On Conditional Value at Risk (CoVaR) for tail-dependent copulas

Size: px
Start display at page:

Download "On Conditional Value at Risk (CoVaR) for tail-dependent copulas"

Transcription

1 Depend Model 2017; 5:1 19 Research Article Special Issue: Recent Developments in Quantitative Risk Management Open Access Piotr Jaworski* On Conditional Value at Risk CoVaR for tail-dependent copulas DOI /demo Received September 15, 2016; accepted December 5, 2016 Abstract: The paper deals with Conditional Value at Risk CoVaR for copulas with nontrivial tail dependence We show that both in the standard and the modified settings, the tail dependence function determines the limiting properties of CoVaR as the conditioning event becomes more extreme The results are illustrated with examples using the extreme value, conic and truncation invariant families of bivariate tail-dependent copulas Keywords: Copulas, Tail dependence, Value-at-Risk VaR, Conditional Value-at-Risk CoVaR, Conditional quantiles MSC: 62H05, 60E05, 91B30, 91G40 1 Introduction This paper is based on the Profit/Loss P/L approach as for example in [1, 4, 15] We will study random variables X, Y,, which are modeling: welfare of financial institutions; financial positions; investment profits; rates of returns of stock prices and indices So generally The higher value of X, Y,, the better" We recall that Value-at-Risk, at a given significance level 0, 1, of a P/L random variable X, is defined as follows [15]: VaR X = inf{v R : PX + v < 0 } When X is modelling a financial position, VaR X is the smallest amount of capital v that ensures that X + v is solvent with probability at least equal to 1 Alternatively when X is modelling the gain from the investment, VaR X is the probabilistic" answer to the question How much may I lose?" ie it is the largest loss that one is exposed to with a confidence level of 1 compare [36] 11 The above can be expressed in terms of quantiles Namely Value-at-Risk at a level is equal to the negative upper quantile of X or lower 1 quantile of the loss X VaR X = Q + X = Q 1 X To switch to the alternative Loss/Profit L/P approach applied for example in [5, 18, 33] when random variables are modelling losses from the financial investments, actuarial risks or high water levels in hydrology, *Corresponding Author: Piotr Jaworski: Institute of Mathematics, University of Warsaw, Poland, PJaworski@mimuwedupl 2017 Piotr Jaworski, published by De Gruyter Open This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 30 License

2 2 Piotr Jaworski for example, it is is enough to change the sign of the variables L = X, and remember that, by convention, the subscript is changed The significance level is replaced by the confidence level c = 1, that is VaR c L = VaR X Now assume that we are measuring the risk basing on some extra information For example, we want to determine what size bailout would be required to keep a financial institution Y solvent with probability at least 1 β when a financial institution X would incur significant losses Conditional Value at Risk CoVaR introduced by Adrian and Brunnermeier in 2008 [1] and its later modifications proved to be very useful tools for quantifying such phenomena Let X and Y be random variables modelling positions CoVaR is defined as VaR of Y conditioned by X Specifically CoVaRY X := VaR β Y X E, where a Borel subset of the real line E represents an adverse event concerning X Most often E consists of one point a threshold or is a half-line bounded by a threshold From the definition of CoVaR, it is clear that one has to model the dependence between Y and X, and this can be achieved by means of copulas In this paper we continue the research started in [4], but we pay special attention to the case when Y and X are tail dependent, ie the tail dependence function of their copula is positive Since we are following a P/L approach, we are interested in the shape of the copula close to the origin, ie in the lower tail We will consider three families of copulas having nontrivial lower tail: the survival extreme value copulas, the survival conic copulas and left truncation invariant copulas Adrian and Brunnermeier [1] applied the construction with E consisting of one point To differentiate from their approach, we will call their CoVaR the standard one Specifically, the standard Conditional-VaR at a level, β is defined as VaR at level β of Y under the condition that X = VaR X Definition 11 CoVaR =,βy X = VaR β Y X = VaR X The above can be expressed in terms of conditional quantiles Namely CoVaR =,βy X = Q + βy X = Q + X The main technical drawback of this approach is the dependence of the CoVaR on the choice of the version of the conditional probability compare [2, Thm 331] Since we expect that a risk measure should uniquely assign a real number to a random pair, one has to extend definition 11 by providing an algorithm selecting the version of the conditional probability In section 4 we will follow the approach from [4] Under the mild assumption that the univariate distribution functions of X and Y are continuous, we will use the theory of copulas to provide a canonical way how to select a version of the conditional probability, which allows us to redefine CoVaR in a mathematically correct way, ie as a univocal risk measure Besides the issues described above, there are also some more practical drawbacks of the standard CoVaR pointed out for example by Mainik and Schaanning in [33], which motivate a modification of the original definition The main objection is due to the fact that the standard CoVaR is not compatible with concordance ordering Hence it is breaking" the paradigm: more dependence, more systemic risk The modified Conditional-VaR at a level, β is defined as the VaR at level β of Y under the condition that X VaR X Definition 12 CoVaR,βY X = VaR β Y X VaR X

3 On Conditional Value at Risk CoVaR for tail-dependent copulas 3 The above can be expressed in terms of quantiles Namely CoVaR,βY X = Q + βy X Q + X Modified CoVaR was introduced by Girardi and AT Ergün in 2013 [16] and Mainik and Schaanning in 2014 [33], both in the L/P setting The paper is organized as follows: In section 2 we recall the basic facts about copulas, their geometric transformations and tail expansions Next we describe the tail behaviour of extreme value, conic and left truncation invariant copulas In section 3 we show how to express modified CoVaR in terms of copulas Following [4], we study the threshold w *, β, C such that CoVaR Y = VaR w* Y We discuss the compatibility of modified CoVaR with concordance ordering of copulas ie with the strength of dependence between the conditioned and conditioning variable and provide approximate bounds for the thresholds Next we deal with copulas with a nontrivial tail expansion We show that for such copulas the first order limiting properties of w * are fully determined by the tail dependence function We illustrate the above on the example of extreme value, conic and left truncation invariant copulas The last section is devoted to the standard CoVaR First we refine the definition to make it univocal even when the dependence between X and Y is described by a singular copula Following [4], we study the threshold v *, β, C such that CoVaR = Y = VaR v* Y We discuss the incoherent response of standard CoVaR to concordance ordering of copulas We show that for almost all copulas for sufficiently small β, the threshold v *, β, C is smaller than that of the comonotonic copula M Next we deal with copulas with regular nontrivial tail expansion We show that for such copulas the first order limiting properties of v * are determined by the tail dependence function We illustrate the above on the example of extreme value, conic and left truncation invariant copulas 2 Copulas 21 Basic notation To fix notation, we recall some basic facts about copulas For more details the reader is referred to standard texts such as [8, 11, 13, 24 27, 34, 35] We recall that the function C : [0, 1] 2 [0, 1] is called a copula if the following three properties hold: c1 u 1, u 2 [0, 1] Cu 1, 0 = 0, C0, u 2 = 0; c2 u 1, u 2 [0, 1] Cu 1, 1 = u 1, C1, u 2 = u 2 ; c3 u 1, u 2, v 1, v 2 [0, 1], u 1 v 1, u 2 v 2 Cv 1, v 2 Cu 1, v 2 Cv 1, u 2 + Cu 1, u 2 0 Alternatively we can characterize copulas in a more probabilistic way Namely, a function C is a copula if and only if there exist random variables U, V, which are uniformly distributed on [0, 1], such that C is a restriction to the unit square [0, 1] 2 of their joint distribution function Random variables U and V are called the representers of the copula C Proved half a century ago, Sklar s Theorem remains crucial for applications of the copula theory It states that any multivariate distribution function may be expressed as a composition of a copula and its univariate marginals It allows to split the study of the multivariate phenomena into the study of marginals and the study of dependence Since in our study of CoVaR we will rely very much on Sklar s Theorem for random pairs, we recall how it is formulated:

4 4 Piotr Jaworski Theorem 21 Let F be a 2-dimensional distribution function and F 1, F 2 its marginal distribution functions, then there is a copula C such that for each x = x 1, x 2 R 2 Fx 1, x 2 = CF 1 x 1, F 2 x 2 Furthermore, the copula C is uniquely determined when the boundary distribution functions F i are continuous Conversely, if C is a 2-dimensional copula and F 1, F 2 are univariate distribution functions then the function F is a 2-dimensional distribution function and F 1, F 2 are its boundary distribution functions Remark 21 [20] Pr1 Let C be a copula of a random pair X 1, X 2 Suppose that F i s, the distribution functions of X i s, are continuous, then the random variables are representers of the copula C F 1 X 1, F 2 X 2 One more premise to use copulas to model systemic risk follows from the fact that the copulas are true measures of interdependence between random phenomena Namely they do not depend on the scale in which these phenomena are quantified Indeed, if C is a copula of a random pair X = X 1, X 2, and the functions h 1, h 2 are defined and strictly increasing on the supports of X 1, X 2, then C is also a copula of the transformed random pair Y = h 1 X 1, h 2 X 2 22 Geometrical transformations of copulas We recall that there exist eight linear isometric transformations of the unit square [0, 1] 2 : two mirror reflections with respect to the diagonals, two mirror reflections with respect to bisectors, one point reflection, two rotations ±π/2 90 and 270 degrees and identity They induce the transformations of copulas Namely let random variables U 1, U 2 be representers of a copula C and σ : [01] 2 [0, 1] 2 be an isometry, then random variables V 1, V 2 given by V 1, V 2 = σu 1, U 2 are uniformly distributed on the unit interval [0, 1] The copula C σ of the pair V 1, V 2 is called a reflection or respectively rotation of the copula C The copulas obtained by the point reflection are better known under the name survival copulas" and denoted by Ĉ Ĉp, q = p + q 1 + C1 p, 1 q Note that in this case V 1 = 1 U 1, V 2 = 1 U 2 Survival copulas are a useful tool when one is switching from P/L to L/P setting Indeed when C is a copula of gains X and Y, then the survival copula Ĉ is a copula of losses L 1 = X and L 2 = Y

