Bayesian Vector Autoregressive (BVAR) dalam Meramal Mata Uang Cina, India dan Indonesia terhadap Mata Uang Amerika Serikat

Abstract

Currency exchange rates can be used to measure the economic level of a country. Forecasting exchange rates is necessary so that investors can determine the future economic status of an investment destination country. In this research, the Bayesian Vector Autoregressive (BVAR) method is used to model, forecast, and compare the forecast results of the Rupee, Rupiah, and Yuan exchange rates against the US Dollar using two different priors: the Litterman-Minnesota Prior and the Normal-Flat Prior. The objective of this study is to obtain a model and compare the forecasting results of the BVAR method from each prior. Based on previous research, Bayesian Vector Autoregressive (BVAR) has been proven capable of providing superior forecasting results. The forecasting results from each prior indicate that both the Litterman-Minnesota prior and the Normal-Flat prior produce accurate forecasts based on low Mean Absolute Percentage Error (MAPE) values. Furthermore, after comparison, in the case of forecasting the Rupee, Rupiah, and Yuan exchange rates against the US Dollar, the Litterman-Minnesota prior yielded a smaller MAPE value, meaning its results are more accurate.

Keywords: Bayesian Vector Autoregressive (BVAR), Exchange rate, Litterman-Minnesota Prior, Mean Absolute Percentage Error (MAPE), Normal-Flat Prior.

ABSTRAK

Nilai tukar mata uang dapat digunakan untuk mengukur tingkat perekonomian suatu negara. Peramalan nilai tukar diperlukan agar investor dapat mengetahui tingkat perekonomian negara tujuan investasi di masa datang. Pada penelitian ini digunakan metode Bayesian Vector Autoregressive (BVAR) dalam memodelkan, melakukan peramalan,dan membandingkan hasil ramalan antara nilai tukar Rupee, Rupiah, dan Yuan terhadap Dolar Amerika Serikat dengan dua prior yang berbeda yaitu Litterman-Minnesota Prior dan Normal-Flat Prior. Tujuan dari penelitian ini adalah mendapatkan model, serta membandingkan hasil peramalan dengan metode BVAR dari masing-masing prior. Bayesian Vector Autoregressive (BVAR) berdasarkan penelitian sebelumnya,terbukti mampu memberikan hasil peramalan yang lebih unggul. Hasil peramalan dari masing-masing prior menyatakan bahwa baik Litterman-Minnesota prior dan Normal-Flat prior menghasilkan peramalan yang akurat berdasarkan nilai Mean Absolute Percentage Error (MAPE) yang kecil. Kemudian, setelah dibandingkan, pada kasus peramalan nilai tukar Rupee, Rupiah, dan Yuan terhadap Dolar Amerika Serikat, Litterman-Minnesota prior memiliki nilai MAPE lebih kecil yang artinya hasilnya lebih akurat.