A Simulation Analytics Approach to Dynamic Risk Measurement

Project: Research

View graph of relations


Financial risk measures are important indicators of risk of financial portfolios and even stabilityof financial institutions. Therefore, it is critical to be able to estimate them accurately andpromptly. Estimating risk measures is typically a challenging task due to the complicatedstructures of securities contained in the portfolios. Previous work in the literature focusesalmost exclusively on estimating risk measures in a static way, and considers the estimationproblem only once. Therefore, almost all methods are designed to be implemented over nightor weekend. They provide accurate estimates of risk measures, but cannot be applied in realtime. Taking advantage of the recent development of statistical learning methods and big dataanalytics, we propose to take an alternative view and formulate the dynamic risk estimationproblem as a statistical learning problem using simulated data. This approach is generallyknown as simulation analytics and has been proposed only very recently. Our goal is to providefast and accurate estimates of risk measures so that they can be used to monitor portfolio risksin real time.In this project we propose to build regression models (e.g., linear regression models or logisticregression models) using data generated in past simulation experiments and to use the modelsto predict portfolio risk measures and conduct risk monitoring or control. We also explorestatistical learning tools, such as regularization, to improve the prediction accuracy by betterbalancing the tradeoff between bias and variance and to select important risk factors fordynamic risk monitoring.The statistical learning based methods can in general be viewed as black-box methods. Theyconduct data mining on the simulated data without exploring the information of the simulationmodel itself. However, in the field of risk management, analysts often possess a lot ofinformation on the simulation model itself, and we propose to use this information to furtherimprove the prediction quality of the regression model. To do that, we propose to considerseveral issues, including how to manipulate the simulated sample paths so that they appear inimportant regions, how to use neighborhood time information, and how to incorporate stylizedmodels into basis functions of regression models, and evaluate these enhancement methodsboth theoretically and empirically.


Project number9042570
Grant typeGRF
Effective start/end date1/01/1823/12/20