A STEP BY STEP APPROACH TO VALUE AT RISK RELEVANCE FOR INDIAN MARKET

Golaka.C.Nath*

Background Banks and financial institutions experience different types of risk, such as, business risk, strategic risk, and financial risk. Financial risk has been given more importance as it can be quantifiable while other risks have some amount of subjectivity. It is caused by movements in financial markets. The literature distinguishes four major categories of financial risk, viz., credit risk, operational risk, liquidity risk and market risk. Credit risk generally relates to the potential loss due to the default on the part of the counterparty to meet its obligations at designated time. It has three basic components: credit exposure, probability of default and loss in the event of default. Operational risk takes into account the errors that can be made in instructing payments or settling transactions, and includes the risk of fraud and regulatory risks. Liquidity risk is caused by an unexpected large and stressful negative cash flow over a short period. If a firm has highly illiquid assets and suddenly needs some liquidity, it may be compelled to sell some of its assets at a discount. Finally, market risk estimates the loss of an investment portfolio due to the changes in market price of

component securities. Banks and financial institutions in developed as well as in underdeveloped countries are generally required by their respective regulators to pass through minimum Capital Adequacy. The 1988 Basel Capital Accord set capital requirements on the basis of risk adjusted assets, defined as the sum of asset positions multiplied by asset-specific risk weights. These risk weights were intended to reflect primarily the credit risk associated with a given asset. In 1996 the Accord was amended to include r market risk, defined as the risk arising from movements in the market prices of trading positions (Basel Committee on Banking Supervision, 1996a). The 1996 Amendments Internal Models Approach (IMA) determines capital requirements on the basis of the output of the financial institutions internal risk measurement systems. Financial institutions are required to report daily their VaRs at the 99% confidence level over a one-day horizon and over a two-week horizon (ten trading days). Simply stated, VaR is the maximum loss of a trading portfolio over a given horizon, at a given confidence level. Financial institutions are allowed to derive

*Shri Golaka C Nath is the Advisor, Economic Research & Surveillance Department, The Clearing Corporation of India Limited.

their two-week VaR measure by scaling up the daily VaR by the square root of ten (see: Basel Committee on Banking Supervision, 1996b, p. 4). The new Capital Accord (commonly known as BASEL II) likely to be implemented in 2006 would require international banks to provide for operations risk. The minimum capital requirement on a given day is then equal to the sum of a charge to cover general market risk and a charge to cover credit risk (or idiosyncratic risk), where the market-risk charge is equal to a multiple of the average reported two-week VaRs in the last 60 trading days and the credit-risk charge is equal to 8% of riskadjusted assets. The US regulated banks and OTC derivatives dealers are subject to capital requirements determined on the basis of the IMA. More precisely, the market-risk charge is equal to the larger of: (i) the average reported two-week VaRs in the last 60 trading days times a multiplier and (ii) the last-reported two-week VaR. However, since the multiplier is not less than 3 (see below), the average of the reported VaRs in the last 60 trading days times a multiplier typically exceeds the last-reported VaR. Of course, the reliance on the financial institutions self-reported VaR to determine capital requirements creates an adverse selection and moral hazard problem, since the institution has an incentive to underreport its true VaR in order to minimize capital requirements The procedure

suggested by the Basel Committee to address this problem relies on backtesting (Basel Committee on Banking Supervision, 1996c): regulators should evaluate on a quarterly basis the frequency of exceptions (that is, the frequency of daily losses exceeding the reported VaR) for every financial institution in the most recent twelve-month period and the multiplier used to determine the market risk charge should be increased (according to a given scale varying between 3 and 4) if the frequency of exceptions is high. The reason to base backtesting on a daily VaR measure in spite of the fact that the market risk charge is based on a two-week VaR measure is due to the fact that VaR measures are typically computed ignoring portfolio revisions over the VaR horizon. The basic point to remember here is that the bank is assumed to hold the same portfolio over a historical time horizon from the date of VaR estimation. According to the Basel Committee, it is often argued that revalueat-risk measures cannot be compared against actual trading outcomes, since the actual outcomes will inevitably be contaminated by changes in portfolio composition during the holding period. This argument is persuasive with regard to the use of Value-atRisk measures based on price shocks calibrated to longer holding periods. That is, comparing the ten-day, 99th percentile risk measures from the internal models capital requirement with actual ten-day trading outcomes would probably not be a meaningful exercise. In particular, in any

given ten day period, significant changes in portfolio composition relative to the initial positions are common at major trading institutions. For this reason, the backtesting framework described here involves the use of risk measures calibrated to a one-day holding period. (Basel Committee on Banking Supervision (1996c, p. 3). Additional corrective actions in response to a high number of exceptions are left to the discretion of regulators. In recent years, VaR has become the standard measure that financial analysts use to quantify the market risk. The great popularity that this instrument has achieved among financial practitioners is essentially due to its conceptual simplicity: VaR reduces the (market) risk associated with any portfolio to just one number that is the loss associated with a given probability and horizon. VaR measures can have many applications. It evaluates the performance of risk takers and satisfies the regulatory requirements. VaR has become an indispensable tool for monitoring risk and an integral part of methodologies that allocate capital to various lines of business. Today regulators all over the globe have been forcing institutions to adopt internal models and calculate the required capital charge based on VaR methodologies. The Basel Committee on Banking Supervision of the BIS imposes requirements on banks to meet capital requirements based on the VaR estimated

through Internal Model Approach. Under this approach, regulators do not provide any specific VaR measurement technique to their supervised banks the banks are free to use their own model. But to eliminate the possible inertia of supervised banks to underestimate VaR so as to reduce the capital requirements, BIS has prescribed minimum standard of VaR estimates and also certain tests, such as backtesting, of VaR models. If VaR model of a bank fails in backtesting, a penalty is imposed resulting to higher capital charge. Thus, providing accurate estimates of VaR is of crucial importance for all stakeholders. If the underlying risk is not properly estimated, this may lead to a sub-optimal capital allocation with consequences on the profitability or the financial stability of the institutions. A bank would like to pick up a model that would generate as low VaR as possible but pass through the backtesting. There has been voluminous work done on VaR in financial market all over the world, the task of estimating/forecasting VaR still remains challenging. The major difficulty lies in modeling the return series which has normally heteroskedistic properties (being skewed and/or having fatter tails than normal distribution). Available VaR models can be classified into four broad categories: the historical simulation method, the Monte Carlo simulation method, modelling return distribution (including the variance/covariance method, which assumes

