Fast and accurate method for estimating portfolio CVaR risk

The present invention relates to portfolio risk management, and more particularly to calculating the Conditional Value-at-Risk (CVaR), a widely used risk measure, for a portfolio.

One of the main objectives of portfolio risk management is to evaluate and improve the performance of the portfolio while reducing exposure to a financial loss. A financial portfolio refers to a collection of investments owned by an individual or an organization. An investment includes, but is not limited to, a stock, a bond, a currency, a derivative, a mutual fund, a hedge fund, cash equivalents, etc. A risk refers to a likelihood of losing investment values in a portfolio. Estimating the risk of a portfolio through a simulation (e.g., Monte Carlo simulation or any other equivalent simulation), is a fundamental task in portfolio risk management. Different measures of risk call for different simulation techniques.

A standard benchmark for a measurement of a risk is “Value-at-Risk” (VaR). For a given confidence level β(0<β<1, typical β=95%), the β-level VaR is the loss in the portfolio's value that is exceeded with the probability 1−β. However, as a risk measure, VaR lacks coherency in the sense that it does not necessarily encourage diversification. This is because the VaR value of a combination of two portfolios can be greater than the sum of VaR values of the individual portfolios. Philippe Artzner, et al. “Coherent Measure of Risk,” Mathematical Finance, vol. 9, no. 3, July 1999, pp. 203-228, wholly incorporated by reference, describes VaR in detail.

An alternative risk measure to VaR is “Conditional Value-at-Risk” (CVaR), which is also known as “Average Value-at-Risk”, “Mean Excess Loss”, “Mean Shortfall” or “Tail VaR”. For a given level β, the β-level CVaR value is the conditional expectation of the loss above the β-level VaR value. The value of CVaR is always greater than or equal to that of the corresponding VaR. CVaR can be calculated by generating random samples to simulation losses of a portfolio, and then averaging those samples that are greater than the VaR value.

In one embodiment, the present invention describes a system, method and computer program product for measuring a risk of a portfolio.

In one embodiment, there is provided a system for measuring a risk of a portfolio. The system comprises at least one memory device and at least one processor connected to the memory device. The system estimates the CVaR (Conditional Value-at-Risk) of the portfolio.

In a further embodiment, there is provided a method for measuring a risk of a portfolio, the method comprising of estimating, by a computing system, a β-level CVaR of a portfolio where β is a real number between 0 and 1.

In a further embodiment, the portfolio comprises n number of assets, and a i , i=1, . . . , n, is the number of shares invested in an asset i.

In a further embodiment, a Gaussian copula model captures the interdependency between the assets in the portfolio. The Gaussian copula model is represented by n marginal Cumulative Distribution Functions (CDF) F i (·), and a n×n matrix Σ Z , wherein F i (·) is a marginal CDF of the potential loss of asset i, and Σ Z is a correlation matrix that captures interdependencies among asset losses.

