Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Study on the Dependence Conditional Risk in Financial Portfolio YI Wende1, HUANG Aihua2 1. Dept. of Math & Computer Science, Chongqing University of Arts and Sciences; 2. Dept. of planning finance, Chongqing University of Arts and Sciences Yongchuan Chongqing 402160, China [email protected] Abstract: It’s very interesting to investigate the dependence structure and measures of conditional dependence risk of time series. In this paper, some dependence probabilities and conditional dependence risk density functions are proposed to study the conditional dependence risk of time series. Keywords: Copula; Conditional dependence; Conditional density; Conditional probability 1 Introduction Dependence risk measurement in portfolio’s risk management is one of the most important issues. It is well known that linear correlation has some deficiencies that one is that it is not invariant under non-linear strictly increasing transformation, and the other is that it can not measure the dependence structure of two time series, such as non-linear correlation, kurtosis, fat-tail and tail-dependence. While the dependence measures derived from copulas can overcome this shortcoming. From the time when Embrechts et al. (2001)[1] proposed the application of copula functions in order to model the multivariate distribution of returns, the copula functions have had broader applications in finance and the theory of copulas has experienced rapid development[2,3,4] with the development of computer software and information technology. This paper presents some conditional dependence risk probabilities and conditional dependence risk density functions of time series using copula functions to investigate the conditional dependence risk of financial portfolio assets. Our research focuses on the condition such as X = x or Y = y and X ≤ x or Y ≤ y and X > x or Y > y or ( X > x, Y > y ) . The structure of this paper is as follows. Section 2 introduces the notation of copula and some tail dependence measures. Section 3 presents several conditional dependence risk probability functions and conditional dependence risk density functions for studying the conditional dependence structure and the conditional dependence degree of portfolio assets. Section 4 concludes the paper. 2 Copulas and Tail Dependence Measures 2.1 Copulas The most interesting fact in the copula framework is the Sklar’s (1959)[2] theorem. A copula function is defined as a multivariate distribution with standard uniform marginal distributions or as a function that maps values from the unit hypercube to value in the unit interval. If the joint distribution of X and Y is H ( ⋅, ⋅ ) with continuous marginal distributions F ( ⋅ ) and G ( ⋅ ) , respectively, then there exists a unique copula C ( ⋅, ⋅ ) such that the joint distribution can be written as H ( x , y ) = C ( F ( x ) , G ( y ) ) , ( x , y ) ∈ ℜ2 . From this expression, we can see that any copula C ( ⋅, ⋅ ) pertaining to function H ( ⋅, ⋅ ) can also be expressed as C ( u , v ) = H ( F −1 ( u ) , G −1 ( v ) ) , 620 [ ] where u = F ( x ) , v = G ( y ) , u , v ∈ 0,1 . 