normality of the return distribution, and methods under Extreme Value Theory (EVT). All these VaR estimation methods adopt the classical approach: they deal with the statistical distribution of time series of returns. The main objective of this paper is to give a theoretical orientation of the subject and to discuss the steps involved in implementation of VaR models with backtesting as per BIS/RBI norms. Reserve Bank of India (RBI) has issued detailed guidelines on market risk for banks on the basis of the BIS framework though it has not become mandatory for banks to use VaR models1. It is interesting to note that a significant number of banks in developed countries are opting for the Internal Ratings Based methods, while their counterparts in developing countries (including India) follow the Standardized Approach. The main issue for moving towards the Standardized Approach is the absence of rating culture in emerging markets like India for which the unrated claims on Sovereigns, PSEs, banks, Security Firms and Corporates would attract 100% risk weighting and hence higher capital charge requirement. The BIS capital accord, for capital purposes, recognizes only external credit ratings, to the

total exclusion of banks internal rating system. However, RBI has issued detailed guidelines to Primary Dealers (PDs) for mandatory implementation of VaR methods while calculating the capital charge required2. The recent guidelines of RBI for PDs states that the capital requirement will be the higher of (i) the previous days valueat-risk number measured according to the above parameters specified in this section and (ii) the average of the daily value-at-risk measures on each of the preceding sixty business days, multiplied by a multiplication factor prescribed by Reserve Bank of India (3.30 presently). We have restricted our analysis to only application of VaR methodologies using a popular GOI bond index (IBEX Principal Return Index). The data period is from August 1994 to June 2004, about 11 years of data with 3622 data point. The data period is sufficiently large to cover various cyclical and seasonal factors in the economy. For the limited purpose of the study, we have used Normal (Variance Covariance), Historical Simulation and a weighted Historical Simulation method for simplicity and parsimony, though it is not difficult to estimate VaR using EVT and Weighted Normal.

RBI circular BP./21.04.103/2001 dated March 26, 2002 and RBI/2004/263/DBDO.No.BP.BC.103/21.04.151/2003-04 dated June 24, 2004. RBI circular IDMC.PDRS.PDC.3/03.64.00/2000-01 dated December 11, 2000 and RBI / 2004/7/IDMD 1/(PDRS) 03.64.00 /2003-04 dated January 7, 2004.
2

Literature Review There has been large volume of literature on VaR methodologies as well as on its implementation. The concept received tremendous response from banks all over the world. Banks management can apply the VaR concept to set capital requirements because VaR models allow for an estimate of capital loss due to market risk (see Duffie and Pan, 1997; Jackson, Maude and Perraudin, 1997; Jorion, 1997; Saunders, 1999; Friedmann and Sanddrof-Kohle, 2000; Hartmann-Wendels, et al., 2000; Simons, 2000, among others). The computation of volatility is the most important aspect of any VaR estimation. The volatility estimation should take care of the most stylized facts of any financial asset class the important ones being fat tailed property, volatility clustering and asymmetry of return distribution. Once these issues are identified in the distribution, then calculating volatility is easy. Today GARCH family models have been increasingly used by researchers to model volatility. An important documentation in this regard has been the J P Morgans RiskMetrics that applied declining weights to past returns to compute volatility with a decay factor 0.94 which is a variant of IGARCH. Other measures of volatility, which differs from the estimation of return variance, include Garman and Klass (1980), and Gallant and Tauchen (1998), who incorporate daily high and low quotes, and

Andersen and Bollerslev (1998) and Andersen, et al. (1999), who use average intraday squared returns to estimate daily volatility. Several studies such as Danielsson and de Vries (1997), Christoffersen (1998), and Engle and Manganelli (1999) have found significant improvements possible when deviations from the relatively rigid RiskMetrics framework are explored. Choosing an appropriate VaR measure is an important and difficult task, and risk managers have coined the term Model Risk to cover the hazards from working with potentially mis-specified models. Beder (1995), for example, compares simulationbased and parametric models on fixed income and stock option portfolios and finds apparently economically large differences in the VaRs from different models applied to the same portfolio. Hendricks (1996) finds similar results analyzing foreign exchange portfolios. In Indian context, Darbha (2001) made a comparative study of three models Normal, HS and Extreme Value Theory while studying the portfolio of Gilts held by PDs. Nath and Samanta (2003a and b) looked at the issues in implemention as well as selection of VaR methodologies for the portfolio of Government of India securities. Theoretical Issues As stated earlier, VaR is the maximum amount of money that may be lost on a portfolio over a given period of time, with a

given level of confidence and typically calculated for a one-day time horizon with 95% or 99% confidence level. The capital charge would be different for different holding period. BIS requires that VaR be computed daily by Banks, using a 99th percentile, one-tailed confidence interval with a minimum price shock equivalent to ten trading days (holding period) and the model incorporate a historical observation period of at least one year. The capital charge for a bank that uses a proprietary model will be higher of (i) The previous days VaR and (ii) an average of the daily VaR of the preceding sixty business days, multiplied by a multiplication factor. The multiplication factor may be 3 and this may go up if the regulators feel that 3 is not sufficient to account for potential weaknesses in the modeling process. In the case of PDs, RBI prescribes all these above criteria except that (i) minimum holding period would be 15 trading days; (ii) the minimum length of the historical observation period used for calculating VaR should be one year or 250 trading days. For PDs who use a weighting scheme or other methods for the historical observation period, the effective observation period must be at least one year (that is, the weighted average time lag of the individual observations cannot be less than 6 months); and (iii) the multiplication factor is presently fixed at 3.3. The RBI guidelines of January 2004 read as:

