Mengukur resiko lima indeks pasar saham asean dengan metode normal dan cornish fisher value at risk

Oetomo, Bangkit (2016) Mengukur resiko lima indeks pasar saham asean dengan metode normal dan cornish fisher value at risk. Masters thesis, Institut Pertanian Bogor.

[img]
Preview
Text
R53-01-Oetomo-Cover.pdf - Published Version

Download (1MB) | Preview
[img]
Preview
Text
R53-03-Oetomo-Summary.pdf - Published Version

Download (1MB) | Preview
[img]
Preview
Text
R53-02-Oetomo-Ringkasan.pdf - Published Version

Download (1MB) | Preview
[img]
Preview
Text
R53-04-Oetomo-Daftarisi.pdf

Download (1MB) | Preview
[img]
Preview
Text
R53-05-Oetomo-Pendahuluan.pdf - Published Version

Download (2MB) | Preview
[img] Text
Tesis.pdf
Restricted to Registered users only

Download (2MB)

Abstract

Value at Risk is a method to measure, quantify, and forecast market risk in particular time interval. This methodology tries to quantify risk on certain degree of level of confidence applied by the VaR user. For example, by using 5% level of confidence then VaR users are predict that 95% probability return will be higher than VaR value. Therefore, the VaR value represents the worst 5% probability that might occur. This method is one of the key components in risk governance under Basel Committee on Banking Supervision risk framework. There are still numbers of VaR studies, especially after 2008 subprime mortgage crisis, try to apply various econometric methods to find the best suitable model in certain market accomplied by actual characteristics of those market. Unfortunately there is no “one fit for all” method in measuring risk under VaR model due to the difference of data characteristic. The market movement is generally determined by external condition and internal conditions. However, the VaR model should not only considering about its accuracy but also its practicality. The objective of this research is to develop accurate and practical VaR model to quantify and compare the level of risk posses by ASEAN-5 indices that consist of Indonesia (IDX), Malaysia (MYX), Singapore (SGX), Phillipines (PSE), and Thailand (SET) using two VaR approaches that distinguished by its distribution models. The VaR model that assumes normality in its return distribution, called Normal VaR, has a limittation in capture the extreme deviation due to its normal density function value that underestimate extreme deviation. To encounter normal distribution sortages, this study employs Cornish Fisher expansion model which will modify normal density function value according to skewness and kurtosis of actual distribution. The VaR model under this modification function (CFVaR) is expected to overcome the limitation of Normal VaR. In general there two types of markets in ASEAN-5 markets; developed market which consist of SGX and emerging markets which consist of IDX, MYX, PSE, and SET. There are significant numbers of studies found that stock market return posses asymmetric distribution with fat tail. The existance of fat tail distribution indicates there are extreme return deviations from the mean. Therefore, develop an accurate model is essential to capture this extreme deviation characteristic. There are two main concentrations in VaR model development; (i) designing conditional volatility model, and (ii) designing accurate distribution model to accomodate the high variance data characteristic. To capture dynamic and clustered volatility, this study employs Autoregressive Moving Average (ARMA), Autoregressive Conditional Heteroscedasticity (ARCH), and Generalized ARCH (GARCH) to modeled the conditional volatility. These models are able to capture non constant variance movement which require dynamic capability. These models assumed that variance follows non constant patterns and posses correlation between return series. The performance of these two approaches need to be tested for each market to find the best application in every markets. The time interval taken in this study is start from 5th January 2009 to 19th January 2016. This research finds that the distribution of ARMA residuals in all markets show non normal skewness and kurtosis. The ARMA-GARCH (1,1) model perform best in emerging ASEAN due to strong autocorrelation meanwhile SGX which catagorized as developed market shows efficient market characteristic where historical return no longer able to predict future movement. The GARCH (0,1) model on SGX does not have dynamic capability due to the absence of ARCH component. The CFVaR model produces higher risk calculation in markets with significant skewness such as IDX, PSE, and SET. Meanwhile SGX and MYX residuals show close character with normal distribution but with substantial kurtosis which have made the CFVaR model calculate lower risk than Normal VaR This research reveals that distribution quantile modification model under Cornish Fisher expansion is not neccessarily perform better compare with Normal VaR. The CFVaR model overestimates risk in PSE, therefore this model is not suitable to measure and forecast market risk in this market. On the other hand, CFVaR model have superiority compare with Normal VaR in IDX and SET where Normal VaR perform better for PSE, MYX, and SGX.

Item Type: Thesis (Masters)
Additional Information: 2(53)Oet m
Uncontrolled Keywords: ARMA-GARCH, distribusi normal, modifikasi distribusi, value at risk. ARMA-GARCH, normal distribution, distribution modification, value at risk.
Subjects: Manajemen Keuangan
Depositing User: SB-IPB Library
Date Deposited: 25 Jan 2017 03:58
Last Modified: 24 Oct 2019 02:12
URI: http://repository.sb.ipb.ac.id/id/eprint/2788

Actions (login required)

View Item View Item