2.2 Tail Dependence Measures The dependence structure of X and Y is characterized by the copula, moreover, the measure of dependence is invariant under strictly increasing transformations of random variables. For example, Kendall’s tau (τ ) and Spearman’s rho ( ρ ) are the most widely used scale invariant measure of concordance. (1) Tail dependence Another important concept related to copulas is the concepts of upper tail dependence and lower tail dependence. The upper tail dependence is defined as ( ) lim+ λU (α ) = lim+ P X > F −1 (1 − α ) Y > G −1 (1 − α ) , α →0 α →0 if this limit exists. The two variables X and Y are said to be asymptotically dependent in the upper tail if λU ∈ ( 0,1] , and asymptotically independent if λU = 0 . In finance the concept of upper tail dependence can be interpreted by λU = lim P ( X > VaRα ( X ) Y > VaRα (Y ) ) , α → 0+ where VaRα ( X ) = F −1 (1 − α ) is the value at risk. In a similar way, the lower tail dependence is defined as ( ) lim λL (α ) = lim+ P X < F −1 (α ) Y < G −1 (α ) . α → 0+ In terms of copulas, α →0 λU and λL can be expressed using the pertaining copula. 1 − 2u + C ( u , u ) , u →1 1− u C ( u, u ) . λL = lim u →0 u λU = lim (2) Conditional tail risk measures In risk management, VaR (value-at-risk) usually acts as an important tool of risk measurement. Briefly speaking, VaR is the maximal potential loss of a position and a portfolio in some investment horizon under a given confidential level and over a certain period. While VaR is a powerful tool for risk management, it is not a coherent risk measure and sub-additive. To overcome this problem, a conditional VaR (denote CVaR or ES)[5]is proposed to modify the risk measure of VaR. CVaRα = E X X > VaRα ( X ) , (1) where X is the variable representing the negative return (loss) of asset and FX is its distribution. Artzner et al. (1998)[6]proposed the concept of the expected shortfall eX ∗ shortfall eX (α ) (α ) which are respectively defined as eX (α ) = E X − VaRα ( X ) X > VaRα ( X ) , e∗X (α ) = Median X − VaRα ( X ) X > VaRα ( X ) . 3 Dependence Conditional Risk Density Functions 621 and the median (2) (3) Let random variables X and Y represent the negative returns (loss) value of two financial assets. If the joint distribution of X and Y is H ( ⋅, ⋅ ) with continuous marginal distributions F ( ⋅ ) and G ( ⋅ ) , and the unique copula C ( ⋅, ⋅ ) , whose density functions are f ( ⋅) , g ( ⋅) and c ( ⋅, ⋅ ) , respectively. Given certain values of X and Y , such as x and y , we consider some dependence risk density functions under the conditions such as X = x or Y = y and X ≤ x or Y ≤ y and X > x or Y > y or ( X > x, Y > y ) . These dependence risk density functions give the instantaneous probabilities for the individual security and the portfolio assets that have attained the values x and y . P { X ≤ x Y = y} = lim P { X ≤ x y − ∆y < Y ≤ y} ∆y → 0 P { X ≤ x, y − ∆y < Y ≤ y} ∆y → 0 P { y − ∆y < Y ≤ y} = lim x h ( u, t ) du dt ∫y −∆y ∫−∞ = lim y ∆y → 0 ∫ g ( t ) dt y y −∆y x = ∫ h ( u, y ) du −∞ g ( y) ∫ c ( F ( x ) , G ( y ) ) dF ( x ) . = x −∞ The function f ( x ) = X Y g ( y) c ( F ( x ) , G ( y )) f ( x ) g (y) (4) is called as the conditional density function under the condition Y = y . Similarly, we can obtain the conditional density function gY X ( y ) . P {X ≤ x Y ≤ y} = P { X ≤ x , Y ≤ y} H ( x , y ) C ( F ( x ) , G ( y ) ) . (5) = = G (y) G (y) G (y) f X Y( ( x ) = where C 1 ( u , v ) = ∂C (u , v ) ∂u C1 ( F ( x ) , G ( y ) ) f ( x ) G (y) (6) . P { x < X ≤ x + ∆ x X > x} = F ( x + ∆x ) − F ( x ) 1− F (x) ≈ The function , f ( x ) ∆x . 1− F (x) (7) f (x) is interpreted as a conditional probability density function: it gives the value of 1− F (x) the conditional probability density function of X at exact value x , given survival to that value. Let’s 622 denote it as f X X ( x ) , which is usually called the hazard rate function. In the follows, we discuss the dependence scenario of two assets. P { y < Y ≤ y + ∆y X > x} = = gY X ( y ) = lim+ X 1− F (x) ∂H ( x, y ) ∂C ( u , v ) , and C 2 ( u , v ) = . ∂y ∂v P {x < , (9) X ≤ x + ∆ x , y < Y ≤ y + ∆ y} 1 − F (x ) H ( x + ∆x , y + ∆ y ) − H ( x + ∆ x, y ) − H ( x, y + ∆ y ) + H ( x, y ) 1− F (x) = hXY g ( y ) − C 2 ( F ( x ) , G ( y )) g ( y ) < X ≤ x + ∆ x, y < Y ≤ y + ∆ y X > x} = = .