The PDs should calculate the capital requirement based on their internal Value at Risk (VaR) model for market risk, as per the following minimum parameters: (a) Value-at-risk must be computed on a daily basis. (b) In calculating the value-at-risk, a 99th percentile, one-tailed confidence interval is to be used. (c) A n i n s t a n t a n e o u s p r i c e s h o c k equivalent to a 15-day movement in prices is to be used, i.e. The minimum holding period will be fifteen trading days. (d) Interest rate sensitivity of the entire portfolio should be captured on an integrated basis by including all f i x e d income securities like G o v e r n m e n t securities, Corporate/PSU bonds, CPs and derivatives like IRS, FRAs, Interest rate futures etc., based on the mapping of the cash flows to work o u t the portfolio VaR. Wherever d a t a f o r calculating volatilities is not available, PDs may calculate the volatilities of such instruments using the G-Securities curve with appropriate spread. H o w e v e r, t h e d e t a i l s o f s u c h instruments and the spreads applied have to be reported and consistency of methodology s h o u l d b e e n s u re d . Instruments which are part of trading book, but f o u n d d i ff i c u l t t o b e subjected to measurement of market

risk may be applied a flat market risk measure of 15%. The instruments likely to be applied the flat market risk measure are units of MF, Unquoted Equity, etc., and added arithmetically to the measure obtained under VaR in respect of other instruments. (e) U n d e r w r i t i n g c o m m i t m e n t s a s explained at the beginning of the Appendix should also be mapped into t h e Va R f r a m e w o r k f o r r i s k measurement purposes. (f) The unhedged foreign exchange position arising out of the foreign currency borrowings under FCNR(B) loans scheme would carry a market risk of 15% as hitherto and the measure obtained will be added0 arithmetically to the VaR measure obtained for other instruments. (g) The choice of historical observation period (sample period) for calculating value-at-risk will be constrained to a minimum length of one year and not less than 250 trading days. For PDs who use a weighting scheme or other methods for the historical observation period, the effective observation period must be at least one year (that is, the weighted average time lag of the individual observations cannot be less than 6 months).

(h) The capital requirement will be the higher of (i) the previous days value-atrisk number measured according to the above parameters specified in this section and (ii) the average of the daily value-at-risk measures on each of the preceding s i x t y b u s i n e s s d a y s , multiplied by a multiplication factor prescribed by Reserve Bank of India (3.30 presently). (i) No particular type of model is prescribed. So long as the model used captures all the material risks run by the PDs, they will be free to use models, based for example, on variancecovariance matrices, historical simulations, or Monte Carlo simulations or EVT etc. The RBI has not restricted PDs for any particular model but the intention is that PDs should use the model which gives them the best comfort level with regard to capital requirement while not violating the regulatory norms on backtesting. The weaknesses of the VaR models may be due to (a) market prices often display patterns (heteroskedastic) that differs from the statistical simplifications used in modeling, (b) past not being always a good approximation of the future (October 1987 crash happened that did not have parallel in historical data), (c) most of the models take ex-post volatility and not ex-ante, (d) VaR estimations normally is based on end-of-day

positions and not take into account intraday risk, (e) models can not adequately capture event risk arising from exceptional market circumstances. Since VaR heavily relies on the availability of historical market price data on the portfolio to understand its effectiveness, it would be appropriate to use the long historical data to see if the stress conditions can be replicated. The logic behind VaR is to consider todays portfolio and find out what would have been its historical values (time series) and then construct the return series and then calculate the VaR numbers. We have used a bond index that represents the most liquid basket of underlying Government securities and has a long historical price sereis. Basic Statistics Related to VaR The portfolio consists of many securities and in our case we are concerned with only Gilts. The basic price equation of the portfolio can be written as follows: (1) and the return on the portfolio is at time defined as
n

(3) Where sij,t+1is the covariance and rij,t+1 is the correlation between security i and j on day t+1 and for rij,t+1 = 1 and we write sij,t+1 = s2i,t+1 for all i. The VaR of the portfolio is simply

VaR

PF, t +1

= P F, t +1 * Fp- 1

(4)

where Fp-1 is the pth quantile of the rescaled portfolio returns. When we use HS method, we write the VaR equation as
VaR
p PF , t +1

= - percentile {{RP F ,t +1 }r =1 ,100 p}

Select VaR Methodologies There are few VaR methodologies that are very simple and easy to implement, to name a few are (a) Normal (parametric using variance and covariance approach) and (b) Historical simulation. Cleverly these simple methods have been extended with application of weights recent events are given more weight and past is given less. However, different people have used different weighting methodologies. Riskmetrics has used exponentially moving average where the decay factor (l) has been considered as 0.94 while Boudoukh, et al. (1997) fixed it at 0.98. Variance-Covariance (Normal) Method The Variance-Covariance (Normal) method

R pf ,t +1 =

w
i =1

* Ri , t +1

(2)

Where the sum is taken over n securities in the portfolio, wi denotes the proportionate value of the holding of security i at the end of day t. And the variance of the portfolio should be written as

is the easiest of the VaR methodologies. Since we are considering a sovereign bond index for our analysis, it is known that interest rate movement in sovereign bond market is unidirectional at any point of time. For our purpose, the plain standard deviation would be useful to calculate the require VaR. But whether to take static variance of the entire time series or conditional variance is a point for debate. It is argued that variance changes over time horizons and hence we should not rely on unconditional variance for measuring VaR. Historical Simulation Method Historical simulation approach provides some advantages over the normal method, as it is not model based, although it is a statistical measure of potential loss. The main benefit is that it can cope with all portfolios that are either linear or nonlinear. The method does not assume any specific form of the distribution of price change/return. The method captures the characteristics of the price change distribution of the portfolio, as VaR is estimated on the basis of actual distribution. This is very important, as the HS method would be on the basis of available past data. If the past data does not contain highly volatile periods, then HS method would not be able to capture the same. Hence, HS should be applied when we have very large data points that are sufficiently large to take into account all possible cyclical events. HS method takes a portfolio at a point of time

and then revalues the same using the historical price series. Once we calculate the daily returns of the price series, then sorting the same in an ascending order and find out the required data point at desired percentiles. Linear interpolation can be used if the required percentile falls in between 2 data points. The moot question is what length of price series should be used to compute VaR using HS method and what we should do if the price history is not available. It has to be kept in mind that HS method does not allow for time-varying volatility. Another variant of HS method is a hybrid approach put forward by Boudhoukh, et al. (1997), that takes into account the exponential declining weights as well as HS by extimating the percentiles of the return directly, using declining weights on past data. As described by Boudhoukh et al. (1997, pp. 3), the approach starts with ordering the returns over the observation period just like the HS approach. While the HS approach attributes equal weights to each observation in building the conditional empirical distribution, the hybrid approach attributes exponentially declining weights to historical returns. The process is simplified as follows:Calculate the return series of past price data of the security or the portfolio from t-1 to t. To each most recent K returns: R(t), R(t1), R(t-K+1) assign a weight