5 On Conditional Value at Risk CoVaR for tail-dependent copulas 5 23 Copulas with nontrivial tail expansions In risk management one has to deal with extreme events and the interdependencies between them This leads to the study of the tail behaviour of a copula, ie of the possible approximations of a copula close to the vertices of the unit square Since applying a proper geometric transformation one may map any vertex of the unit square [0, 1] 2 to the selected one, we restrict ourselves to the vertex 0, 0 the origin Definition 21 We say that a copula C has a tail expansion at the vertex 0, 0 of the unit square if the limit exists for all nonnegative x, y The function L : [0, ] 2 [0,, Ctx, ty lim t 0 + t Ctx, ty Lx, y = lim, t 0 t is called the tail dependence function or the leading term of the tail expansion The second naming follows from the fact, proved in [21]: if L exists then we have a decomposition of a copula where R is bounded and Cu, v = Lu, v + Ru, vu + v, lim Ru, v = 0 u,v 0,0 The above can be applied to other vertices as well It is enough to reflect the copula For example for the upper tail the vertex 1,1 we get Ĉtx, ty tx + ty 1 + C1 tx, 1 ty Lx, y = lim = lim t 0 t t 0 t Note that L1, 1 and L1, 1 are equal to the lower and upper tail dependence coefficients We recall the basic properties of the tail dependence functions for details see [6, 7, 19 23, 28, 29, 31] Lemma 22 [20, 21]The tail dependence function induced by a copula C, Lu = lim t 0 + Ctu, u [0, + 2, is: t 1 homogeneous of degree 1, 2 2-nondecreasing and nondecreasing with respect to every variable, 4 nonnegative and bounded by the smaller coordinate of u: 5 Lipschitz with Lipschitz constant 1: 0 Lu minu 1, u 2 Lv Lu v 1 u 1 + v 2 u 2 6 concave: λ 1, λ 2 0, λ 1 + λ 2 = 1 Lλ 1 u + λ 2 v λ 1 Lu + λ 2 Lv Due to homogeneity, the leading term L is uniquely described by vertical sections like lt = L1, t

6 6 Piotr Jaworski Theorem 23 Let l : [0, [0, 1], l0 = 0, be a nondecreasing, concave function, such that lt t Then the function { L : [0, + 2 xl y [0, +, Lx, y = x for x > 0, 0 for x = 0 is a leading term of some copula The function lt = L1, t will be called a generator of the leading term The proof of theorem 23 follows from the examples below see [20] s21 and [19] for detailed calculations Since l is concave, the one-sided partial derivatives of L exist everywhere in 0, + 2 For left-sided derivatives we get Lu, v v v = l l v +, u u u u Lu, v = l v v u The same is valid for right-sided derivatives, by just switching +" and " We put Obviously for u > 0 and v 0 It = lt tl t+ = L1, t u Lu, v u v = I u Lemma 24 The function is: 1 nondecreasing; 2 I0 = l0 = 0 and I+ = l+ ; 3 for t such that lt < l+ It < lt I : [0, + [0, 1] Proof Since Lu, v is two-nondecreasing, nondecreasing in u and Lipschitz with constant 1, its derivative with respect to u is nondecreasing in v, nonnegative and bounded by 1 Hence so is I It is nondecreasing, nonnegative and bounded by 1 Since Lu, 0 = 0, I0 = L1, 0 u Since Lu, v is concave and homogeneous of degree 1, its derivative with respect to the first variable is homogeneous of degree 0, and we get = 0 L1, t L, 1 I+ = lim It = lim = lim = L0+, 1 t + t + u 0 + u u L, 1 = lim = lim L1, t = lim lt = l+ 0 + t + t + To show the last point we observe that l is a concave, nondecreasing function Hence l t + > 0 if lt < l+ Therefore It = lt tl t + < lt In the following we will use a generalized inverse of I I [ 1] s = inf{t : It > s}

7 On Conditional Value at Risk CoVaR for tail-dependent copulas 7 24 Examples 241 Extreme value copulas Let l : [0, + ] [0, 1] be a concave, nondecreasing function, such that lt t, then the function C l : [0, 1] 2 lnv [0, 1], C l u, v = uv exp lnul lnu is a copula see for example [35] 334, [11] 66 or [17] with an upper tail dependence function [20] 54 y Lx, y = xl x Note that copulas C l satisfy the following property C l u n, v n = C l u, v n, n > 0 They are called extreme value copulas, and when we put we get the well known Gumbel family of copulas The survival copula is given by lt = 1 + t 1 + t θ 1/θ, θ 1, Ĉ l u, v = u + v u1 v exp v = ul + ou + v u ln1 v ln1 ul ln1 u 242 Conic copulas Let l : [0, + ] [0, 1] be a concave, nondecreasing function, such that lt t, then the function C l : [0, 1] 2 v [0, 1], C l u, v = max ul, u + v 1 u is a copula with a tail dependence function y Lx, y = xl x Copulas of this form were used in [20] to prove the existence of copulas with given lower and upper tail dependence functions The survival copulas are given by Ĉ l : [0, 1] 2 1 v [0, 1], Ĉ l u, v = max u l 1 v 1 u 1 u 1, 0 They are known under the name conic copulas" see [14, 30] 243 LTI copulas Let f : [0, + ] [0, 1] be a surjective, concave and nondecreasing function and g its right inverse f gy = y Then the function C f : [0, 1] 2 0 for x = 0, [0, 1], C f x, y = xf gy x for x > 0

8 8 Piotr Jaworski is a copula introduced and considered in [10, 12] It belongs to the class of copulas that are invariant under left truncation For a suitable generator f, the popular Clayton copulas belong to this class Namely f t = 1 + t θ 1/θ, θ > 0 Furthermore see [12] proposition 41 the leading term of C f equals Lx, y = xf g 0 + y x Note that since g is a convex increasing function its right sided derivative at 0 exists and is nonnegative Furthermore L is nonzero if and only if g 0 + > 0 Then the generator lt equals f g 0 + t For Clayton copulas with positive θ we get y θ 1 θ Lx, y = x 1 + = x θ + y θ 1 θ, x which follows as well from the general results for Archimedean copulas see [20] Th6, [21] Pr10 or [7] Th31 3 Modified CoVaR by copulas We follow the P/L approach from [4] For the L/P setting the reader is referred to [33] Th 31b Let Cu, v be a copula of random variables X and Y having continuous distribution functions F X and F Y, then Therefore PY y X Q + X = PY y X Q+ X CoVaR,βY X = VaR w* Y, where w * = w *, β, C is the largest solution of the equation C, w * = β = C, F Yy Note that: w * = CoVaR,βF Y Y F X X Furthermore, as was observed in [33] Th34 for the L/P setting, modified CoVaR is compatible with the concordance ordering of copulas The same is valid for the P/L setting Theorem 31 Let C i u, v, i = 1, 2, be a copula of random variables X i and Y i having continuous distribution functions F Xi and F Yi and, β 0, 1] some fixed thresholds If u, v [0, 1] 2 C 1 u, v C 2 u, v then w *, β, C 1 w *, β, C 2 31 If furthermore t, + F Y1 t F Y2 t then Proof Since CoVaR,βY 1 X 1 CoVaR,βY 2 X 2 32 C 1, w *, β, C 1 = β = C 2, w *, β, C 2

9 On Conditional Value at Risk CoVaR for tail-dependent copulas 9 and C 1 C 2 we get w 1 = w *, β, C 1 w *, β, C 2 = w 2 Now since F Y1 F Y2 and w 1 w 2, we get CoVaR,βY 1 X 1 = VaR w1 Y 1 VaR w2 Y 1 VaR w2 Y 2 = CoVaR,βY 2 X 2 Theorem 31 implies the rough bounds for the threshold w * Due to the Fréchet-Hoeffding bounds see [11, 35], we have Mu, v Cu, v Wu, v, where Mu, v = minu, v is the comonotonic copula and Wu, v = u + v 1 + the countermonotonic one Since w *, β, M = β and w *, β, W = 1 + β [4] 31, we get: Corollary 32 Let C be any bivariate copula Then β w *, β, C 1 1 β If furthermore we assume that copula C is PQD positively quadrant dependent ie C dominates the independence copula Πu, v = uv u, v [0, 1] 2 Cu, v uv = Πu, v, we may improve the upper bound Indeed, since w *, β, C = β, we get: Corollary 33 Let C be a PQD copula then β w *, β, C β 31 Copulas with nontrivial tail expansions Theorem 34 Let the copula C have a nonzero tail dependence function L Ctu, tv v lim = Lu, v = ul t 0 t u Then for β < l Proof We have to solve the equation lim w *, β, C = 0, 0 w lim *, β, C = l 1 β 0 C, w * = β 33 First we show that for sufficiently small, w *, β, C is bounded by some linear function of We choose β 1 from the interval β, l+ We obtain So, for smaller than sufficiently small 1 C, l 1 β lim 1 = L1, l 1 β 0 1 = β 1 > β C, l 1 β 1 > β

