Abstract:Determining contributions by sub-portfolios or single exposures to portfolio-wide economic capital for credit risk is an important risk measurement task. Often economic capital is measured as Value-at-Risk (VaR) of the portfolio loss distribution. For many of the credit portfolio risk models used in practice, the VaR contributions then have to be estimated from Monte Carlo samples. In the context of a partly continuous loss distribution (i.e. continuous except for a positive point mass on zero), we investigate how to combine kernel estimation methods with importance sampling to achieve more efficient (i.e. less volatile) estimation of VaR contributions.

What problem does this paper attempt to address?

This paper attempts to solve the problem of how to effectively estimate the Value - at - Risk (VaR) contribution under a partially continuous loss distribution (i.e., continuous except for a positive probability mass at zero) in the credit risk portfolio. Specifically, the paper focuses on how to combine the kernel estimation method and importance sampling to estimate VaR contributions more efficiently (i.e., with less volatility) when the loss distribution is partially continuous. ### Background and Problem Description In many financial institutions, a common method for measuring portfolio risk is through economic capital (EC). EC is usually defined as the capital buffer required to cover the losses that a lending institution may suffer, ensuring with a high probability that these losses will not exceed the capital buffer. A commonly used EC measurement method is VaR based on the portfolio loss distribution. However, determining the contribution of a single exposure or sub - portfolio to the entire portfolio's EC is a more challenging task because most practically used credit risk models cannot provide a closed - form solution for VaR contributions and thus need to be estimated by simulation methods (such as Monte Carlo simulation). ### Core Problems 1. **Partially Continuous Loss Distribution**: For a partially continuous loss distribution (i.e., continuous except for a positive probability mass at zero), directly applying existing VaR contribution estimation methods may lead to unstable estimation results. 2. **Efficient Estimation**: To reduce the estimation volatility, the paper proposes combining the kernel estimation method and importance sampling to improve the estimation efficiency. ### Solutions The main contributions of the paper are: - **Combining Kernel Estimation and Importance Sampling**: The kernel estimation method can handle non - parametric expected value estimation problems, and importance sampling can reduce the estimation variance by changing the sampling distribution. The combination of the two can provide more stable VaR contribution estimates when dealing with partially continuous loss distributions. - **Theoretical Derivation and Algorithm Design**: The paper elaborately derives the mathematical basis for applying kernel estimation and importance sampling under a partially continuous loss distribution and designs the corresponding algorithm steps. ### Formula Summary 1. **Definition of VaR**: \[ \text{EC} = \text{VaR}_{P,\alpha}(L)-E_P[L] \] where \(\text{VaR}_{P,\alpha}(L)\) is the \(\alpha\) - quantile of loss \(L\): \[ \text{VaR}_{P,\alpha}(L)=\min\{\ell:P[L\leq\ell]\geq\alpha\} \] 2. **Definition of Risk Contribution**: \[ \text{VaR}_{P,\alpha}(L_i|L)=\left.\frac{d\text{VaR}_{P,\alpha}(L + hL_i)}{dh}\right|_{h = 0} \] Under certain conditions, this can be expressed as a conditional expectation: \[ \text{VaR}_{P,\alpha}(L_i|L)=E_P[L_i|L=\text{VaR}_{P,\alpha}(L)] \] 3. **Kernel Estimator**: - Density Estimation: \[ \hat{f}_h(x)=\frac{1}{hT}\sum_{t = 1}^T\phi\left(\frac{x - x_t}{h}\right) \] - Conditional Expectation Estimation: \[ \hat{E}_h[Y|X = x]=\frac{\sum_{t = 1}^T y_t\phi\left(\frac{x - x_t}{h}\right)}{\sum_{t = 1}^T\phi\left(\frac{x - x_t}{h}\right)} \] 4. **Importance Sampling**: - Likelihood Ratio: \[

Capital allocation for credit portfolios with kernel estimators

Conditional Tail-Related Risk Estimation Using Composite Asymmetric Least Squares and Empirical Likelihood

Optimal Capital Allocation Based on Weighted-Mean-Variance Principle

Estimation of risk contributions with MCMC

Kernel density estimation based distributionally robust mean-CVaR portfolio optimization

Portfolio Optimization with Nonparametric Value at Risk: A Block Coordinate Descent Method.

Efficient Nested Estimation of CoVaR: A Decoupled Approach

Estimating Sensitivities of Portfolio Credit Risk Using Monte Carlo

Efficient Risk Measures Calculations for Generalized CreditRisk+ Models

Nonparametric Inference of Value-at-Risk for Dependent Financial Returns

Efficient Risk Estimation for the Credit Valuation Adjustment

Fair estimation of capital risk allocation

Efficient global estimation of conditional-value-at-risk through stochastic kriging and extreme value theory

Estimating Systemic Risk within Financial Networks: A Two-Step Nonparametric Method

Efficient algorithms for calculating risk measures and risk contributions in copula credit risk models

Dynamic Mean–Variance Portfolio Optimization with Value-at-Risk Constraint in Continuous Time

Markov Chain Monte Carlo Methods for Estimating Systemic Risk Allocations

Kernel Smoothing for Nested Estimation with Application to Portfolio Risk Measurement.

A fast wavelet expansion technique for Vasicek multi-factor model of portfolio credit risk

Capital allocation for credit portfolios under normal and stressed market conditions