[(1 - l) /(1 - l k )], [(1 - l) / (1- lk )] l,.....[(1 - l) / (1 - lk )] lk 1

-

respectively. The constant [(1 - l) /(1 - l k )] simply ensures that the weights sum to 1. S o r t t h e re t u rn s i n a s c e n d i n g order. In order to obtain p% VaR of the portfolio, start from the lowest return and keep accumulating the weights until p% is reached. Linear interpolation may be used to achieve exactly p% of the distribution. In many studies lambda (l) has been used as 0.98. Another Hybrid method is a weighting scheme suggested by Hull and White that allows us to transform the returns by multiplying the return series with a vector of ratios of last days (the day for which VaR is estimated) VaR and conditional volatilities (using the Riskmetrics method with a decay factor) calculated for previous n days. Then take the appropriate percentile values to represent the VaR numbers. The weighting scheme is justified on the ground that if the volatility on a previous day in the sample is lower than the current period volatility, VaR would be underestimated. The weighting scheme makes them comparable during the entire period. This is a method of normalizing the return series. Extreme Value Theory (EVT) Risk managers are primarily concerned with

the risk of low-probability events that could lead to catastrophic losses. Yet traditional VaR methods tend to ignore extreme events and focus on risk measures that accommodate the whole empirical distribution of returns. For example, it is often assumed that returns are normally or lognormally distributed, and little attention is paid to the distribution of the extreme returns we are most concerned about. The danger is then that our models are prone to fail just when they are needed most in large market moves, when we can suffer very large losses. One response to this problem is to use stress tests and scenario analyses. These can simulate the changes in the value of our portfolio under hypothesized extreme market conditions. These are certainly very useful. However, they are inevitably limited we cannot explore all possible scenarios and by definition give us no indication of the likelihoods of the scenarios considered. This type of problem is not unique to risk management, but also occurs in other disciplines as well, particularly in hydrology and structural engineering, where the failure to take proper account of extreme values can have devastating consequences. Researchers and practitioners in these areas handle this problem by using Extreme Value Theory (EVT) a specialist branch of statistics that attempts to make the best possible use of what little information we have about the extremes of the distributions in which we are

interested. The key to EVT is the extreme value theorem a cousin of the better-known central limit theorem which tells us what the distribution of extreme values should look like in the limit, as our sample size increases. Suppose we have some return observations but do not know the density function from which they are drawn. Subject to certain relatively innocuous conditions, this theorem tell us that the distribution of extreme returns converges asymptotically to:
Hxms( x ) = ,, 1x exp( - [ 1 + x ( x - m ) / s ] - / -(x-m)/s) exp( - e

If

x0 x=0

The parameters and s correspond to the mean and standard deviation, and the third parameter, x gives an indication of the heaviness of the tails: the bigger, x the heavier the tail. This parameter is known as the tail index, and the case of most interest in finance is where , x>0 which corresponds to the fat tails commonly founded in financial return data. In this case, our asymptotic distribution takes the form of a Frchet distribution. This theorem tells us that the limiting distribution of extreme returns always has the same form whatever the distribution of the parent returns from which our extreme returns are drawn. It is important because it allows us to estimate extreme probabilities and extreme quantiles, including VaRs, without having to make strong assumptions about an unknown parent distribution.

To put it in simple, suppose we have 1000 data points of historical prices and we have some idea how to divide the period into smaller time buckets of, say,10-15 days. Once we have a rationality of diving the entire period into optimal time buckets (we may use some statistical tests to identify the optimal number), we take the extreme high and low values and combine all these extreme high values one from each time bucket, to estimate the distribution properties of the extreme series and estimate the VaR using this extreme series as these are the tails in which we are interested rather than taking the entire sample of 1000 data points to estimate the tail. EVT provides a natural approach to VaR estimation, given that VaR is primarily concerned with the tails of our return distributions. To apply to VaR, we first estimate the parameters of the distribution, and there are a number of standard estimators available (in one of the earlier work, the author has used Gauss codes to estimate EVT VaR). Once we have these, we can plug them into a number of alternative formulas to obtain VaR estimates. To give a simple example, if we want to estimate a VaR that is out of (i.e., more extreme than) our sample range, we can project the tail out from an existing in-sample quantile Xk+1 where Xk+1 is the k+1-th most extreme observation in our sample and infer the (asymptotic) VaR from the projected tail using the formula:

VaR = [CL/k]

-x

Xk+1

Robust VaR Models - Backtesting Any method used for VaR estimation need to satisfy the criteria of back testing using the current data set. Suppose we calculate the VaR numbers with probability level 0.01. We can check the accuracy of a VaR model by counting the number of times VaR estimate fails (i.e. actual loss exceeds estimated VaR), say in 100 days. If we want to calculate VaR of a one-day holding period with 99% confidence level, logically, we are allowing 1 failure in 100 days. But if the number is more than 1, then the model is under predicting VaR numbers and if we find less number of failures the model is over predicting. The Basle Committee provides guidelines for imposing penalty leading to higher multiplication factor, when the number of failure is too high. However, no penalty is imposed when the failure occurs with less frequency than the expected number. Thus, selection of VaR model is a very difficult task. A model, which overestimates VaR, may result in reduced number of failure but increase the required capital charge directly. On the other hand if a model underestimates VaR numbers, the number failures may be too large which ultimately increases the multiplying factor and hence the required capital charge. Thus an ideal VaR model would be the one, which produces VaR estimates, as minimum as possible and also pass through the backtesting. Samanta and Nath (2003) have discussed the issue of selecting models on the basis of robust loss functions.

Where CL is the confidence level on which the VaR is predicated. EVT also gives us expressions for the confidence intervals associated with our VaR estimates. The EV approach to VaR has certain advantages over traditional parametric and non-parametric approaches to VaR. Parametric approaches estimate VaR by fitting some distribution to a set of observed returns. However, since most observations lie close to the centre of any empirical distribution, traditional parametric approaches tend to fit curves that accommodate the mass of central observations, rather than accommodate the tail observations that are more important for VaR purposes. Traditional parametric approaches also suffer from the drawback that they impose distributions that make no sense for tail estimation and fly in the face of EV theory. By comparison, the EV approach is free of these problems and specifically designed for tail estimation. Non-parametric or historical simulation approaches estimate VaR by reading off the VaR from an appropriate histogram of returns. However, they lead to less efficient VaR estimates than EV approaches, because they make no use of the EV theory that gives us some indication of what the tails should look like. More importantly, these approaches also have the very serious limitation that they can tell us nothing whatever about VaRs beyond our sample range.

