Perbaikan Model NAGARCH Menggunakan Data Frekuensi Tinggi

Abstract

Volatilitas adalah salah satu indikator risiko terpenting bagi pelaku dan pengamat pasar kuangan. Generalized Autoregressive Conditional Heteroskedasticity (GARCH) adalah salah satu model populer yang digunakan untuk mengukur volatilitas, dan telah dikembangkan menjadi berbagai varian seperti GARCH-X, GARCH-CJ, dan Realized GARCH (RGARCH). Studi ini berfokus pada pengembangan model Realized Non-Linear Asymmetric GARCH (RNAGARCH) dan Realized Non-Linear Asymmetric GARCH-CJ (RNAGARCH-CJ), serta mengevaluasi kinerja kedua model dalam memprediksi voltilitas menggunakan data Tokyo Stock Price Indeks (TOPIX) dari 2004 hingga 2011. Metode Adaptive Random Walk Metropolis (ARWM) dalam algoritma Markov Chain Monte Carlo (MCMC) digunakan untuk mengestimasi parameter model. Hasil penelitian menujukkan bahwa model RNAGARCH-CJ memiliki kinerja terbaik dibandingkan model RNAGARCH dan NAGARCH, berdasarkan nilai Log-Likelihood dan empat kriteria pemilihan model lainnya.

Volatility is one of the most important risk indicators for market participants and observers. The Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model is one of the popular models used to measure volatility and has evolved into various variants such as GARCH-X, GARCH-CJ, and Realized GARCH (RGARCH). This study focuses on the development of the Realized Non-Linear Asymmetric GARCH (RNAGARCH) and Realized Non-Linear Asymmetric GARCH-CJ (RNAGARCH-CJ) models, as well as evaluating the performance of these models in predicting volatility using Tokyo Stock Price Index (TOPIX) data from 2004 to 2011. The Adaptive Random Walk Metropolis (ARWM) method in the Markov Chain Monte Carlo (MCMC) algorithm is used to estimate the model parameters. The results show that the RNAGARCH-CJ model performs the best compared to the RNAGARCH and NAGARCH models, based on Log-Likelihood values and four other model selection criteria.