10 10 Piotr Jaworski Since C is continuous and monotonic in the second variable, we get that for 0, 1 the solution w * of 33 is between 0 and l 1 β 1 Hence lim 0 w *, β, C = 0 To show the second equality we decompose C Cu, v = Lu, v + Ru, vu + v As was shown in [21], R is bounded and has a limit at zero lim Ru, v = 0, u,v 0,0 ie for any two sequences of numbers u n and v n from the unit interval, which are tending to 0 when n, the sequence R n = Ru n, v n tends to 0 as well We can rewrite equation 33 as L, w * + R, w * + w * = β We divide both sides by Hence Since for < 1 w * < l 1 β 1, we get l w* = β R, w * w * = l 1 β R, w * 1 + w * 1 + w * lim R, w * 1 + w * = 0 0 Since l 1 is continuous we obtain w lim *, β, C = l 1 β Survival conic copulas We need to solve the equation w max l, 1 + w 1 = β We get w * = { minl 1 β, 1 1 β for β < l, 1 1 β for β l, Hence for β < l and 1 1+l 1 β β w * = l 1 β 312 LTI copulas We need to solve the equation We get [4] f gw = β w * = f gβ,

11 On Conditional Value at Risk CoVaR for tail-dependent copulas 11 lim w * = f 0 = 0, 0 w lim * 0 = lim 0 w * = f 0 + gβ Note that since f is concave, nondecreasing its derivative may be finite or infinite In the first case C f has a nontrivial leading term and w * = f 0 + gβ + o 313 Survival extreme value copulas We get the equation ln1 w* + w * w * exp ln1 l = β ln1 Solving this with respect to l gives ln1 w* l = ln1 + β w * ln1 ln1 w * ln1 ln1 = β + O Next we invert l For β < l we obtain w * = 1 exp ln1 l 1 ln1 + β w* ln1 ln1 w * ln1 = l 1 β + O 2 To get a better approximation of w *, one may apply the following recurrence: Lemma 35 Let the functions w 1, w 2 : [0, 1] [0, 1] satisfy w 2 = 1 exp ln1 l 1 ln1 + β w1 ln1 ln1 w 1 ln1 If w * w 1 = O k, k 2, then w * w 2 = O k+1 Proof We have for sufficiently small and some constants C and C 1 w * w 2 = exp ln1 l 1 ln1 + β w* ln1 ln1 w * ln1 exp ln1 l 1 ln1 + β w1 ln1 ln1 w 1 ln1 C ln1 w * w 1 C 1 O k = O k+1 4 Standard CoVaR by copulas The discrepancy following the non-uniqueness of the conditional probability is usually overcome by an assumption that the pair X, Y has a continuous density f x, y, which allows us to select, as the density of Y X = x, the following function φ x y = { f x,y f X x when f X x > 0, f Y y when f X x = 0 or f X x = +, 41 where the density of X, denoted by f X x, is given by a formula f X x = f x, ydy

12 12 Piotr Jaworski and similarly the density of Y Note that, since the versions of the density of a given random pair may differ only at the set of Lebesgue measure 0, there may exist at most one continuous version Therefore the above determines uniquely a version of the conditional density But when every version of the density f x, y is discontinuous, there is no such canonical choice Furthermore, the choice of the version of the density f in formula 41 may affect CoVaR significantly We illustrate this by the following simple example: Example 41 For 0, 1, consider a pair of normally distributed random variables X and Y, X, Y N0, 1 coupled by a copula being the ordinal sum see [11, 35] of two copies of the independence copula with respect to the intervals [0, ] and [, 1] The resulting probability distribution has a discontinuous density We select two versions of the density, left+down continuous" and right+up continuous": and f 1 x, y = f 2 x, y = 1 1 2π exp π exp 1 1 2π exp 1 1 2π 1 exp x2 +y 2 2 x2 +y 2 2 when x, y Φ 1, when x, y > Φ 1, 0 otherwise x2 +y 2 2 x2 +y 2 2 when x, y < Φ 1, when x, y Φ 1, 0 otherwise, where Φ denotes the distribution function of the standard normal probability law The first choice implies via formula 41 CoVaR =,βx, Y = Φ 1 β, while the second CoVaR =,βx, Y = Φ 1 + β β As we see when + β = 2β, the results are quite different, which is not acceptable for risk measures In what follows, we will show how to avoid the discrepancy from the above example and moreover how to deal with singular distributions We keep only the assumption that random variables X and Y have continuous distribution functions F X and F Y We will redefine the standard CoVaR without any assumptions concerning the linking copula We start with so called technicalities" There is a well known relation between conditional probabilities and partial derivatives of copulas But in general copulas are only Lipschitz functions, which may not be differentiable Therefore we apply the notion of Dini derivatives see [32] From four possibilities we choose a left-sided upper one By D u C we denote the partial left-sided upper Dini derivative of copula Cu, v with respect to the first variable Cu, v Cu h, v D u Cu, v = lim sup h 0 + h We recall that for u 0, 1] D u Cu, 0 = Cu, 0 Cu h, lim = lim = 0, h 0 + h h 0 + h D u Cu, 1 = Cu, 1 Cu h, 1 u u h lim = lim = 1 h 0 + h h 0 + h Furthermore D u Cu, v is nondecreasing in v compare [9] The Dini derivative D u may be rewritten in the following way D u Cu, v = lim sup ν 0 + 0<h ν Cu, v Cu h, v, h

13 On Conditional Value at Risk CoVaR for tail-dependent copulas 13 which implies that it is a Borel measurable function Therefore the composition D u CF X X, v is a well defined σx measurable random variable Since a Dini derivative of a Lipschitz function is almost everywhere equal to the classical" derivative, the random variable D u CF X X, F Y y is a version of the conditional expected value see [2] 34 of the indicator function D u CF X X, F Y y = E1l Y y X as Putting F Y X=x = PY y X = x = lim η y + D ucf X x, F Y η we fix the version of the conditional probability This leads to the following definition of CoVaR see [4] and [33] Th 31a for L/P approach, compare [3] for conditional quantile setting where CoVaR,β Y X = sup{y : F Y y = v * } = Q + v * Y = VaR v* Y, When C is continuously differentiable we get Note that: v * = inf{v : D u C, v > β} 42 C u, F Y CoVaR,β Y X = β v * = CoVaR,β F Y Y F X X The L/P versions of the above two formulas can be found in [33] and [18] The next proposition is in line with the approach presented in [33] We provide conditions under which copula C, for given 0, 1 and sufficiently small β, is more stress testing sensitive" than the comonotonic copula M, although the latter dominates in concordance ordering We recall that v *, β, M = see for example [4] 31 Proposition 41 For any pair, β 0, 1, if then β < D u C, = lim η D uc, η, v *, β, C < = v *, β, M Proof We fix and β From 42 we have D u, v * β < D u C, Since the Dini derivative D u Cu, v is nondecreasing in the second variable, v * is smaller than Proposition 41 has interesting consequences Let X and Y be linked by a copula C, such that for a given 0, 1 D u C, > 0 Proposition implies that for sufficiently small β CoVaR =,βy X > CoVaR =,βy c X c, where X c and Y c are comonotonic versions of X and Y, which is quite intuitive A more precise information may bound the risk

14 14 Piotr Jaworski 41 Copulas with regular tail dependence In this section we will discuss the case of heavy tails We add to the assumption of the convergence of the copula C to its leading part L the assumption that the partial derivative of C with respect to the first variable is converging to the partial derivative of L We will base on the notation from section 23 Lu, v = ul v u, L v u u, v = I u Theorem 42 Let the copula C have a nonzero tail dependence function L and v lim D u Cu, v I = 0 u,v 0,0 u Then for β < l+ and lim v *, β, C = 0 0 v lim η β I[ 1] η lim inf *, β, C v lim sup *, β, C I [ 1] β 0 0 Proof We have to find the infimum v * = infv : D u C, v > β 43 First we show that for sufficiently small, v *, β, C is bounded by some linear function of We choose β 1 from the interval β, l+ ], such that I [ 1] β 1 > I [ 1] β note that I = l We obtain So, for smaller than sufficiently small 1 On the other hand lim D uc, I [ 1] β 1 = II [ 1] β 1 β 1 > β 0 D u C, I [ 1] β 1 > β lim D uc, l 1 β = Il 1 β = β l 1 βl l 1 β < β 0 Hence l 1 β does not belong to the half-line {v : D u C, v > β} and l 1 β inf{v : D u C, v > β} = v * Since D u C is nondecreasing in the second variable, we get that for 0, 1 the infimum v * from 43 lies between l 1 β and I 1 β 1 Hence lim 0 v *, β, C = 0 To show the first order estimate we decompose D u C We rewrite 43 Hence where D u Cu, v = Iv/u + R 1 u, v v * / = inf{v/ : Iv/ + R 1, v > β} = inf{v : Iv + R 1, v > β} I [ 1] β r = inf{v : Iv > β r} v * / inf{v : Iv > β + r} = I [ 1] β + r, r = sup{ R 1, v : l 1 β < v < I [ 1] β 1 } Since Ru, v tends to 0 with u, v 0, 0, we get that v lim η β I[ 1] η lim inf *, β, C 0