The BIS requires that models must incorporate past 250 days data points (one year assuming Saturday/Sundays being nontrading days). In Indian market, RBI has issued guidelines for PDs to use one year and not less than 250 trading days for VaR estimation. Since Saturday is a trading day in bond market in India, we have taken 290 days (a period of about one year) for our analysis. Accordingly the capital charge is the higher of (i) the previous days value-atrisk number measured according to the above parameters specified in this section and (ii) the average of the daily value-at-risk measures on each of the preceding sixty business days, multiplied by a multiplication factor prescribed by RBI (3.30 presently for Pds). Basle Committee (1996b) provides following backtesting criteria for an internal VaR model (see van den Goorbergh and Vlaar, 1999; Wong et al., 2003, among others) (1) One-day VaRs are compared with actual one-day trading outcomes. (2) One-day VaRs are required to be correct on 99% of backtesting days. There should be at least 250 days (around one year) for backtesting. (3) A VaR model fails in Backtesting when it provides 5% or more incorrect VaRs. (4) If a bank provides a VaR model that fails in backtesting, it will have its capital multiplier adjusted upward, thus increasing the amount of capital charges.

For carrying out the Backtesting of a VaR model, realized day-to-day returns of the portfolio are compared to the VaR of the portfolio. The number of days when actual portfolio loss was higher that VaR provides an idea about the accuracy of the VaR model. For a good VaR model, this number would approximately be equal to the 1 per cent (i.e. 100 times of VaR probability) of back-test trading days. If the number of violation (i.e. number of days when loss exceeds VaR) is too high, a penalty is imposed by raising the multiplying factor (which is at least 3), resulting in an extra capital charge. The penalty directives provided by the Basle Committee for 250 back-testing trading days is as follows; multiplying factor remains at minimum (i.e. 3) for number of violation upto 4, increases to 3.4 for 5 violations, 3.5 for 6 violations, 3.65 for violations 8, 3.75 for violations 8, 3.85 for violation 9, and reaches at 4.00 for violations above 9 in which case the bank is likely to be obliged to revise its internal model for risk management (van den Goorbergh and Vlaar, 1999). For the limited purpose of this paper, to do the back testing, we can think of an indicator variable I(t) which is one if return of the day is more than the VaR for the previous day and zero otherwise. Average of the indicator variable should be our VaR percent. Data We have estimated VaRs as on June 30, 2004 for two important asset classes relevant for

banks and institutions. We have used GOI bond index (IBEX) developed by ICICI Securities and the INR-USD Exchange rate. The exchange rate has been taken from various RBI publications and website and the data is from march 01, 1993. We have chosen this data as our starting point because the unified exchange rate system was introduced from this date. This IBEX index is widely used by market participants and has a long time series. The index as of June 2004 has 19 underlying liquid bonds covering all the time buckets upto 15 years. The longest maturity underlying is 2019 while the shortest underlying is 2005. For the details on construction of IBEX, readers are requested to look at the website of ICICI Securities. While calculating VaR and doing

the backtesting, we have used the full period as well as various rolling periods of 500, 750, 1000, 1250, 1500 days. For example, rolling 500 days means, we will take 1 to 500 observations and compute the VaR and compare the same with next days losses and then 2 to 501,and so on. However, the use should see how the models work over various back periods with respect to estimation and backtesting. All our calculations are for June 30, 2003. The Chart-I gives the movement of IBEX PRI index used for the study while Chart-2 gives the daily return distribution during the period. The Chart-3gives the movement of exchange rate while Chart 4 gives the daily returns of exchange rates.

Chart-1: Movement of IBEX (August 1994-June 2004)

1500

1400

1300

1200 P RI 1100 1000 900 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Chart-2: Distribution of Daily Returns

The Table -1 gives the summary statistics of The daily returns. The figures shows that the series can not be said to have properties that can fit a normal distribution. Table-1 - Summary Statistics for Data: IBEX PRINCIPAL RETURN Min 1st Qu. Mean Median 3rd Qu. Max Total N NAs Variance Std Dev. Sum SE Mean : : : : : : : : : : : : -3.780000e+000 -2.000000e-002 8.845941e-003 0.000000e+000 5.000000e-002 3.830000e+000 3.622000e+003 0.000000e+000 5.127412e-002 2.264379e-001 3.204000e+001 3.762486e-003

P R IR T N

-1

-3

-5 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

LCL Mean : UCL Mean : Skewness : Kurtosis :

1.469139e-003 1.622274e-002 -1.355666e-001 6.392652e+001

Table-2 - Summary Statistics for Data: INR-USD Exchange Rate Rtn Min 1st Qu. Mean Median 3rd Qu. Max Total N NAs Variance Std Dev. Sum SE Mean : : : : : : : : : : : : -3.297790e+000 -4.728692e-002 1.242997e-002 0.000000e+000 5.766430e-002 2.976490e+000 2.891000e+003 0.000000e+000 8.281272e-002 2.877720e-001 3.593505e+001 5.352103e-003

LCL Mean : UCL Mean : Skewness :

1.935647e-003 2.292430e-002 -5.563513e-002

to estimate the VaR numbers. We first compute 1-day VaR numbers for all methods and as well as the average of 1-day VaRs in last 60 days in our sample. All VaR estimates correspond to the probability level 0.01 (equivalently correspond to the confidence level 0.99). For a given security/portfolio, maximum of these two VaRs (i.e. 1-day VaR in last day and 60-day average of 1-day VaR) has been adjusted to arrive at VaR numbers corresponding to two alternative holding periods, viz., 10-days and 15-days1. For calculating capital charge corresponding to a holding period h, h=10days or 15-days, the VaR with h-days holding period has been multiplied by the multiplication factor 3.3 (as given in the RBI circular for Pds).

Kurtosis: 2.495990e+001 Estimates of VaRs and Capital Charge In this section we report our estimated VaR figures and corresponding capital charges required All calculations are restricted to lefttail (one tailed) of return distribution and probability level is fixed at 0.01 (equivalently confidence level of VaR estimates is set to 0.99) strictly as per the RBI guidelines. Thus, the estimates we provide here actually refer to long-investment positions assuming that investment has been on the basis of the index. However, it is not difficult to take individual bonds or the portfolio of bonds

Chart-3: Movement of in INR-USD Exchange Rate (March 1993-June 2004)

50

45

40

Price
35 30

500

1000

1500

2000

2500

3000

Chart-4: Distribution of Daily Returns of Exchange Rate

3

The estimated results are given in Annexure1 for Normal Method and Annexure -2 for Historical Simulation Method and Annexure-3 for Weighted Historical Simulation (Hull-White) Method for the GOI bond index IBEX and Annexure-4 for Normal Method and Annexure -5 for Historical Simulation Method and Annexure-6 for Weighted Historical Simulation (Hull-White) Method for the INR-USD exchange rate data. We notice from the above tables that if we

