1. Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 1 Distributions and Copulas for Integrated Risk Management Elements of Financial Risk Management Chapter 9 Peter Christoffersen

2. Overview In Chapter 6 we built univariate standardized nonnormal distributions of the shocks Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 2 where z t = r t /  t and where D( * ) is a standardized univariate distribution In this chapter we want to build multivariate distributions for our shocks where z t is a vector of asset specific shocks, z i,t = r i,t /  i,t and where  t is the dynamic correlation matrix We assume that the individual variances and the correlation dynamics have already been modeled

3. Overview The chapter proceeds as follows: First, we define and plot threshold correlations, which will be our key graphical tool for detecting multivariate nonnormality Second, we review the multivariate standard normal distribution, and introduce multivariate standardized symmetric t distribution and the asymmetric extension Third, we define and develop copula modeling idea. Fourth, we consider risk management and integrated risk management using the copula model Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 3

4. Threshold Correlations Bivariate threshold correlation is useful as a graphical tool for visualizing nonnormality in multivariate case Consider the daily returns on two assets, for example the S&P 500 and the 10-year bond return Consider a probability p and define the corresponding empirical percentile for asset 1 to be r 1 (p) and similarly for asset 2, we have r 2 (p) These empirical percentiles, or thresholds, can be viewed as the unconditional VaR for each asset Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 4

5. Threshold Correlations The threshold correlation for probability level p is Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 5 Here we compute the correlation between the two assets conditional on both of them being below their pth percentile if p 0.5 In a scatterplot of the two assets we include only the data in square subsets of the lower-left quadrant when p 0.5

6. Threshold Correlations If we compute the threshold correlation for a grid of values for p and plot the correlations against p then we get the threshold correlation plot Threshold correlation is informative about dependence across asset returns conditional on both returns being either large and negative or large and positive They therefore tell us about the tail shape of the bivariate distribution Next we compute threshold correlations for the shocks Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 6

7. Figure 9.1: Threshold Correlation for S&P 500 versus 10-Year Treasury Bond Returns Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 7

8. Figure 9.2: Threshold Correlation for S&P versus 10-Year Treasury Bond GARCH Shocks Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 8

9. Multivariate Distributions In this section we consider multivariate distributions that can be combined with GARCH (or RV) and DCC models to provide accurate risk models for large systems of assets We will first review the multivariate standard normal distribution, then the multivariate standardized symmetric t distribution, and finally an asymmetric version of the multivariate standardized t distribution Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 9

10. Multivariate Standard Normal Distribution In the bivariate case we have the standard normal density with correlation defined by Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 10 where 1-  2 is the determinant of the bivariate correlation matrix

11. Figure 9.3: Simulated Threshold Correlations from Bivariate Normal Distributions with Various Linear Correlations Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 11

12. Multivariate Standard Normal Distribution In the multivariate case with n assets we have the density with correlation matrix  Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 12 Note that each pair of assets in the vector z t will have threshold correlations that tend to zero for large thresholds The 1-day VaR is easily computed via where we have portfolio weights w t and the diagonal matrix of standard deviations D t+1

13. Multivariate Standard Normal Distribution The 1-day ES is also easily computed using Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 13 In multivariate normal distribution, linear combination of multivariate normal variables is normally distributed The multivariate normal distribution does not adequately capture the (multivariate) risk of returns This means that convenience of normal distribution comes at a too-high price for risk management purposes We therefore consider the multivariate t distribution

14. Multivariate Standardized t Distribution In Chapter 6 we considered the univariate standardized t distribution that had the density Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 14 where the normalizing constant is

15. Multivariate Standardized t Distribution The bivariate standardized t distribution with correlation takes the following form: Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 15 where Note that d is a scalar here and so the two variables have the same degree of tail fatness

16. Figure 9.4: Simulated Threshold Correlations from the Symmetric t Distribution with Various Parameters Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 16

17. Multivariate Standardized t Distribution In the case of n assets we have the multivariate t distribution Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 17 Where Using the density definition we can construct the likelihood function which can be maximized to estimate d

18. Multivariate Standardized t Distribution The correlation matrix can be preestimated using Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 18 The correlation matrix  can also be made dynamic, and can be estimated using the DCC approach An easier estimate of d can be obtained by computing the kurtosis,  2, of each of the n variables The relationship between excess kurtosis and d is

19. Multivariate Standardized t Distribution Using all the information in the n variables we can estimate d using Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 19 where  2,i is sample excess kurtosis of ith variable The standardized symmetric n dimensional t variable can be simulated as follows

20. Multivariate Standardized t Distribution where W is a univariate inverse gamma random variable, where U is a vector of multivariate standard normal variables, where U and W are independent The simulated z will have a mean of zero, a standard deviation of one, and a correlation matrix  Once we have simulated MC realizations of vector z we can simulate MC realizations of the vector of asset returns and from this the portfolio VaR and ES can be computed by simulation as well Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 20

