Evaluating the accuracy of value at risk in Indonesia Stock Exchange comparison between conventional method and cornish-fisher expansion

KESUMA, Yudhistira, Mamduh M. Hanafi, Dr., MBA

2008 | Tesis | S2 Magister Manajemen

Penelitian ini bertujuan untuk mengevaluasi akurasi dari model VaR, terutama membandingkan rnetode penyimpangan normalitas dengan metode konvensional, yaitu imbal hasil yang terdistribusi normal, dalam perhitungan VaR menggunakan rnetode variance-covariance. Sampel yang digunakan dalam penelitian ini dipilih dari perusahaanperusahaan yang telah terdaftar di bursa pada September 2007, dibagi menjadi 2 periode amatan, yang masing-masing melingkupi 255 hari perdagangan. Terdapat 128 saham yang digunakan sebagai sampel. Terdiri dari 3 subgrup: dari anggota LQ45 (43), anggota KompaslOO (45) dan IHSG (40). Berdasarkan temuan, dapat disimpulkan bahwa akurasi dari metode varince-covariance dalam menghitung VaR menggunakan Cornish-Fisher expansion telah meningkat, dibandingkan dengan pendekatan konvensional yaitu menggunakan critical value dari distribusi normal. Ditemukan juga bahwa asumsi "waktu lampau akan tercerrnin di masa data" tidak terpenuhi. Meskipun asumsi tersebut adalah prinsip dasar dari VaR parametrik. Selain itu, variabel keanggotaan index pasar dan penerimaan model VaR adalah independen Kata kunci: value at risk, penyimpangan normaJitas, variance-covariance, Cornish-Fisher expansion

This research planned to evaluate the accuracy of VaR model, especially comparing the normality deviation with conventional method, which is normally distributed returns, in variance-covariance approach ofVaR computation. The samples for this research are selected from companies that have been listing at September 2007, divided into 2 period of observation; each was covering 255 trading days. There are 128 stocks use as samples of study. It consists ofthree subgroups: from LQ45 member (43), KompaslOO member (45) and Composite Stock Price Index ( 40). On the basis of the finding, it can be concluded that the accuracy of variance-covariance method of computing Value at Risk by using Cornish-Fisher expansion have increasing, compare with conventional approach that using critical value under normal distribution. It also found that the assumption "what happens in the past will happen in the future" was violated. Event ought the assumption is the underlying principle of parametric VaR. Meanwhile the variable of membership in market index is independent with the acceptance of variancecovariance approach ofVaR computation by particular stock. Keywords: value at risk, normality deviation, variance-covariance, cornish-fisher expanston

Kata Kunci : Value at risk,Penyimpangan normalitas,Variance,convariance,Cornish,fisher expansion