R tn
3

-1

-3

-5 0 500 1000 1500 2000 2500 3000

take only last one year data for backtesting, The Normal method requires less capital charge vis--vis Historical simulation. But when we increase the backtesting horizon, Historical simulation gives better results and is more stable. Form a regulatory point of view, Historical simulation would be preferred. An important issue need to be mentioned here is that all VaR estimates provided in the tables are in percentage form, and thus, may actually be termed as the relative VaR (Wong,

As per the Basle Committee guideline (1996), capital charge should be derived based on VaR numbers for probability level 0.01 and holding periods 10-days. The VaR for 10-days holding period, however, are calculated based on 1-day VaR numbers computed daily basis. In India, guidelines issued to PDs maintain all attributes for capital charge computation except that VaR should have 15-days holding period (rather than 10-days holding period prescribed in the Basle Committee).

et al., 2003), which refers to the percentage of a portfolio value which may be lost after hholding period with a specified probability (i.e. the probability level of VaR). The absolute VaR (i.e. the VaR expressed in Rupees term) can easily be computed by multiplying the portfolio values with the estimated relative VaR. Similarly, the capital charge can also be represented in two alternative forms, viz., relative (i.e. in percentage) or absolute (i.e. in rupees terms). The additional information we require to convert a relative VaR/capital charge in a day to a corresponding absolute term (i.e. rupees term) figures is the value of the portfolio. The results show that normal method does not provide better results and if fails in backtesting when we apply the VaR methods at 99%. Historical Simulation and Weighted Historical Simulation methods provide better results. Conclusion: This paper has experimented with two most widely used VaR models, such as, variancec o v a r i a n c e / n o rm a l a n d h i s t o r i c a l simulation for estimating VaR using GOI bond index IBEX as well as INR-USD exchange rate data. The results are given in annexure I to 6. Historical simulation and weighted historical simulations methods provide better results in terms of back testing in general and they require higher capital charge while normal method requires less capital charge. It is upto the banks to decide which method to choose depending on the

tradeoff between the failures and higher capital charge. From a regulatory point of view Historical Simulations and weighted historical simulations would be useful. Reference: Altzner, P., F.Delbaen, J-M. Eber and D. Heath, 1999, Coherent Measures of Risk, Mathematical Finance, 9, pp. 203-208. Andersen, T. and T. Bollerslev (1998), Answering the Critics: Yes, ARCH Models do Provide Good Volatility Forecasts, International Economic Review, 39, 885-905. Andersen, T., T. Bo llerslev, F. Diebold and P. Labys (1999), The Distribution of Exchange Rate Volatility, Journal of the American Statistical Association, Website: http://citeseer.nj.nec.com/andersen99distri bution.html Basel Committee on Banking Supervision, 1996a Amendment to the Capital Accord to Incorporate Market Risk. Basel Committee on Banking Supervision, 1996b Overview of the Amendment to the Capital Accord to Incorporate Market Risk. Basel Committee on Banking Supervision, 1996c Supervisory Framework for the Use of Backtesting in conjunction with the Internal Models Approach to Market Risk Capital Requirements. Basel Committee on Banking Supervision, 2001, The Standardized Approach to Credit

Risk. Basle Committee on Banking Supervision (1995), An Internal Model-Based Approach to Market Risk Capital Requirements, Basle, Bank for International Settlements. Basle Committee on Banking Supervision (1996), Supplement to the Capital Accord to Incorporate Market Risks, Basle, Bank for International Settlements. Boudoukh J., Matthew Richardson, and R. F. Whitelaw (1997), The Best of both Worlds: A Hybrid Approach to Calculating Value at Risk, Stern School of Business, NYU Christoffersen, P. (1998), Evaluating Interval Forecasts, International Economic Review, 39, 841-862. Cruz, M (2002), Modeling, measuring and hedging operational Risk, John Wiley & Sons, Ltd. ISBN no. 0471515604 Danielsson, J. 2000, The Emperor has no clothes: limits to risk modelling, Mimeograph, London School of Economics. (Internet site http:// www.riskresearch.org). Danielsson, J. and C.G. de Vries, 2000, Value-at-Risk and Extreme Returns, Mimeograph, London School of E c o n o m i c s . ( I n t e rn e t s i t e h t t p : / / www.riskresearch.org. Darbha G, 2001, Value-at-Risk for Income portfolios: A comparison of alternative models, (www.nseindia.com)

Duff, D. and J. Pan (1997), An Overview of Value at Risk, Journal of Derivatives, 4, 7-49. Embrechts, P. [Ed.] (2000), Extremes and Integrated Risk Management, UBS Warburg. Engle, R. and S. Manganelli (1999), CAVaR: Conditional Autoregressive Value at Risk by Regression Quantiles, Manuscript, UCSD. Gallant, R. and G. Tauchen (1996), Which Moments to Match?, Econometric Theory, 12,657-681. Gallant, R. and G. Tauchen (1998), Reprojecting Partially Observed Systems with Application to Interest Rate Diffusions, Journal of the American Statistical Association, 93, 10-24. Garman, M. and M. Klass (1980), On the Estimation of Security Price Volatilities from Historical Data, Journal of Business, 53, 67-78. Hull, John and Allan White, 1998, Value at Risk When daily Changes in market Variables are not normally distributed Journal of Derivatives (Spring), 9-19. Hull, John and Allan White, 1998, Incorporating Volatility Updating into the Historical Simulation Method for Value at Risk Journal of Risk (Fall), 5-19. Hendricks, D. (1996), Evaluation of Valueat-Risk Models Using Historical Data, Federal Reserve Bank of New York Economic Policy Review, April, 39-70.

Hendricks, D., and B. Hirtle, 1997, Bank Capital requiremnts for market risk: The Internal models approach., Federal Reserve Bank of New York Economic Policy Review, 4, 1-12. Lopez, J.A, 1999, Regulatory Evaluation of Value-at-Risk models, Journal of Risk, 1, 201-242. Longin, F., 1996, The asymptotic distribution of extreme stock market returns, Journal of Business, 63, 383-406. Longin, F., 2000, From Value-at-Risk to Stress testing: The Extreme Value Approach, in Embrechts, P. [Ed.], Extremes and Integrated Risk Management, UBS Warburg. Nath, G C and Samanta, G P, 2003, Value at Risk: Concept and Its Implementation for Indian Banking System, UTIICM conference paper (http:// gloriamundi.org/ detailpopup.asp?ID=453056842) Nath G C and Reddy Y V, 2003, Value at Risk: Issues and Implementation in Forex Market in India, ICFAI Journal of Applied Finance, Nov 2003, http://gloriamundi.org/ detailpopup.asp?ID=453056841 Nelson, C.R. and A. F. Siegel, 1987, Parsimonious Modelling of Yield Curves, Journal of Business, Vol. 60, pp. 473-89.