In a further embodiment, the computing system applies a singular value decomposition or other equivalent matrix decomposition technique on the correlation matrix Σ Z to decompose it as Σ Z =U T DU, where D is a diagonal matrix with non-negative diagonal entries, and U is a unitary matrix (i.e., U T U=UU T =I n , where I n is the n×n identify matrix). The computing system generates J number of sample points V 1 , . . . , V J from a standard n-dimensional multivariate normal distribution whose mean value is zero and whose correlation matrix is I n . The computing system then creates J number of points Z 1 , . . . , Z J , by multiplying D 1/2 and U T to V 1 , . . . , V J as Z j =D 1/2 U T V j , where an index j ranges from 1 to J. The computing system further creates J number of points X 1 , . . . , X J by calculating X i j =F i −1 (Φ(Z i j )), where the asset i ranges from 1 to n, the sample point index j ranges from 1 to J, Z i j is the i-th entry of Z j , X i j is the i-th entry of X j , F i −1 (·) is the inverse function of F i (·), and Φ(·) is the univariate standard normal CDF. The computing system computes empirical losses L 1 , . . . , L J as L j =a T X j , where the index j ranges from 1 to J. The computing system then sorts L 1 , . . . , L J in an ascending order. Let L (1) , . . . , L (J) denote the sorted L 1 , . . . , L J with L (1) ≦ . . . ≦L (J) , and K denote the largest integer such that J−K≧J (1−β), i.e., K=max{j|J−j≧J(1−β), j=1, . . . , J}. The computing system estimates a β-level VaR of total portfolio loss L as L (K) . The computing system divides points X 1 , . . . , X J into two groups, a first group and a second group. The first group includes those X j 's that satisfy a T X j ≧L (K) . The second group includes remainders. The computing system divides points V 1 , . . . , V J into two groups, a third group and a fourth group. The third group includes those V j 's whose corresponding X j 's belong to the first group. The fourth group includes remaining V j 's. The computing system finds a hyper-plane that separates the third group and the fourth group. In one embodiment, to find the separating hyper-plane, the computing system applies a binary classification technique or other classification technique to the points V 1 , . . . , V J . The computing system represents the separating hyper-plane as f(x)=k T x−b=0, where k is a unit normal vector (i.e., k T k=1) of the function, |b| (i.e., the absolute value of b) is a distance from the origin (0,0) to the hyper-plane. The computing system computes a shifting amount ΔZ as ΔZ=bD 1/2 U T k. The computing system shifts the points Z 1 , . . . , Z J by ΔZ, and creates points Y 1 , . . . , Y J from the shifted points as Y i j =F i −1 (Φ(Z i j +ΔZ i )), where the asset i ranges from 1 to n, the index j ranges 1 to J. The computing system computes a set of likelihood ratios w 1 , . . . , w J as

w j = ϕ Z ⁡ ( Z j + Δ ⁢ ⁢ Z ) ϕ Z + Δ ⁢ ⁢ Z ⁡ ( Z j + Δ ⁢ ⁢ Z ) ,

where the index j ranges from 1 to J, φ Z (·) is the joint probability density function (PDF) of a n-dimensional multivariate normal distribution whose mean value is zero, and whose correlation matrix is the correlation matrix Σ Z , and φ Z+ΔZ (·) is the joint PDF of a n-dimensional multivariate normal distribution whose mean value is ΔZ=bD 1/2 U T k and whose correlation matrix is the correlation matrix Σ Z . The computing system computes “exaggerated” empirical losses {tilde over (L)} 1 , . . . , {tilde over (L)} J as {tilde over (L)} j =a T Y j , j=1, . . . , J, and sorts {tilde over (L)} 1 , . . . , {tilde over (L)} J in an ascending order, and denotes the sorted a {tilde over (L)} 1 , . . . , {tilde over (L)} J as {tilde over (L)} (1) , . . . , {tilde over (L)} (J) with {tilde over (L)} (1) ≦ . . . ≦{tilde over (L)} (J) . Let w (j) be the corresponding likelihood ratio of the j-th smallest element {tilde over (L)} (j) . The computing system finds the largest integer S between 1 and J such that the sum of w (j) from S to J is larger than J(1−β), i.e.,

The computing system estimates the β-level CVaR value of the total portfolio loss L, CVaR β (L) as

The accompanying drawings are included to provide a further understanding of the present invention, and are incorporated in and constitute a part of this specification.

FIGS. 1A-1B illustrate a flow chart that describes method steps for measuring a risk of an asset portfolio in one embodiment.

FIG. 2 illustrates an exemplary hardware configuration for implementing the flow chart depicted in FIGS. 1A-1B in one embodiment.

A portfolio may comprise an arbitrary number of assets. The potential loss of each asset is a random variable that may follow an arbitrary probability distribution. Gaussian copula model or other equivalent models captures interdependence among asset losses in the portfolio. The present invention describes a system, method and computer program product to estimate CVaR (Conditional Value-at-Risk) of the portfolio.

