can be seen by considering Portfolio A. Portfolio A is not fully diversified and is likely to be concentrated in relatively few loans or bonds. By adding loans or bonds that have low or negative correlation with the existing assets to Portfolio A, the credit risk of the portfolio can be reduced from o,, to ,9 While maintaining the same expected portfolio retum. Portfolio B is referred to as the minimum risk portfolio~the combination of assets that mini entire portfolio the variance of the Alternatively, if the financial institution is willing to accept the credit risk of Portfolio A, the portfolio can be rebalanced to increase its expected return to that of Portfolio C. ewe tre 9 ema. 163FECT WWW.GFEDUNET esi yk ohm Portfolios B and C are both said to be more efficient than Portfolio A. Thus, it can be seen that a financial institution can adjust the proportion of loans or bonds in a portfolio. to achieve any risk/return combination above and to the right of the minimum variance portfolio. Portfolios that are below the efficient frontier are not efficient and do not provide the best possible risk/retum combination. nei) <—_S Efficient fone 4A ie oe Application of Portfolio Theory Loan volume data may be used to construct variations of pure MPT models and allow managers to assess portfolio concentration and credit risk exposure. Loan volume data may be obtained from commercial bank call reports and/or the shared national credit database, Commercial bank call reports are prepared by banks for the Federal Reserve (Fed). These reports classify loans based on borrower type (e.g., commercial and industrial (C8), real estate, and agriculture). Aggregating these reports across all indi estimate of the national allocation of loans among loan categories, jual banks provides an Shared National Credit (SNC) is a national database on C&I loans that segregates loans on the basis of the borrower's two-digit standard industry classification (SIC) code. This database provides a picture of how the national C&I loan portfolio is allocated among, industries. The SNC database is essentially the market portfolio for C&tl loans. The aggregated portfolios formed via these data sources provide a national benchmark against which banks can compare their portfolios. Large deviations from the market allocations to any given industry sector or loan type indicate areas of potentially undesirable loan concentration. Regional benchmark portfolios may be constructed using the same data sources as those used for the national benchmark portfolio when regional 1 wR OH AEASty www.GFEDUNET Bik k -H comparisons are deemed more meaningful Loan Loss Ratios-Based Models ‘A mode) based on MPT that uses historic loan loss ratios can also be applied to the ‘measurement and control of loan concentration risk. This model provides estimates of the systematic loan loss risk of a given SIC sector relative to the loan loss risk of a bank’s total loan portfolio. The time series regression equation that this model uses to estimate systematic Joan loss risk is as follows: lossesinSICsectori__, gp totalloanlosses totalloanstoSICsectori totalloans Where: a= intercept term Bi ‘Suppose that regression estimates a beta value for the real estate sector equal to 0.25( fi.) = the systematic loss sensitivity of loans made to sector i and a beta for the C&E sector equal to 1.6( Ac.) By definition, the loss rate beta for the entire bank loan portfolio is 1.0(f,). Since zg, is greater than 1, it can be concluded that loan fosses to the C&I sector are systematically higher relative to the total loan losses of the entire bank portfolio. The opposite can be said for the Joan losses to the real estate sector. Since fy is less than 1, loan losses to the real estate sector are less sensitive to systematic factors than loan losses for the entire portfolio. Example: Barium Bank used time-series regression analysis of historic loan losses to obtain the following estimate: Xe = 0,005 + 1.3Xp; where Xe = loss rate on consumer loans, and Xp = loss rate on Barium’s total loan portfolio, Assuming Barium’s total loan loss rates increase to 10 percent, calculate the expected loss rate to the consumer sector. Answer: Given a 10 percent loss rate for the overall portfolio, the loss rate for the consumer sector will be: 0.005 + (J.3 x 0.10) = 13.5% ‘The implication of this model is that portfolios can tolerate concentrations in sectors with low betas because these sectors have lower loan loss sensitivity to systematic (economic) factors. In practice, itis often advisable to decrease portfolio concentrations of loan types that have relatively high systematic loan loss sensitivities (betas), especially when the economy is expected to move into a recession. Suk ATTA 16s