Pagan, A., 1998, The Econometrics of Financial Markets, Journal of Empirical Finance, 1, 1-70. Reserve Bank of India (RBI), Handbook of Statistics, 2002-03 and various other publications and circulations. Samanta, G P & Nath G C, 2003, Selecting Value-at-Risk Models for Government of India Fixed Income Securities, ICFAI Journal of Applied Finance (forthcoming) (http://gloriamundi.org/detailpopup.asp?I D=453056896) Tsay, Ruey S. , 2002, Analysis of Financial Time Series, Wiley Series in Probability and Statistics, John Wiley & Sons, Inc. van den Goorbergh, R.W.J. and P.J.G. Vlaar (1999), Value-at-Risk Analysis of Stock Returns Historical Simulation, Variance Techniques or Tail Index Estimation?, DNB Staff Reports, No. 40, De Nederlandsche Bank. Wong, Michael Chak Sham, Wai Yan Cheng and Clement Yuk Pang Wong (2003), Market Risk Management of Banks: Implications from the Accuracy of Value-atRisk Forecasts, Journal of Forecasting, Vol. 22, pp. 23-33. Amexur 2: VaR estimation using Normal Method as of June 30, 2004 (IBEX)

Annexure 1: VaR estimation using Normal Method as of June30, 2004 (IBEX) Variance-Covariance (Normal) Method Full (95%) 0.4248 1.4001 1.4001 4.4275 5.4226 7.8507 6.0425 7.8507 6.0425 7.7009 6.4355 6.7943 6.4100 4.9337 6.4100 4.9337 6.2877 5.2546 5.5476 4.6370 5.6791 2.0270 1.5602 2.0270 1.5602 1.9884 1.6616 1.7543 1.4663 2.0270 1.5602 2.0270 1.5602 1.9884 1.6616 1.7543 1.4663 0.5647 0.4738 0.5504 0.4609 0.6006 0.5025 0.5381 0.4502 0.4874 1.5870 1.5870 5.0186 6.1465 Rolling500 (99%) Rolling 500 (95%) Rolling 750 (99%) Rolling 750 (95%) Rolling 1000 (99%) Rolling 1000 (95%) Rolling 1500 (99%) Rolling 1500 (95%) Rolling 2000 (99%) Rolling 2000 (95%) 0.4076 1.3262 1.3262 4.1937 5.1362

Description of Estimates

Full (99%)

DEaR

0.5187

60-day Average3.3

1.7075

1.7075

5.3996

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 3 14 23 49 56 62 86 91 124 75 108 66 87 57 81 58 70 44 56 49 68 55 34 42 37 47 40 10 14 13 18 14 20 54 77 85 102 6 7 6 9 7 11 1 2 1 2 1 3 2 9 18 42 47 52 52 3 15 26 57 67 75 75

6.6131

Over 1Year (4/19)

3 10 18 41 47 49 49

4 16 25 53 64 68 68

Over 500 days (5/25)

Over 750 days (8/38)

16

Over 1000 days (10/50)

34

Over 1500 days (15/75)

38

Over 2000 days (20/100)

40

Full (31/156)

54

For IBEX, data is continuous and hence 365 days are taken for 1 years

Annexure 2: VaR estimation using Historical Simulation as on June 30, 2004 (IBEX) Historical Simulation Full (95%) 0.2536 0.8071 0.8071 2.5523 3.1259 10.4226 3.9720 9.9207 3.6043 9.7364 4.7116 9.7364 3.77629 8.5100 3.2432 8.1002 2.9429 7.9497 3.8470 7.9497 3.08333 2.6911 1.0256 2.5615 0.9306 2.5139 1.2165 2.5139 0.97503 2.5049 7.9212 9.7015 2.6911 1.0256 2.5615 0.9306 2.5139 1.2165 2.5139 0.97503 2.5049 0.6663 0.2892 0.7613 0.2768 0.7998 0.3776 0.7618 0.32401 0.7591 0.2706 0.8588 0.8588 2.7157 3.326 Rolling5 00 (99%) Rolling500 (95%) Rolling750 (99%) Rolling7 50 (95%) Rolling100 0 (99%) Rolling 1000 (95%) Rolling200 0 (99%) Rolling2000 (95%) Rolling1 500 (99%) Rolling 1500 (95%)

Description of Estimates

Full (99%)

DEaR

0.7285

60-day Average3.3

2.4134

2.4134

7.6319

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 11 25 49 107 137 161 243 42 249 33 226 35 27 159 22 153 27 14 103 19 121 25 12 72 14 76 19 88 139 165 212 4 24 7 26 7 31 4 15 6 14 6 17 6 9 21 23 25 25 1 8 1 4 1 4 1 8 21 45 103 132 162 163

9.3471

Over 1Year (4/19)

1 6 8 17 19 21 21

10 23 44 101 133 147 147

Over 500 days (5/25)

Over 750 days (8/38)

Over 1000 days (10/50)

16

Over 1500 days (15/75)

18

Over 2000 days (20/100)

20

Full (31/156)

31

For IBEX, data is continuous and hence 365 days are taken for 1 years

Annexure 3: VaR estimation using WHS Method as on June 30, 2004 (IBEX) Weighted Historical Simulation Method(Hull-White - Lambda = 0.94) Rolling500 Rolling500 (99%) (95%) 0.5770 2.0815 2.0815 6.5823 8.0616 4.6571 7.2575 4.0229 7.5492 4.2010 7.8455 4.2197 3.8025 5.9257 3.2847 6.1639 3.4301 6.4058 3.4454 1.2024 1.8739 1.0387 1.9492 1.0847 2.0257 1.0895 2.0256 6.4054 7.8450 1.2024 1.8739 1.0387 1.9492 1.0847 2.0257 1.0895 2.0256 0.4040 0.5765 0.3365 0.5766 0.3459 0.6213 0.3497 0.6213 0.3496 1.0855 1.0855 3.4328 4.2043 Rolling Rolling750 Rolling1000 Rolling1000 Rolling1500 Rolling1500 Rolling2000 Rolling2000 750 (95%) (99%) (95%) (99%) (95%) (99%) (95%) (99%)

Description of Estimates 0.3425 1.1002 1.1002 3.4790 4.2609

Full (99%)