More specifically, let n denote the number of assets included in the portfolio, random variable Q 1 , i=1, . . . , n, denote a potential loss of an asset i, and a i , i=1, . . . , n, denote the number of shares invested in the asset i. A total portfolio loss L can be represented by L=a 1 Q 1 + . . . +a n Q n =a T Q, where a=[a 1 , . . . , a n ] T and Q=[Q 1 , . . . , Q n ] T are column vectors of a i 's and Q i 's, and a T represents the transpose of column vector a. The present invention describes a system, method and computer program product to estimate, CVaR β (L), a β-level CVaR, of the portfolio, where β is a real number between 0 and 1.

In one embodiment, interdependence among asset losses Q i 's is captured by a Guassian copula model or other equivalent models. The Gaussian copula model consists of n number of culmulative distribution functions (CDF) corresponding to n number of random variables Q i 's, and a n×n correlation matrix. Let F i (·) denote the CDF of Q i , and Σ Z denote the correlation matrix.

FIGS. 1A-1B illustrate a flow chart that describes method steps for measuring a risk of an asset portfolio in one embodiment. At step 100 , a user inputs β, a i , F i (·), and Σ Z to a computing system (e.g., a computing system 200 in FIG. 2 ), e.g., via a user interface (not shown), a keyboard (e.g., a keyboard 224 in FIG. 2 ), etc.

At step 110 , the computing system applies a singular value decomposition or other equivalent matrix decomposition technique on the correlation matrix Σ Z to decompose it as Σ z =U T DU, where D is a diagonal matrix with non-negative diagonal entries, and U is a unitary matrix (i.e., U T U=UU T =I n , where I n is an n×n identify matrix). Note that this is possible because Σ Z is a correlation matrix and thus is positive semi-definite.

At step 115 , the computing system generates J number of sample points V 1 , . . . , V T from a standard n-dimensional multivariate normal distribution whose mean value is zero and whose correlation matrix is I n . The computing system then creates J number of points Z 1 , . . . , Z J , e.g., by multiplying D c and U T to V 1 , . . . , V J as Z j =D 1/2 U T V j , where an index j ranges from 1 to J.

At step 120 , the computing system further creates J number of points X 1 , . . . , X J , e.g., by calculating X i j =F i −1 (Φ(Z i j )), where the asset i ranges from 1 to n, a sample point index j ranges from 1 to J, Z i j is the i-th entry of Z j , X i j is the i-th entry of X j , F i −1 (·) is the inverse function of F i (·), and Φ(·) is the univariate standard normal CDF.

At step 125 , the computing system computes empirical losses L 1 , . . . , L J as L j =a T X j , where the index j ranges from 1 to J. The computing system then sorts L 1 , . . . , L J , for example, in an ascending order. Let L (1) , . . . , L (J) denote the sorted L 1 , . . . , L J with L (1) ≦ . . . ≦ . . . L (J) , and K denote the largest integer such that J−K≧J(1−β), i.e.,

K =max{ j|J−j≧J (1−β), j= 1 , . . . , J}.

At step 130 , the computing system estimates a β-level VaR (Value-at-Risk) of the total portfolio loss Las L (K) . At step 135 , the computing system divides points X 1 , . . . , X J , for example, into two groups, a first group and a second group. The first group includes those X j 's that satisfy a T X j ≧L (K) . The second group includes remainders. At step 140 , the computing system divides points V 1 , . . . , V J into two groups, a third group and a fourth group. The third group includes those V j 's whose corresponding X j 's belong to the first group. The fourth group includes remaining V j 's.

At step 150 , the computing system finds a hyper-plane that separates the third group and the fourth group. In one embodiment, to find the separating hyper-plane, the computing system applies a binary classification technique or other classification technique to the points V 1 , . . . , V J . See, for example, S. B. Kotsiantis, “Supervised Machine Learning: A Review of Classification Techniques,” Informatica 31, 2007, pp. 249-268, wholly incorporated by reference as if set forth herein, for details on classification techniques. The computing system represents the separating hyper-plane as f(x)=k T x−b=0, where k is a unit normal vector (i.e., k T k=1) of the hyper-plane, |b| (i.e., the absolute value of b) is a distance from the origin (0,0) to the hyper-plane.