(8) ∆y = {x 1− F (x) g ( y ) − H Y ( x, y ) 1− F (x) = P G ( y + ∆ y ) − G ( y ) − H ( x , y + ∆ y ) − H ( x , y ) P { y < Y ≤ y + ∆y X > x} ∆y → 0 where H Y ( x, y ) = P { y < Y ≤ y + ∆ y } − P { X ≤ x , y < Y ≤ y + ∆ y} 1− F (x) (x, y ) = h ( x , y ) ∆ x ∆y . 1− F ( x) lim + (10) P (x < X ≤ x + ∆x, y < Y ≤ y + ∆y X > x ) ∆ x∆ y ∆x→ 0 ∆y → 0+ = h ( x, y ) 1 − F (x) = c ( F ( x ) , G ( y )) f 1− F (x) (x ) g ( y) . (11) Similarly, the corresponding density functions under the condition Y > y can be calculated, omitted here. Now we consider the scenario under the condition ( X > x, Y > y ) . P (x < X ≤ x + ∆x X > x ,Y > y ) = F ( x + ∆ x ) − F ( x ) − H ( x + ∆ x , y ) − H 1 − F (x ) − G ( y ) + H (x, y ) = f X XY ( x ) = lim+ ∆x → 0 = f ( x ) ∆ x − H X ( x, y ) ∆ x . 1 − F ( x ) − G ( y ) + H ( x, y ) ( x , y ) (12) P ( x < X ≤ x + ∆x X > x , Y > y ) ∆x f ( x ) − C1 ( F ( x ) , G ( y ) ) f ( x ) 1 − F ( x ) − G ( y ) + C ( F ( x ), G ( y )) 623 . (13) P ( x < X ≤ x + ∆x, y < Y ≤ y + ∆ y X > x , Y > y ) = = P {x < X ≤ x + ∆ x, y < Y ≤ y + ∆y} 1 − F ( x ) − G ( y ) + H ( x, y ) h ( x, y ) ∆x ∆y 1 − F ( x ) − G ( y ) + H ( x, y ) = c ( F ( x ) , G ( y )) f ( x ) g ( y ) ∆ x ∆y 1 − F ( x ) − G ( y ) + C ( F ( x ), G ( y )) hXY XY ( x, y ) = lim+ (14) P ( x < X ≤ x + ∆x, y < Y ≤ y + ∆y X > x , Y > y ) ∆x∆y ∆x →0 ∆y → 0+ = ; c ( F ( x ) , G ( y )) f (x) g ( y) . 1 − F ( x ) − G ( y ) + C ( F ( x ), G ( y )) (15) In this section, we achieve some dependence conditional probability functions and its density functions. These results are very useful in risk management, since the risk analysts need to analyse the conditional dependence structure and conditional dependence measure of portfolio assets according to given risk conditions. Especially, the conditional dependence risk probability and density functions have particular meaning when the threshold values equal some certain values such as x = 0, y = 0 and x = VaRα ( X ) , y = VaRα ( Y ) . 4 Conclusions In this paper, we have studied the conditional dependence risk probability and the conditional dependence risk density functions. We mainly focus on the scenarios under the conditions such as X = x or Y = y and X ≤ x or Y ≤ y and X > x or Y > y or ( X > x, Y > y ) . A member of conditional dependence risk relationships is completely characterized by the marginal distribution and the copulas of random variables. These results can be applied to investigate the conditional dependence structure and the conditional dependence measure of portfolio assets. References [1]Embrechts, P., McNeil, A.J., Straumann, D.. Correlation and dependency in risk management: properties and pitfalls[M]. In: Dempster, M., Moffatt, H. (Eds.), Risk Management. Cambridge University Press, New York, pp. 176-223. [2]R. B. Nelsen. An Introduction to Copulas[M]. Springer, New York, 1998. [3]R. B. Nelsen. Dependence and order in families of Archimedean copulas[J]. Journal of Multivariate Analysis, 1997, 60, 111-122. [4]C. Genest, L. P. Rivest. On the multivariate probability integral transformation[J]. Statistics and Probability Letters 53(2001) 391-399. [5] C. Acerbi, D. Tasche, On the coherence of expected shortfall[J]. Journal of Banking and Finance, 2002,26,1487-1503. [6]Artzner, Delbaen, Eber, &Heath, Coherent measures of risk[J]. Mathematical Finance, 1998,9(3), 203-228. 624