15 On Conditional Value at Risk CoVaR for tail-dependent copulas 15 and lim sup 0 v *, β, C I [ 1] β Corollary 43 Let the copula C have a nonzero tail dependence function L and v lim D u Cu, v I = 0 u,v 0,0 u Then for β < l+ and sufficiently small If furthermore then for sufficiently small Proof We compare the limits of v * / and of w * / Since β < l, from Lemma 24 we get v *, β, C w *, β, C β < I1 = lim η 1 Iη v *, β, M = > v *, β, C Il 1 β < ll 1 β = β On the other hand, since I is right-sided continuous, I lim η β I[ 1] η β Hence Finally, from Theorems 34 and 42 we get that l 1 β < lim η β I[ 1] η v lim inf *, β, C w *, β, C > lim 0 + η β I[ 1] η l 1 β > 0, which concludes the proof of the first inequality The condition β < lim η 1 Iη implies that I [ 1] β < 1 Therefore, due to Theorem 42, for sufficiently small v *, β, C < Corollary 43 has interesting consequences Let X and Y be linked by a tail dependent copula C Then the first inequality implies that for sufficiently small and β, modified CoVaR dominates the standard one CoVaR =,βy X CoVaR,βY X The adverse event X VaR X is worse for Y than X = VaR X The second inequality implies that for C = M and for sufficiently small and β CoVaR =,βy X > CoVaR =,βy c X c, where X c and Y c are comonotonic versions of X and Y Although M dominates C in concordance ordering, C is more stress testing sensible" as regards standard CoVaR Note, that when we add one more assumption to those in Corollary 43, namely that the copula C has a density which is bounded from 0 on some neighbourhood of the diagonal = {t, t : t [0, 1]}, then there exists a global" bound β 0, such that for all 0, 1 and all β 0, β 0 ] v *, β, M = > v *, β, C

16 16 Piotr Jaworski 42 Examples 421 Clayton copulas, θ > 0 We recall that the Clayton copula with positive θ is given by C Cl u, v; θ = u θ + v θ 1 θ 1 From this we derive We get [4] Note that if then for all 0, 1 C Cl u, v; θ u = u θ + v θ 1+θ θ 1 u θ 1 v * = β θ 1+θ 1 + θ 1 θ = β θ 1 θ 1+θ 1 + O 1+θ β < 2 θ+1 θ, v *, β, C Cl ; θ < = v *, β, M Since for fixed and β lim v *, β, C Cl ; θ =, θ we observe that v *, β, C Cl ; θ is not a decreasing function with respect to θ Moreover for sufficiently large θ it is increasing Thus we observe for Clayton copulas the same phenomenon as described in [33] for normal distributions The standard CoVaR is decreasing when the increasing in concordance ordering family of copulas is approaching the maximal copula M 422 Survival conic copulas Since v C l u, v = max u l, u + v 1, u it can be easily derived that D u C l, v = { I v when l v + v 1, 1 when l v < + v 1, where It = lt tl t + is a nondecreasing function introduced in section 23 1 Thus, we get for β < I and I [ 1] β Furthermore if β < I1 than for all 0, 1 v * = I [ 1] β β < D u C l, and Proposition 41 implies v *, β, C l < = v *, β, M

17 On Conditional Value at Risk CoVaR for tail-dependent copulas LTI copulas Since we get where We have to solve the equation C f x, y = { D u C f x, y = f xf 0 for x = 0, for x > 0, gy x gy x gv I f = β, I f t = f t tf t +, f + gy gy x x We obtain [4] For fixed β 0, 1 we get v * = f I [ 1] f β, lim v * = f 0 = 0, 0 v lim * 0 = f 0 + I [ 1] f β Furthermore when g 0 + > 0 and β < I f g 0 +, then due to Proposition 41, for all 0, 1 v *, β, C f < = v *, β, M 424 Survival extreme value copulas Since we get ln1 v Ĉ l u, v = u + v u1 v exp ln1 ul, ln1 u D u Ĉ l u, v = v exp ln1 ul = I ln1 v + Ou + v, ln1 u ln1 v 1 + I ln1 u ln1 v ln1 u where It = lt tl t + is a nondecreasing function introduced in section 23 We get the following formula determining v * { } ln1 v ln1 v v * = inf v : v exp ln1 l 1 + I > β ln1 ln1 We rearrange this to give v * = inf { v : I ln1 v > 1 1 β ln1 1 v exp ln1 l } ln1 v ln1 For simplicity we assume that generator l is strictly concave and twice differentiable with positive l t This implies that I is continuous and strictly increasing, hence invertible In such a case we get for β < I v * = 1 exp ln1 I β ln1 v* exp ln1 l 1 v * ln1 = I 1 β + O 2

18 18 Piotr Jaworski To get a better approximation one may apply the same approach as in Lemma 35 Furthermore since for all 0, 1 D u Ĉ l, I1, Proposition 41 implies that for all β < I1 and 0, 1 v *, β, Ĉl < = v *, β, M Acknowledgement: The author acknowledges the support from National Science Centre, Poland, via project 2015/17/B/HS4/00911 He would like also to thank the guest editor Steven Vanduffel and the anonymous referees for many valuable comments and suggestions References [1] Adrian, T and MK Brunnermeier 2016 CoVaR Am Econ Rev 1067, [2] Billingsley, P 1979 Probability and Measure, John Wiley & Sons, Chichester [3] Bernard, C and C Czado 2015 Conditional quantiles and tail dependence, J Multivariate Anal 138, [4] Bernardi, M, F Durante, and P Jaworski 2017 CoVaR of families of copulas Stat Probabil Lett 120, 8-17 [5] Bernardi, M, F Durante, P Jaworski, L Petrella, and G Salvadori 2016 Conditional risk based on multivariate hazard scenarios Preprint [6] Charpentier, A and A Juri 2006 Limiting dependence structures for tail events, with applications to credit derivatives J Appl Probab 432, [7] Charpentier, A and J Segers 2009 Tails of multivariate Archimedean copulas J Multivariate Anal 100, [8] Cherubini, U, E Luciano, and W Vecchiato 2004 Copula Methods in Finance John Wiley & Sons, Chichester [9] Durante, F and P Jaworski 2010 A new characterization of bivariate copulas Comm Statist Theory Methods 3916, [10] Durante, F and P Jaworski 2012 Invariant dependence structure under univariate truncation Statistics 462, [11] Durante, F and C Sempi 2016 Principles of Copula Theory CRC Press, Boca Raton FL [12] Durante, F, P Jaworski, and R Mesiar 2011 Invariant dependence structures and Archimedean copulas Stat Probabil Lett 8112, [13] Embrechts, P 2009 Copulas: a personal view J Risk Insur 763, [14] Fernández-Sánchez, J and M Úbeda-Flores 2016 The distribution of the probability mass of conic copulas Fuzzy Set Syst 284, [15] Föllmer, H and A Schied 2004 Stochastic Finance An Introduction in Discrete Time Second edition Walter de Gruyter, Berlin [16] Girardi, G and TA Ergün 2013 Systemic risk measurement: multivariate GARCH estimation of CoVar J Bank Financ 378, [17] Gudendorf, G and J Segers 2010 Extreme-value copulas In Copula Theory and Its Applications, pp Springer, Heidelberg [18] Hakwa, B, M Jäger-Ambrozewicz, and B Rüdiger 2015 Analysing systemic risk contribution using a closed formula for conditional Value at Risk through copula Commun Stoch Anal 91, [19] Jaworski, P 2003 Asymptotics of bivariate copulas in Polish Matematyka Stosowana 4, [20] Jaworski, P 2004 On uniform tail expansions of bivariate copulas Appl Math 314, [21] Jaworski, P 2006 On uniform tail expansions of multivariate copulas and wide convergence of measures Appl Math 332, [22] Jaworski, P 2010 Tail behaviour of copulas In Copula Theory and its Applications, pp Springer, Heidelberg [23] Jaworski, P 2013 The limiting properties of copulas under univariate conditioning In Copulae in Mathematical and Quantitative Finance, pp Springer, Heidelberg [24] Jaworski, P, F Durante, W Härdle, and T Rychlik, editors 2010 Copula Theory and its Applications Springer, Heidelberg [25] Jaworski, P, F Durante, and W Härdle, editors 2013 Copulae in Mathematical and Quantitative Finance Springer, Heidelberg [26] Joe, H 1997 Multivariate Models and Dependence Concepts Chapman & Hall, London [27] Joe, H 2014 Dependence Modeling with Copulas Chapman & Hall/CRC, Boca Raton FL [28] Joe, H, H Li, and A Nikoloulopoulos 2010 Tail dependence functions and vine copulas J Multivariate Anal 101 1, [29] Juri, A and MV Wüthrich 2002 Copula convergence theorems for tail events Insurance Math Econom 303, [30] Jwaid, T, B De Baets, J Kalická, and R Mesiar 2011 Conic aggregation functions Fuzzy Set Syst 1671, 3-20