21. Multivariate Asymmetric t Distribution Let be an n×1 vector of asymmetry parameters The asymmetric t distribution is then defined by Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 21 where

22. Multivariate Asymmetric t Distribution where is the so-called modified Bessel function of the third kind, which can be evaluated in Excel using the formula besselk(x, (d+n)/2) Note that the vector and matrix are constructed so that the vector of random shocks z will have a mean of zero, a standard deviation of one, and the correlation matrix  Note also that if = 0 then Note that the asymmetric t distribution will converge to the symmetric t distribution as the asymmetry parameter vector goes to a vector of zeros Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 22

23. Figure 9.5: Simulated Threshold Correlations from the Asymmetric t Distribution with Various Linear Correlations Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 23

24. Multivariate Asymmetric t Distribution From the density we can construct the likelihood function Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 24 which can be maximized to estimate the scalar d and vector The correlation matrix can be preestimated using The correlation matrix  can be made dynamic,  t, and can be estimated using the DCC approach

25. Multivariate Asymmetric t Distribution Simulated values of asymmetric t distribution can be constructed from inverse gamma and normal variables We now have Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 25 where W is an inverse gamma variable U is a vector of normal variables, U and W are independent Note that the asymmetric t distribution generalizes the symmetric t distribution by adding a term related to the same inverse gamma random variable W, which is now scaled by the asymmetry vector

26. Multivariate Asymmetric t Distribution The simulated z vector will have the following mean: Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 26 The variance-covariance matrix of the simulated shocks will be

27. Multivariate Asymmetric t Distribution The asymmetric t distribution allows for much more flexibility than the symmetric t distribution because of the vector of asymmetry parameters, For a large number of assets, n estimating the n different s may be difficult. Note that the scalar d and the vector have to describe the n univariate distributions as well as the joint density of the n assets. Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 27

28. Copula Modeling Approach In multivariate t distribution the condition that the d parameter is the same across all assets is restrictive The asymmetric t distribution is more flexible but it requires estimating many parameters simultaneously So we use copula functions Consider n assets with potentially different univariate distributions, f i (z i ) and cumulative density functions (CDFs) u i = F i (z i ) for i = 1,2,…., n Note that u i is simply the probability of observing a value below z i for asset i Our goal is to link the marginal distributions across the assets to generate a valid multivariate density Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 28

29. Sklar’s Theorem It states that for a very general class of multivariate cumulative density functions, defined as F(z 1,…..,z n ), with marginal CDFs F 1 (z 1 ),….,F n (z n ), there exists a unique copula function, G( * ) linking the marginals to form the joint distribution Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 29 The G(u 1,…,u n ) function is known as the copula CDF

30. Sklar’s Theorem Sklar’s theorem then implies that the multivariate probability density function (PDF) is Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 30 where the copula PDF is defined in the last equation as

31. Sklar’s Theorem Consider now the logarithm of the PDF Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 31 The log likelihood function corresponding to entire copula distribution model is constructed by summing log PDF over the T observations in our sample But if we have estimated the n marginal distributions in a first step then the copula likelihood function is

32. Sklar’s Theorem The upshot of this is that we only have to estimate parameters in the copula PDF function g(u 1,t,…, u n,t ) in a single step We can estimate all the parameters in the marginal PDFs beforehand. This makes high-dimensional modeling possible Sklar’s theorem is very general and not very specific It does not say anything about the functional form of G(u 1,…,u n ) and thus g(u 1,t,..,u n,t ) To implement copula modelling approach we need to make specific modeling choices for the copula CDF Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 32

33. Normal Copula We can build the normal copula function from the standard normal multivariate distribution In the bivariate case we have Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 33 where  * is the correlation between  -1 (u 1 ) and  -1 (u 2 ) and we will refer to it as copula correlation  -1 ( * ) denotes univariate standard normal inverse CDF

34. Normal Copula Note that if the two marginal densities, F1 and F2, are standard normal then we get Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 34 which is simply the bivariate normal distribution Note - If marginal distributions are NOT normal then normal copula does NOT imply normal distribution The normal copula is much more flexible than the normal distribution because the normal copula allows for the marginals to be nonnormal

35. Normal Copula In order to estimate the normal copula we need the normal copula PDF. It can be derived as Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 35 where denotes bivariate standard normal PDF and. denotes univariate standard normal PDF

36. Normal Copula The copula correlation,  *, can now be estimated by maximizing the likelihood Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 36 where we have u 1,t = F 1 (z 1,t ) and u 2,t = F 2 (z 2,t ).

37. Normal Copula In the general case with n assets we have the multivariate normal copula CDF and copula PDF Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 37 where u is the vector with elements (u 1,…,u n ), and where I n is an n-dimensional identity matrix that has ones on the diagonal and zeros elsewhere

38. Normal Copula The correlation matrix,  *, in the normal copula can be estimated by maximizing the likelihood Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 38 If the number of assets is large then  * contains many elements to be estimated and numerical optimization will be difficult.