At step 155 , the computing system computes a shifting amount ΔZ as ΔZ=bD 1/2 U T k.

At step 160 , the computing system shifts the points Z 1 , . . . , Z J by ΔZ, and creates points Y 1 , . . . , Y J from the shifted points as Y i j =F i −1 (Φ(Z i j +ΔZ i )), where the asset i ranges from 1 to n, the index j ranges 1 to J. At step 165 , the computing system computes a set of likelihood ratios w 1 , . . . , w J as

w j = ϕ Z ⁡ ( Z j + Δ ⁢ ⁢ Z ) ϕ Z + Δ ⁢ ⁢ Z ⁡ ( Z j + Δ ⁢ ⁢ Z ) ,

where the index j ranges from 1 to J, Φ z (·) is the joint probability density function (PDF) of a n-dimensional multivariate normal distribution whose mean value is zero, and whose correlation matrix is the correlation matrix Σ Z , and Φ Z+ΔZ (·) is the joint PDF of a n-dimensional multivariate normal distribution whose mean value is ΔZ=bD c U T k and whose correlation matrix is the correlation matrix Σ Z .

At step 170 , the computing system computes “exaggerated” empirical losses {tilde over (L)} 1 , . . . , {tilde over (L)} J as {tilde over (L)} j =a T Y j , j=1, . . . , J, and sorts {tilde over (L)} 1 , . . . , {tilde over (L)} J , for example, in an ascending order, and denotes the sorted {tilde over (L)} 1 , . . . , {tilde over (L)} J as {tilde over (L)} (1) , . . . , {tilde over (L)} (J) with {tilde over (L)} (1) ≦ . . . ≦{tilde over (L)} (J) . Let w (j) be the corresponding likelihood ratio of the j-th smallest element {tilde over (L)} (j) . At step 175 , the computing system finds the largest integer S between 1 and J such that the sum of w (j) from S to J is larger than J(1−β), i.e.,

At step 180 , the computing system estimates the β-level CVaR value of the total portfolio loss L, CVaR β (L) as

CVaR β ⁡ ( L ) = ( ∑ j = S J ⁢ w ( j ) ⁢ a T ⁢ Y j ) / ( ∑ j = S J ⁢ w ( j ) ) .

The estimated β-level CVaR value of the total portfolio loss L reflects a possible loss in the portfolio. Thus, a user (e.g., a fund manager, a stock portfolio manager, etc.) may utilize this estimated β-level CVaR value of the total portfolio loss L to find out a possible or potential loss in an asset portfolio.

FIG. 2 illustrates an exemplary hardware configuration of a computing system 200 running and/or implementing the method steps in FIG. 1 . The hardware configuration preferably has at least one processor or central processing unit (CPU) 211 . The CPUs 211 are interconnected via a system bus 212 to a random access memory (RAM) 214 , read-only memory (ROM) 216 , input/output (I/O) adapter 218 (for connecting peripheral devices such as disk units 221 and tape drives 240 to the bus 212 ), user interface adapter 222 (for connecting a keyboard 224 , mouse 226 , speaker 228 , microphone 232 , and/or other user interface device to the bus 212 ), a communication adapter 234 for connecting the system 200 to a data processing network, the Internet, an Intranet, a local area network (LAN), etc., and a display adapter 236 for connecting the bus 212 to a display device 238 and/or printer 239 (e.g., a digital printer of the like).

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or computer program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” Furthermore, aspects of the present invention may take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied thereon.

Any combination of one or more computer readable medium(s) may be utilized. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with a system, apparatus, or device running an instruction.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with a system, apparatus, or device running an instruction.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may run entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

Aspects of the present invention are described below with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which run via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer program instructions may also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which run on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more operable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be run substantially concurrently, or the blocks may sometimes be run in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.