19 On Conditional Value at Risk CoVaR for tail-dependent copulas 19 [31] Li, H and P Wu 2013 Extremal dependence of copulas: a tail density approach J Multivariate Anal 114, [32] Łojasiewicz, S 1988 An Introduction to the Theory of Real Functions Third edition John Wiley & Sons, Chichester [33] Mainik G and E Schaanning 2014 On dependence consistency of CoVaR and some other systemic risk measures Stat Risk Model 311, [34] McNeil, AJ, R Frey, and P Embrechts 2005 Quantitative Risk Management Concepts, Techniques and Tools Princeton University Press [35] Nelsen, RB 2006 An Introduction to Copulas Second edition Springer, New York [36] Risk Metrics - Technical Document 1996 Morgan Guaranty Trust Company of New York

On the Conditional Value at Risk (CoVaR) from the copula perspective

On the Conditional Value at Risk (CoVaR) from the copula perspective On the Conditional Value at Risk (CoVaR) from the copula perspective Piotr Jaworski Institute of Mathematics, Warsaw University, Poland email: P.Jaworski@mimuw.edu.pl 1 Overview 1. Basics about VaR, CoVaR

More information

Lecture Quantitative Finance Spring Term 2015

Lecture Quantitative Finance Spring Term 2015 on bivariate Lecture Quantitative Finance Spring Term 2015 Prof. Dr. Erich Walter Farkas Lecture 07: April 2, 2015 1 / 54 Outline on bivariate 1 2 bivariate 3 Distribution 4 5 6 7 8 Comments and conclusions

More information

arxiv:physics/ v1 [physics.soc-ph] 18 Aug 2006

arxiv:physics/ v1 [physics.soc-ph] 18 Aug 2006 arxiv:physics/6819v1 [physics.soc-ph] 18 Aug 26 On Value at Risk for foreign exchange rates - the copula approach Piotr Jaworski Institute of Mathematics, Warsaw University ul. Banacha 2, 2-97 Warszawa,

More information

Behaviour of multivariate tail dependence coefficients

Behaviour of multivariate tail dependence coefficients ACTA ET COMMENTATIONES UNIVERSITATIS TARTUENSIS DE MATHEMATICA Volume 22, Number 2, December 2018 Available online at http://acutm.math.ut.ee Behaviour of multivariate tail dependence coefficients Gaida

More information

ON UNIFORM TAIL EXPANSIONS OF BIVARIATE COPULAS

ON UNIFORM TAIL EXPANSIONS OF BIVARIATE COPULAS APPLICATIONES MATHEMATICAE 31,4 2004), pp. 397 415 Piotr Jaworski Warszawa) ON UNIFORM TAIL EXPANSIONS OF BIVARIATE COPULAS Abstract. The theory of copulas provides a useful tool for modelling dependence

More information

A Brief Introduction to Copulas

A Brief Introduction to Copulas A Brief Introduction to Copulas Speaker: Hua, Lei February 24, 2009 Department of Statistics University of British Columbia Outline Introduction Definition Properties Archimedean Copulas Constructing Copulas

More information

Explicit Bounds for the Distribution Function of the Sum of Dependent Normally Distributed Random Variables

Explicit Bounds for the Distribution Function of the Sum of Dependent Normally Distributed Random Variables Explicit Bounds for the Distribution Function of the Sum of Dependent Normally Distributed Random Variables Walter Schneider July 26, 20 Abstract In this paper an analytic expression is given for the bounds

More information

Modelling Dependence with Copulas and Applications to Risk Management. Filip Lindskog, RiskLab, ETH Zürich

Modelling Dependence with Copulas and Applications to Risk Management. Filip Lindskog, RiskLab, ETH Zürich Modelling Dependence with Copulas and Applications to Risk Management Filip Lindskog, RiskLab, ETH Zürich 02-07-2000 Home page: http://www.math.ethz.ch/ lindskog E-mail: lindskog@math.ethz.ch RiskLab:

More information

Tail Approximation of Value-at-Risk under Multivariate Regular Variation

Tail Approximation of Value-at-Risk under Multivariate Regular Variation Tail Approximation of Value-at-Risk under Multivariate Regular Variation Yannan Sun Haijun Li July 00 Abstract This paper presents a general tail approximation method for evaluating the Valueat-Risk of

More information

Copulas and Measures of Dependence

Copulas and Measures of Dependence 1 Copulas and Measures of Dependence Uttara Naik-Nimbalkar December 28, 2014 Measures for determining the relationship between two variables: the Pearson s correlation coefficient, Kendalls tau and Spearmans

More information

Copulas and dependence measurement

Copulas and dependence measurement Copulas and dependence measurement Thorsten Schmidt. Chemnitz University of Technology, Mathematical Institute, Reichenhainer Str. 41, Chemnitz. thorsten.schmidt@mathematik.tu-chemnitz.de Keywords: copulas,

More information

Multivariate Operational Risk: Dependence Modelling with Lévy Copulas

Multivariate Operational Risk: Dependence Modelling with Lévy Copulas Multivariate Operational Risk: Dependence Modelling with Lévy Copulas Klaus Böcker Claudia Klüppelberg Abstract Simultaneous modelling of operational risks occurring in different event type/business line

More information

Multivariate Measures of Positive Dependence

Multivariate Measures of Positive Dependence Int. J. Contemp. Math. Sciences, Vol. 4, 2009, no. 4, 191-200 Multivariate Measures of Positive Dependence Marta Cardin Department of Applied Mathematics University of Venice, Italy mcardin@unive.it Abstract

More information

Modelling Dependent Credit Risks

Modelling Dependent Credit Risks Modelling Dependent Credit Risks Filip Lindskog, RiskLab, ETH Zürich 30 November 2000 Home page:http://www.math.ethz.ch/ lindskog E-mail:lindskog@math.ethz.ch RiskLab:http://www.risklab.ch Modelling Dependent

More information

A measure of radial asymmetry for bivariate copulas based on Sobolev norm

A measure of radial asymmetry for bivariate copulas based on Sobolev norm A measure of radial asymmetry for bivariate copulas based on Sobolev norm Ahmad Alikhani-Vafa Ali Dolati Abstract The modified Sobolev norm is used to construct an index for measuring the degree of radial

More information

VaR bounds in models with partial dependence information on subgroups

VaR bounds in models with partial dependence information on subgroups VaR bounds in models with partial dependence information on subgroups L. Rüschendorf J. Witting February 23, 2017 Abstract We derive improved estimates for the model risk of risk portfolios when additional

More information

Copulas. Mathematisches Seminar (Prof. Dr. D. Filipovic) Di Uhr in E

Copulas. Mathematisches Seminar (Prof. Dr. D. Filipovic) Di Uhr in E Copulas Mathematisches Seminar (Prof. Dr. D. Filipovic) Di. 14-16 Uhr in E41 A Short Introduction 1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 The above picture shows a scatterplot (500 points) from a pair

More information

Stochastic orders: a brief introduction and Bruno s contributions. Franco Pellerey

Stochastic orders: a brief introduction and Bruno s contributions. Franco Pellerey Stochastic orders: a brief introduction and Bruno s contributions. Franco Pellerey Stochastic orders (comparisons) Among his main interests in research activity A field where his contributions are still

More information

A simple graphical method to explore tail-dependence in stock-return pairs

A simple graphical method to explore tail-dependence in stock-return pairs A simple graphical method to explore tail-dependence in stock-return pairs Klaus Abberger, University of Konstanz, Germany Abstract: For a bivariate data set the dependence structure can not only be measured

More information

Quasi-copulas and signed measures

Quasi-copulas and signed measures Quasi-copulas and signed measures Roger B. Nelsen Department of Mathematical Sciences, Lewis & Clark College, Portland (USA) José Juan Quesada-Molina Department of Applied Mathematics, University of Granada

More information

Tail Mutual Exclusivity and Tail- Var Lower Bounds

Tail Mutual Exclusivity and Tail- Var Lower Bounds Tail Mutual Exclusivity and Tail- Var Lower Bounds Ka Chun Cheung, Michel Denuit, Jan Dhaene AFI_15100 TAIL MUTUAL EXCLUSIVITY AND TAIL-VAR LOWER BOUNDS KA CHUN CHEUNG Department of Statistics and Actuarial

More information

VaR vs. Expected Shortfall

VaR vs. Expected Shortfall VaR vs. Expected Shortfall Risk Measures under Solvency II Dietmar Pfeifer (2004) Risk measures and premium principles a comparison VaR vs. Expected Shortfall Dependence and its implications for risk measures

More information

Risk Aggregation with Dependence Uncertainty

Risk Aggregation with Dependence Uncertainty Introduction Extreme Scenarios Asymptotic Behavior Challenges Risk Aggregation with Dependence Uncertainty Department of Statistics and Actuarial Science University of Waterloo, Canada Seminar at ETH Zurich

More information

Introduction to Dependence Modelling

Introduction to Dependence Modelling Introduction to Dependence Modelling Carole Bernard Berlin, May 2015. 1 Outline Modeling Dependence Part 1: Introduction 1 General concepts on dependence. 2 in 2 or N 3 dimensions. 3 Minimizing the expectation