39. Normal Copula Let us define the copula shocks for asset i on day t as follows: Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 39 An estimate of the copula correlation matrix can be obtained via correlation targeting In small dimensions this can be used as starting values of the MLE optimization. In large dimensions it provides a feasible estimate where the MLE is infeasible.

40. Normal Copula Consider again the previous bivariate normal copula. We have the bivariate distribution Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 40 Note that threshold correlations are computed from u 1 and u 2 probabilities and not from z 1 and z 2 shocks The normal copula gives us flexibility but multivariate aspects of the normal distribution remains The threshold correlations go to zero for extreme u 1 and u 2 observations, which is not desirable in a risk management model where extreme moves are often highly correlated across assets

41. Figure 9.6: Simulated Threshold Correlations from the Bivariate Normal Copula with Various Copula Correlations Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 41

42. t Copula t Copula is copula model built from the t distribution Consider first the bivariate case. The bivariate t copula CDF is defined by Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 42 where denotes the symmetric multivariate t distribution denotes the inverse CDF of the symmetric univariate t distribution

43. t Copula The corresponding bivariate t copula PDF is Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 43

44. Figure 9.7: Simulated Threshold Correlations from the Symmetric t Copula with Various Parameters Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 44

45. t Copula The t copula can generate large threshold correlations for extreme moves in the assets Furthermore it allows for individual modeling of the marginal distributions, which allows for much flexibility in the resulting multivariate distribution Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 45

46. t Copula In the general case of n assets we have t copula CDF Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 46 and the t copula PDF

47. t Copula Notice that d is a scalar, which makes the t copula somewhat restrictive but also makes it implementable for many assets Maximum likelihood estimation can again be used to estimate the parameters d and  * in the t copula. We need to maximize Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 47 defining again the copula shocks for asset i on day t as follows:

48. t Copula In large dimensions we need to target the copula correlation matrix, which can be done as before using Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 48 With this matrix preestimated we will only be searching for the parameter d in the maximization of lnL g earlier

49. Other Copula Models An asymmetric t copula can be developed from the asymmetric multivariate t distribution Only a few copula functions are applicable when the number of assets, n, is large So far we have assumed that the copula correlation matrix,  *, is constant across time However, we can let the copula correlations be dynamic using the DCC approach We would now use the copula shocks z* i,t as data input into the estimation of the dynamic copula correlations instead of the z i,t Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 49

50. Figure 9.7: Simulated Threshold Correlations from the Symmetric t Copula with Various Parameters Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 50

51. Risk Management Using Copula Models To compute portfolio VaR and ES from copula models, we need to rely on Monte Carlo simulation Monte Carlo simulation essentially reverses the steps taken in model building Recall that we have built the copula model from returns as follows: First, estimate a dynamic volatility model,  i,t, on each asset to get from observed return R i,t to shock z i,t = r i,t /  i,t Second, estimate a density model for each asset to get the probabilities u i,t = F i (z i,t ) for each asset Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 51

52. Risk Management Using Copula Models` Third, estimate the parameters in the copula model using lnL g =  T t=1 ln g(u 1,t,…,u n,t ) When we simulate data from copula model we need to reverse steps taken in the estimation of the model We get the algorithm: First, simulate the probabilities (u 1,t,…,u n,t ) from the copula model Second, create shocks from the copula probabilities using the marginal inverse CDFs z i,t = F -1 i (u i,t ) on each asset Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 52

53. Risk Management Using Copula Models Third, create returns from shocks using the dynamic volatility models, r i,t =  i,t z i,t on each asset Once we have simulated MC vectors of returns from the model we can easily compute the simulated portfolio returns using a given portfolio allocation The portfolio VaR, ES, and other measures can then be computed on the simulated portfolio returns Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 53

54. Integrated Risk Management Integrated risk management is concerned with the aggregation of risks across different business units within an organization. Senior management needs a method for combining marginal distributions of returns in each business unit In the simplest case, we can assume that the multivariate normal model gives a good description of the overall risk of the firm. If the correlations between all the units are one then we get a very simple result. Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 54

55. Integrated Risk Management Consider first the bivariate case Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 55 where we have assumed the weights are positive The total VaR is simply the (weighted) sum of the two individual business unit VaRs under these specific assumptions.

56. Integrated Risk Management In the general case of n business units we have Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 56 but again only when returns are multivariate normal with correlation equal to one between all pairs of units In general, when the returns are not normally distributed with all correlations equal to one, we need multivariate distribution from individual risk models. Copulas do exactly that and they are therefore very well suited for integrated risk management. But we need to estimate copula parameters and also need to rely on Monte Carlo simulation to compute organization wide VaRs and other risk measures.

57. Summary Multivariate risk models Multivariate normal distribution Threshold correlation Multivariate symmetric t and asymmetric t distribution The normal copula and t copula models Integrated risk management Elements of Financial Risk Management Second Edition © 2012 by Peter Christoffersen 57