Full (95%)

DEaR

0.6657

60-day Average3.3

2.1984

2.1984

6.9519

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 8 21 33 66 77 86 132 39 142 35 137 34 28 94 25 90 26 26 82 23 78 24 79 89 121 24 71 21 67 22 66 13 33 12 34 12 34 8 21 8 22 8 22 6 10 19 21 23 23 2 8 2 9 2 9 1 8 21 33 68 79 88 89

8.5143

Over 1Year (4/19)

1 6 9 18 20 22 22

8 21 33 66 78 84 84

Over 500 days (5/25)

Over 750 days (8/38)

Over 1000 days (10/50)

17

19

Over 1500 days (15/75) Over 2000 days (20/100) Full (31/156)

21

28

For IBEX, data is continuous and hence 365 days are taken for1 years

Annexure-4: VaR estimation using Normal Method as on June 30, 2004 (INR -USD Exchange Rate) Variance-Covariance (Normal) Method Full (95%) 0.6287 1.7832 1.7832 5.6389 6.9062 5.7603 4.8801 4.8823 4.1180 4.8823 3.9629 4.7401 4.7033 3.9846 3.9863 3.3623 3.9863 3.2357 3.8703 3.2471 3.9769 1.4873 1.2600 1.2606 1.0633 1.2606 1.0232 1.2239 1.0268 1.4873 1.2600 1.2606 1.0633 1.2606 1.0232 1.2239 1.0268 1.5721 1.5721 4.9714 6.0887 1.0670 0.8996 0.4213 0.3548 0.3950 0.3318 0.3652 0.3063 0.4851 0.4061 1.3161 1.3161 4.1620 5.0974 Rolling500 (99%) Rolling 500 (95%) Rolling 750 (99%) Rolling 750 (95%) Rolling 1000 (99%) Rolling 1000 (95%) Rolling 1500 (99%) Rolling 1500 (95%) Rolling 2000 (99%) Rolling 2000 (95%)

Description of Estimates

Full (99%)

DEaR

0.7631

60-day Average3.3

2.1694

2.1694

6.8604

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 6 6 6 7 8 23 43 79 100 44 61 56 71 39 55 46 53 31 40 25 37 37 44 50 29 37 23 27 29 27 33 21 28 31 33 50 50 18 19 17 23 19 24 14 15 15 19 15 20 12 12 12 13 14 14 14 15 15 17 18 20 20 20

8.4022

Over 1Year (3/15)

8 8 8 8 8 8 8

10 10 10 10 10 10 10

Over 500 days (5/25)

Over 750 days (8/38)

Over 1000 days (10/50)

Over 1500 days (15/75)

Over 2000 days (20/100)

15

Full (28/140)

29

Annexure - 6: VaR estimation using WHS Method as on June 30, 2004 (INR -USD Exchange Rate) Weighted Historical Simulation Method(Hull-White - Lambda = 0.94) Rolling500 Rolling500 (99%) (95%) 0.9225 3.0444 3.0444 9.6272 11.7909 8.3612 10.8648 7.8537 10.7914 7.6185 10.7083 6.8825 6.8269 8.8711 6.4125 8.8112 6.2205 8.7433 5.6196 2.1588 2.8053 2.0278 2.7863 1.9671 2.7649 1.7771 2.7620 8.7344 10.6974 2.1588 2.8053 2.0278 2.7863 1.9671 2.7649 1.7771 2.7620 0.6546 0.8501 0.6157 0.8378 0.5955 0.8378 0.5385 0.8370 0.5382 1.7768 1.7768 5.6188 6.8816 Rolling Rolling750 Rolling1000 Rolling1000 Rolling1500 Rolling1500 Rolling2000 Rolling2000 750 (95%) (99%) (95%) (99%) (95%) (99%) (95%) (99%)

Description of Estimates 0.5407 1.7875 1.7875 5.6527 6.9231

Full (99%)

Full (95%)

DEaR

0.8298

60-day Average3.3

2.7421

2.7421

8.6713

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 6 6 7 9 13 34 60 29 64 21 46 15 14 37 15 39 15 4 14 5 16 5 3 10 4 12 4 10 14 38 38 2 7 4 7 4 7 2 6 4 6 4 6 2 6 4 6 4 6 4 4 4 4 5 5 5 7 7 8 10 13 13 13

10.6201

Over 1Year (3/15)

4 4 4 4 4 4 4

6 6 7 7 7 7 7

Over 500 days (5/25)

Over 750 days (8/38)

Over 1000 days (10/50)

Over 1500 days (15/75) Over 2000 days (20/100) Full (28/140)

15

30

Annexure - 5: VaR estimation using Historical Simulation as of June 2004 (INR -USD Exchange Rate) Historical Simulation Rolling500 Rolling500 Rolling750 Rolling750 Rolling1000 Rolling1000 Rolling1500 Rolling1500 Rolling2000 Rolling2000 (99%) (95%) (99%) (95%) (99%) (95%) (99%) (95%) (99%) (95%) 0.5054 1.6179 1.6179 5.1161 6.2659 2.4915 5.5227 2.2558 5.1384 2.2378 5.1136 2.1780 2.0343 4.5093 1.8418 4.1955 1.8272 4.1753 1.7783 0.6433 1.4260 0.5824 1.3267 0.5778 1.3203 0.5624 1.6407 5.1884 6.3544 0.6433 1.4260 0.5824 1.3267 0.5778 1.3203 0.5624 1.6407 0.2418 0.4857 0.1922 0.4598 0.1898 0.3981 0.1708 0.5038 0.2069 0.6639 0.6639 2.0994 2.5712

Description of Estimates

Full (99%)

Full (95%)

DEaR

0.9587

0.2950

60-day Average3.3

3.1622

0.9681

3.1622

0.9681

9.9997

3.0615

Max Cap Charge, H=10-day (%) Cap Charge, H=15-day (%) Backtesting -Failures 17 18 24 30 34 57 81 69 191 39 142 28 50 145 34 135 28 41 111 29 102 22 39 102 26 95 20 84 90 132 132 23 68 24 75 18 65 15 44 15 50 15 49 11 32 11 37 12 37 10 10 10 10 10 10 10 35 39 49 61 64 64 64

12.2471

3.7495

Over 1Year (3/15)

8 8 8 8 8 8 8

26 30 40 42 42 42 42

Over 500 days (5/25)

Over 750 days (8/38)

Over 1000 days (10/50)

Over 1500 days (15/75) Over 2000 days (20/100) Full (28/140)

16