More information

GENERAL MULTIVARIATE DEPENDENCE USING ASSOCIATED COPULAS

GENERAL MULTIVARIATE DEPENDENCE USING ASSOCIATED COPULAS REVSTAT Statistical Journal Volume 14, Number 1, February 2016, 1 28 GENERAL MULTIVARIATE DEPENDENCE USING ASSOCIATED COPULAS Author: Yuri Salazar Flores Centre for Financial Risk, Macquarie University,

More information

Statistics and Probability Letters

Statistics and Probability Letters tatistics Probability Letters 80 200) 473 479 Contents lists available at ciencedirect tatistics Probability Letters journal homepage: www.elsevier.com/locate/stapro On the relationships between copulas

More information

ESTIMATING BIVARIATE TAIL

ESTIMATING BIVARIATE TAIL Elena DI BERNARDINO b joint work with Clémentine PRIEUR a and Véronique MAUME-DESCHAMPS b a LJK, Université Joseph Fourier, Grenoble 1 b Laboratoire SAF, ISFA, Université Lyon 1 Framework Goal: estimating

More information

Tail Dependence of Multivariate Pareto Distributions

Tail Dependence of Multivariate Pareto Distributions !#"%$ & ' ") * +!-,#. /10 243537698:6 ;=@?A BCDBFEHGIBJEHKLB MONQP RS?UTV=XW>YZ=eda gihjlknmcoqprj stmfovuxw yy z {} ~ ƒ }ˆŠ ~Œ~Ž f ˆ ` š œžÿ~ ~Ÿ œ } ƒ œ ˆŠ~ œ

More information

Risk Aggregation. Paul Embrechts. Department of Mathematics, ETH Zurich Senior SFI Professor.

Risk Aggregation. Paul Embrechts. Department of Mathematics, ETH Zurich Senior SFI Professor. Risk Aggregation Paul Embrechts Department of Mathematics, ETH Zurich Senior SFI Professor www.math.ethz.ch/~embrechts/ Joint work with P. Arbenz and G. Puccetti 1 / 33 The background Query by practitioner

More information

Convolution Based Unit Root Processes: a Simulation Approach

Convolution Based Unit Root Processes: a Simulation Approach International Journal of Statistics and Probability; Vol., No. 6; November 26 ISSN 927-732 E-ISSN 927-74 Published by Canadian Center of Science and Education Convolution Based Unit Root Processes: a Simulation

More information

Kybernetika. Sebastian Fuchs Multivariate copulas: Transformations, symmetry, order and measures of concordance

Kybernetika. Sebastian Fuchs Multivariate copulas: Transformations, symmetry, order and measures of concordance Kybernetika Sebastian Fuchs Multivariate copulas: Transformations, symmetry, order and measures of concordance Kybernetika, Vol. 50 2014, No. 5, 725 743 Persistent URL: http://dml.cz/dmlcz/144103 Terms

More information

Clearly, if F is strictly increasing it has a single quasi-inverse, which equals the (ordinary) inverse function F 1 (or, sometimes, F 1 ).

Clearly, if F is strictly increasing it has a single quasi-inverse, which equals the (ordinary) inverse function F 1 (or, sometimes, F 1 ). APPENDIX A SIMLATION OF COPLAS Copulas have primary and direct applications in the simulation of dependent variables. We now present general procedures to simulate bivariate, as well as multivariate, dependent

More information

Asymptotics of random sums of heavy-tailed negatively dependent random variables with applications

Asymptotics of random sums of heavy-tailed negatively dependent random variables with applications Asymptotics of random sums of heavy-tailed negatively dependent random variables with applications Remigijus Leipus (with Yang Yang, Yuebao Wang, Jonas Šiaulys) CIRM, Luminy, April 26-30, 2010 1. Preliminaries

More information

ASSOCIATIVE n DIMENSIONAL COPULAS

ASSOCIATIVE n DIMENSIONAL COPULAS K Y BERNETIKA VOLUM E 47 ( 2011), NUMBER 1, P AGES 93 99 ASSOCIATIVE n DIMENSIONAL COPULAS Andrea Stupňanová and Anna Kolesárová The associativity of n-dimensional copulas in the sense of Post is studied.

More information

Overview of Extreme Value Theory. Dr. Sawsan Hilal space

Overview of Extreme Value Theory. Dr. Sawsan Hilal space Overview of Extreme Value Theory Dr. Sawsan Hilal space Maths Department - University of Bahrain space November 2010 Outline Part-1: Univariate Extremes Motivation Threshold Exceedances Part-2: Bivariate

More information

S7. Copulas, Markov Operators and Mass Transportation

S7. Copulas, Markov Operators and Mass Transportation S7. Copulas, Markov Operators and Mass Transportation Organizers: Enrique de Amo (Universidad de Almería, Spain) Carlo Sempi (Università del Salento, Italy) Program: 1. Fabrizio Durante (Free University

More information

Frontier estimation based on extreme risk measures

Frontier estimation based on extreme risk measures Frontier estimation based on extreme risk measures by Jonathan EL METHNI in collaboration with Ste phane GIRARD & Laurent GARDES CMStatistics 2016 University of Seville December 2016 1 Risk measures 2

More information

On tail dependence coecients of transformed multivariate Archimedean copulas

On tail dependence coecients of transformed multivariate Archimedean copulas Tails and for Archim Copula () February 2015, University of Lille 3 On tail dependence coecients of transformed multivariate Archimedean copulas Elena Di Bernardino, CNAM, Paris, Département IMATH Séminaire

More information

Financial Econometrics and Volatility Models Copulas

Financial Econometrics and Volatility Models Copulas Financial Econometrics and Volatility Models Copulas Eric Zivot Updated: May 10, 2010 Reading MFTS, chapter 19 FMUND, chapters 6 and 7 Introduction Capturing co-movement between financial asset returns

More information

Randomly weighted sums under a wide type of dependence structure with application to conditional tail expectation

Randomly weighted sums under a wide type of dependence structure with application to conditional tail expectation Randomly weighted sums under a wide type of dependence structure with application to conditional tail expectation Shijie Wang a, Yiyu Hu a, Lianqiang Yang a, Wensheng Wang b a School of Mathematical Sciences,

More information

Non parametric estimation of Archimedean copulas and tail dependence. Paris, february 19, 2015.

Non parametric estimation of Archimedean copulas and tail dependence. Paris, february 19, 2015. Non parametric estimation of Archimedean copulas and tail dependence Elena Di Bernardino a and Didier Rullière b Paris, february 19, 2015. a CNAM, Paris, Département IMATH, b ISFA, Université Lyon 1, Laboratoire

More information

REMARKS ON TWO PRODUCT LIKE CONSTRUCTIONS FOR COPULAS

REMARKS ON TWO PRODUCT LIKE CONSTRUCTIONS FOR COPULAS K Y B E R N E T I K A V O L U M E 4 3 2 0 0 7 ), N U M B E R 2, P A G E S 2 3 5 2 4 4 REMARKS ON TWO PRODUCT LIKE CONSTRUCTIONS FOR COPULAS Fabrizio Durante, Erich Peter Klement, José Juan Quesada-Molina

More information

CONTAGION VERSUS FLIGHT TO QUALITY IN FINANCIAL MARKETS

CONTAGION VERSUS FLIGHT TO QUALITY IN FINANCIAL MARKETS EVA IV, CONTAGION VERSUS FLIGHT TO QUALITY IN FINANCIAL MARKETS Jose Olmo Department of Economics City University, London (joint work with Jesús Gonzalo, Universidad Carlos III de Madrid) 4th Conference

More information

Estimating Bivariate Tail: a copula based approach

Estimating Bivariate Tail: a copula based approach Estimating Bivariate Tail: a copula based approach Elena Di Bernardino, Université Lyon 1 - ISFA, Institut de Science Financiere et d'assurances - AST&Risk (ANR Project) Joint work with Véronique Maume-Deschamps

More information

arxiv: v1 [q-fin.rm] 11 Mar 2015

arxiv: v1 [q-fin.rm] 11 Mar 2015 Negative Dependence Concept in Copulas and the Marginal Free Herd Behavior Index Jae Youn Ahn a, a Department of Statistics, Ewha Womans University, 11-1 Daehyun-Dong, Seodaemun-Gu, Seoul 10-750, Korea.

