원화이자율스왑(IRS) 거래의 개시증거금에 관한 연구

Abstract

In securities settlement, the central counterparty clearing (CCP) acts as a buyer for all sellers and a seller for all buyers among traders. To contrl for settlement risk, the CCP implements risk management procedures and invests necessary financial resources in a predetermined order, and the margin is composed of variation margin and initial margin. Initial margin refers to the funds required to cover the price fluctuation risk of the position of a particular participant during the period required for the CCP to close out the trading position of that participant due to the failure of that participant. The international guidelines for designing and operating an FMI recommend that the margin be set at a level sufficient to cover the potential future risk to which the CCP is exposed between the final margin charge and the time of position closure due to the failure of the participant. Hence, global CCPs adopt different margin models depending on the financial products they clear and the related market structure. In particular, they use FHS VaR to supplement historical VaR. This paper examines the FHS VaR model adopted by domestic interest rate swap CCPs and constructs an initial margin estimation program to empirically analyze it, thereby examining key variables. The results of the empirical analysis show that the margin size calculated by the current FHS VaR with EWMA method is different from the margin size calculated by the historical method. However, this difference does not determine whether the margin is excessive or insufficient. This paper focuses on constructing an analysis framework (including analysis code) that reflects the current margin system. During the filtering process, parameter values ​​can cause differences in the margin result values. This is especially true when the portfolio is sensitive to the far tail. It may be necessary to examine the properties of residuals and the scaled distribution. It seems that various factors such as the value of the filtering factor, the decay factor, the data window size, and the reflection of the minimum volatility (floor) to alleviate procyclicality have an effect. It is necessary to comprehensively examine and test these factors. The margin model test should not be completed in one go, but should be repeated regularly to maintain the model's appropriateness. Even if the model coefficient values ​​are accepted, they may not be sufficiently sensitive in actual operation. In such cases, it is also necessary to have additional decision criteria rather than setting a specific coefficient value to another coefficient value. Procyclicality is pointed out as this criterion. Although the stability of the margin is very important, a model with less procyclicality may be adopted first. Overseas CCPs calculate the initial margin according to IRS transactions using models such as ES and Principal Component Analysis. It is important to conduct sensitivity analysis or backtests regularly in various ways to maintain the model design and coefficient values ​​appropriately.