More information

arxiv: v1 [math.pr] 9 Jan 2016

arxiv: v1 [math.pr] 9 Jan 2016 SKLAR S THEOREM IN AN IMPRECISE SETTING IGNACIO MONTES, ENRIQUE MIRANDA, RENATO PELESSONI, AND PAOLO VICIG arxiv:1601.02121v1 [math.pr] 9 Jan 2016 Abstract. Sklar s theorem is an important tool that connects

More information

Dependence and Order in Families of Archimedean Copulas

Dependence and Order in Families of Archimedean Copulas journal of multivariate analysis 60, 111122 (1997) article no. MV961646 Dependence and Order in Families of Archimedean Copulas Roger B. Nelsen* Lewis 6 Clark College The copula for a bivariate distribution

More information

A Measure of Monotonicity of Two Random Variables

A Measure of Monotonicity of Two Random Variables Journal of Mathematics and Statistics 8 (): -8, 0 ISSN 549-3644 0 Science Publications A Measure of Monotonicity of Two Random Variables Farida Kachapova and Ilias Kachapov School of Computing and Mathematical

More information

On Kusuoka Representation of Law Invariant Risk Measures

On Kusuoka Representation of Law Invariant Risk Measures MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 1, February 213, pp. 142 152 ISSN 364-765X (print) ISSN 1526-5471 (online) http://dx.doi.org/1.1287/moor.112.563 213 INFORMS On Kusuoka Representation of

More information

Operational Risk and Pareto Lévy Copulas

Operational Risk and Pareto Lévy Copulas Operational Risk and Pareto Lévy Copulas Claudia Klüppelberg Technische Universität München email: cklu@ma.tum.de http://www-m4.ma.tum.de References: - Böcker, K. and Klüppelberg, C. (25) Operational VaR

More information

Markov Switching Regular Vine Copulas

Markov Switching Regular Vine Copulas Int. Statistical Inst.: Proc. 58th World Statistical Congress, 2011, Dublin (Session CPS057) p.5304 Markov Switching Regular Vine Copulas Stöber, Jakob and Czado, Claudia Lehrstuhl für Mathematische Statistik,

More information

A PRACTICAL WAY FOR ESTIMATING TAIL DEPENDENCE FUNCTIONS

A PRACTICAL WAY FOR ESTIMATING TAIL DEPENDENCE FUNCTIONS Statistica Sinica 20 2010, 365-378 A PRACTICAL WAY FOR ESTIMATING TAIL DEPENDENCE FUNCTIONS Liang Peng Georgia Institute of Technology Abstract: Estimating tail dependence functions is important for applications

More information

Lecture 2 One too many inequalities

Lecture 2 One too many inequalities University of Illinois Department of Economics Spring 2017 Econ 574 Roger Koenker Lecture 2 One too many inequalities In lecture 1 we introduced some of the basic conceptual building materials of the course.

More information

Multivariate Distribution Models

Multivariate Distribution Models Multivariate Distribution Models Model Description While the probability distribution for an individual random variable is called marginal, the probability distribution for multiple random variables is

More information

Tail dependence in bivariate skew-normal and skew-t distributions

Tail dependence in bivariate skew-normal and skew-t distributions Tail dependence in bivariate skew-normal and skew-t distributions Paola Bortot Department of Statistical Sciences - University of Bologna paola.bortot@unibo.it Abstract: Quantifying dependence between

More information

On a Class of Multidimensional Optimal Transportation Problems

On a Class of Multidimensional Optimal Transportation Problems Journal of Convex Analysis Volume 10 (2003), No. 2, 517 529 On a Class of Multidimensional Optimal Transportation Problems G. Carlier Université Bordeaux 1, MAB, UMR CNRS 5466, France and Université Bordeaux

More information

Asymptotic Bounds for the Distribution of the Sum of Dependent Random Variables

Asymptotic Bounds for the Distribution of the Sum of Dependent Random Variables Asymptotic Bounds for the Distribution of the Sum of Dependent Random Variables Ruodu Wang November 26, 2013 Abstract Suppose X 1,, X n are random variables with the same known marginal distribution F

More information

Sharp bounds on the VaR for sums of dependent risks

Sharp bounds on the VaR for sums of dependent risks Paul Embrechts Sharp bounds on the VaR for sums of dependent risks joint work with Giovanni Puccetti (university of Firenze, Italy) and Ludger Rüschendorf (university of Freiburg, Germany) Mathematical

More information

Additivity properties for Value-at-Risk under Archimedean dependence and heavy-tailedness

Additivity properties for Value-at-Risk under Archimedean dependence and heavy-tailedness Additivity properties for Value-at-Risk under Archimedean dependence and heavy-tailedness Paul Embrechts, Johanna Nešlehová, Mario V. Wüthrich Abstract Mainly due to new capital adequacy standards for

More information

Theorem 2.1 (Caratheodory). A (countably additive) probability measure on a field has an extension. n=1

Theorem 2.1 (Caratheodory). A (countably additive) probability measure on a field has an extension. n=1 Chapter 2 Probability measures 1. Existence Theorem 2.1 (Caratheodory). A (countably additive) probability measure on a field has an extension to the generated σ-field Proof of Theorem 2.1. Let F 0 be

More information

Sklar s theorem in an imprecise setting

Sklar s theorem in an imprecise setting Sklar s theorem in an imprecise setting Ignacio Montes a,, Enrique Miranda a, Renato Pelessoni b, Paolo Vicig b a University of Oviedo (Spain), Dept. of Statistics and O.R. b University of Trieste (Italy),

More information

Coherent risk measures

Coherent risk measures Coherent risk measures Foivos Xanthos Ryerson University, Department of Mathematics Toµɛας Mαθηµατ ικὼν, E.M.Π, 11 Noɛµβρὶoυ 2015 Research interests Financial Mathematics, Mathematical Economics, Functional

More information

Lifetime Dependence Modelling using a Generalized Multivariate Pareto Distribution

Lifetime Dependence Modelling using a Generalized Multivariate Pareto Distribution Lifetime Dependence Modelling using a Generalized Multivariate Pareto Distribution Daniel Alai Zinoviy Landsman Centre of Excellence in Population Ageing Research (CEPAR) School of Mathematics, Statistics

More information

Tail comonotonicity: properties, constructions, and asymptotic additivity of risk measures

Tail comonotonicity: properties, constructions, and asymptotic additivity of risk measures Tail comonotonicity: properties, constructions, and asymptotic additivity of risk measures Lei Hua Harry Joe June 5, 2012 Abstract. We investigate properties of a version of tail comonotonicity that can

More information

Asymptotic behaviour of multivariate default probabilities and default correlations under stress

Asymptotic behaviour of multivariate default probabilities and default correlations under stress Asymptotic behaviour of multivariate default probabilities and default correlations under stress 7th General AMaMeF and Swissquote Conference EPFL, Lausanne Natalie Packham joint with Michael Kalkbrener

More information

A Goodness-of-fit Test for Copulas

A Goodness-of-fit Test for Copulas A Goodness-of-fit Test for Copulas Artem Prokhorov August 2008 Abstract A new goodness-of-fit test for copulas is proposed. It is based on restrictions on certain elements of the information matrix and

More information

BIVARIATE P-BOXES AND MAXITIVE FUNCTIONS. Keywords: Uni- and bivariate p-boxes, maxitive functions, focal sets, comonotonicity,

BIVARIATE P-BOXES AND MAXITIVE FUNCTIONS. Keywords: Uni- and bivariate p-boxes, maxitive functions, focal sets, comonotonicity, BIVARIATE P-BOXES AND MAXITIVE FUNCTIONS IGNACIO MONTES AND ENRIQUE MIRANDA Abstract. We give necessary and sufficient conditions for a maxitive function to be the upper probability of a bivariate p-box,

More information

Modelling and Estimation of Stochastic Dependence

Modelling and Estimation of Stochastic Dependence Modelling and Estimation of Stochastic Dependence Uwe Schmock Based on joint work with Dr. Barbara Dengler Financial and Actuarial Mathematics and Christian Doppler Laboratory for Portfolio Risk Management

More information

Losses Given Default in the Presence of Extreme Risks

Losses Given Default in the Presence of Extreme Risks Losses Given Default in the Presence of Extreme Risks Qihe Tang [a] and Zhongyi Yuan [b] [a] Department of Statistics and Actuarial Science University of Iowa [b] Smeal College of Business Pennsylvania

More information

Regularly Varying Asymptotics for Tail Risk

Regularly Varying Asymptotics for Tail Risk Regularly Varying Asymptotics for Tail Risk Haijun Li Department of Mathematics Washington State University Humboldt Univ-Berlin Haijun Li Regularly Varying Asymptotics for Tail Risk Humboldt Univ-Berlin

More information

The Instability of Correlations: Measurement and the Implications for Market Risk

The Instability of Correlations: Measurement and the Implications for Market Risk The Instability of Correlations: Measurement and the Implications for Market Risk Prof. Massimo Guidolin 20254 Advanced Quantitative Methods for Asset Pricing and Structuring Winter/Spring 2018 Threshold

More information

arxiv: v2 [math.pr] 23 Jun 2014

arxiv: v2 [math.pr] 23 Jun 2014 COMPUTATION OF COPULAS BY FOURIER METHODS ANTONIS PAPAPANTOLEON arxiv:08.26v2 [math.pr] 23 Jun 204 Abstract. We provide an integral representation for the (implied) copulas of dependent random variables

More information

Calculating credit risk capital charges with the one-factor model

Calculating credit risk capital charges with the one-factor model Calculating credit risk capital charges with the one-factor model Susanne Emmer Dirk Tasche September 15, 2003 Abstract Even in the simple Vasicek one-factor credit portfolio model, the exact contributions

More information

Weak max-sum equivalence for dependent heavy-tailed random variables

Weak max-sum equivalence for dependent heavy-tailed random variables DOI 10.1007/s10986-016-9303-6 Lithuanian Mathematical Journal, Vol. 56, No. 1, January, 2016, pp. 49 59 Wea max-sum equivalence for dependent heavy-tailed random variables Lina Dindienė a and Remigijus

More information

Contents 1. Coping with Copulas. Thorsten Schmidt 1. Department of Mathematics, University of Leipzig Dec 2006

Contents 1. Coping with Copulas. Thorsten Schmidt 1. Department of Mathematics, University of Leipzig Dec 2006 Contents 1 Coping with Copulas Thorsten Schmidt 1 Department of Mathematics, University of Leipzig Dec 2006 Forthcoming in Risk Books Copulas - From Theory to Applications in Finance Contents 1 Introdcution

More information

The multivariate probability integral transform

The multivariate probability integral transform The multivariate probability integral transform Fabrizio Durante Faculty of Economics and Management Free University of Bozen-Bolzano (Italy) fabrizio.durante@unibz.it http://sites.google.com/site/fbdurante

More information

Competitive Equilibria in a Comonotone Market

Competitive Equilibria in a Comonotone Market Competitive Equilibria in a Comonotone Market 1/51 Competitive Equilibria in a Comonotone Market Ruodu Wang http://sas.uwaterloo.ca/ wang Department of Statistics and Actuarial Science University of Waterloo

More information

Relations Between Hidden Regular Variation and Tail Order of. Copulas

Relations Between Hidden Regular Variation and Tail Order of. Copulas Relations Between Hidden Regular Variation and Tail Order of Copulas Lei Hua Harry Joe Haijun Li December 28, 2012 Abstract We study the relations between tail order of copulas and hidden regular variation

More information

Politecnico di Torino. Porto Institutional Repository

Politecnico di Torino. Porto Institutional Repository Politecnico di Torino Porto Institutional Repository [Article] On preservation of ageing under minimum for dependent random lifetimes Original Citation: Pellerey F.; Zalzadeh S. (204). On preservation

More information

Reinsurance and ruin problem: asymptotics in the case of heavy-tailed claims

Reinsurance and ruin problem: asymptotics in the case of heavy-tailed claims Reinsurance and ruin problem: asymptotics in the case of heavy-tailed claims Serguei Foss Heriot-Watt University, Edinburgh Karlovasi, Samos 3 June, 2010 (Joint work with Tomasz Rolski and Stan Zachary

More information

ON DISTRIBUTIONS OF ORDER STATISTICS FOR ABSOLUTELY CONTINUOUS COPULAS WITH APPLICATIONS TO RELIABILITY

ON DISTRIBUTIONS OF ORDER STATISTICS FOR ABSOLUTELY CONTINUOUS COPULAS WITH APPLICATIONS TO RELIABILITY K Y B E R N E T I K A V O L U M E 4 4 2 0 0 8 ), N U M B E R 6, P A G E S 7 5 7 7 7 6 ON DISTRIBUTIONS OF ORDER STATISTICS FOR ABSOLUTELY CONTINUOUS COPULAS WITH APPLICATIONS TO RELIABILITY Piotr Jaworski

More information

Tail negative dependence and its applications for aggregate loss modeling

Tail negative dependence and its applications for aggregate loss modeling Tail negative dependence and its applications for aggregate loss modeling Lei Hua Division of Statistics Oct 20, 2014, ISU L. Hua (NIU) 1/35 1 Motivation 2 Tail order Elliptical copula Extreme value copula

More information

Asymptotic Tail Probabilities of Sums of Dependent Subexponential Random Variables

Asymptotic Tail Probabilities of Sums of Dependent Subexponential Random Variables Asymptotic Tail Probabilities of Sums of Dependent Subexponential Random Variables Jaap Geluk 1 and Qihe Tang 2 1 Department of Mathematics The Petroleum Institute P.O. Box 2533, Abu Dhabi, United Arab

More information

Risk Aggregation. 1 Motivations and preliminaries. Paul Embrechts and Giovanni Puccetti

Risk Aggregation. 1 Motivations and preliminaries. Paul Embrechts and Giovanni Puccetti Risk Aggregation Paul Embrechts and Giovanni Puccetti 1 Motivations and preliminaries Quantitative Risk Management (QRM) standardly concerns a vector of one-period profit-and-loss random variables X =

More information

The main results about probability measures are the following two facts:

The main results about probability measures are the following two facts: Chapter 2 Probability measures The main results about probability measures are the following two facts: Theorem 2.1 (extension). If P is a (continuous) probability measure on a field F 0 then it has a

More information

Operational Risk and Pareto Lévy Copulas

Operational Risk and Pareto Lévy Copulas Operational Risk and Pareto Lévy Copulas Claudia Klüppelberg Technische Universität München email: cklu@ma.tum.de http://www-m4.ma.tum.de References: - Böcker, K. and Klüppelberg, C. (25) Operational VaR

More information

Characterization of Upper Comonotonicity via Tail Convex Order

Characterization of Upper Comonotonicity via Tail Convex Order Characterization of Upper Comonotonicity via Tail Convex Order Hee Seok Nam a,, Qihe Tang a, Fan Yang b a Department of Statistics and Actuarial Science, University of Iowa, 241 Schaeffer Hall, Iowa City,

More information

Supermodular ordering of Poisson arrays

Supermodular ordering of Poisson arrays Supermodular ordering of Poisson arrays Bünyamin Kızıldemir Nicolas Privault Division of Mathematical Sciences School of Physical and Mathematical Sciences Nanyang Technological University 637371 Singapore

More information

Copulas with given diagonal section: some new results

Copulas with given diagonal section: some new results Copulas with given diagonal section: some new results Fabrizio Durante Dipartimento di Matematica Ennio De Giorgi Università di Lecce Lecce, Italy 73100 fabrizio.durante@unile.it Radko Mesiar STU Bratislava,

More information

A multivariate dependence measure for aggregating risks

A multivariate dependence measure for aggregating risks A multivariate dependence measure for aggregating risks Jan Dhaene 1 Daniël Linders 2 Wim Schoutens 3 David Vyncke 4 December 1, 2013 1 KU Leuven, Leuven, Belgium. Email: jan.dhaene@econ.kuleuven.be 2

More information

Parameter estimation of a Lévy copula of a discretely observed bivariate compound Poisson process with an application to operational risk modelling

Parameter estimation of a Lévy copula of a discretely observed bivariate compound Poisson process with an application to operational risk modelling Parameter estimation of a Lévy copula of a discretely observed bivariate compound Poisson process with an application to operational risk modelling J. L. van Velsen 1,2 arxiv:1212.0092v1 [q-fin.rm] 1 Dec

More information

The distribution of a sum of dependent risks: a geometric-combinatorial approach

The distribution of a sum of dependent risks: a geometric-combinatorial approach The distribution of a sum of dependent risks: a geometric-combinatorial approach Marcello Galeotti 1, Emanuele Vannucci 2 1 University of Florence, marcello.galeotti@dmd.unifi.it 2 University of Pisa,

More information

Convex Analysis and Economic Theory AY Elementary properties of convex functions

Convex Analysis and Economic Theory AY Elementary properties of convex functions Division of the Humanities and Social Sciences Ec 181 KC Border Convex Analysis and Economic Theory AY 2018 2019 Topic 6: Convex functions I 6.1 Elementary properties of convex functions We may occasionally

More information

Dependence. Practitioner Course: Portfolio Optimization. John Dodson. September 10, Dependence. John Dodson. Outline.

Dependence. Practitioner Course: Portfolio Optimization. John Dodson. September 10, Dependence. John Dodson. Outline. Practitioner Course: Portfolio Optimization September 10, 2008 Before we define dependence, it is useful to define Random variables X and Y are independent iff For all x, y. In particular, F (X,Y ) (x,

More information

Estimation of multivariate critical layers: Applications to rainfall data

Estimation of multivariate critical layers: Applications to rainfall data Elena Di Bernardino, ICRA 6 / RISK 2015 () Estimation of Multivariate critical layers Barcelona, May 26-29, 2015 Estimation of multivariate critical layers: Applications to rainfall data Elena Di Bernardino,

More information

Simulating Exchangeable Multivariate Archimedean Copulas and its Applications. Authors: Florence Wu Emiliano A. Valdez Michael Sherris

Simulating Exchangeable Multivariate Archimedean Copulas and its Applications. Authors: Florence Wu Emiliano A. Valdez Michael Sherris Simulating Exchangeable Multivariate Archimedean Copulas and its Applications Authors: Florence Wu Emiliano A. Valdez Michael Sherris Literatures Frees and Valdez (1999) Understanding Relationships Using

More information

Estimation of Operational Risk Capital Charge under Parameter Uncertainty

Estimation of Operational Risk Capital Charge under Parameter Uncertainty Estimation of Operational Risk Capital Charge under Parameter Uncertainty Pavel V. Shevchenko Principal Research Scientist, CSIRO Mathematical and Information Sciences, Sydney, Locked Bag 17, North Ryde,

More information

Risk Aggregation and Model Uncertainty

Risk Aggregation and Model Uncertainty Risk Aggregation and Model Uncertainty Paul Embrechts RiskLab, Department of Mathematics, ETH Zurich Senior SFI Professor www.math.ethz.ch/ embrechts/ Joint work with A. Beleraj, G. Puccetti and L. Rüschendorf

More information

An axiomatic characterization of capital allocations of coherent risk measures

An axiomatic characterization of capital allocations of coherent risk measures An axiomatic characterization of capital allocations of coherent risk measures Michael Kalkbrener Deutsche Bank AG Abstract An axiomatic definition of coherent capital allocations is given. It is shown

More information

Jónsson posets and unary Jónsson algebras

Jónsson posets and unary Jónsson algebras Jónsson posets and unary Jónsson algebras Keith A. Kearnes and Greg Oman Abstract. We show that if P is an infinite poset whose proper order ideals have cardinality strictly less than P, and κ is a cardinal

More information