Long memory forecasting of yield spreads using a fractionally integrated ARMA model and its application in Islamic capital market. (2017)

• Record Type: Journal Article
• Authors: Bousalam, Issam; Hamzaoui, Moustapha
• Is Part Of: International journal of bonds and derivatives. Volume 3:Number 1(2017)
• Page Start: 71
• Page End: 92
• Publication Date: 2017
• Subjects: long memory -- ARFIMA model -- Sukuk -- yield spreads analysis -- YSA -- Islamic indexes; Bonds -- Periodicals; Derivative securities -- Periodicals; Capital market -- Periodicals; 332.63205
• Languages: English
• ISSNs: 2050-2281

Abstract

In this paper, we used a modified rescaled range analysis (MRS) to investigate the presence of long memory in three series of absolute yield spreads (AYS) of the Dow Jones Sukuk Indexes from March 1, 2011 to March 1, 2016. The estimated Hurst exponents for the three series are significant and smaller than one providing strong evidence that long range dependence exists in Sukuk's AYS and these can become stationary with fractional differencing. Based on these results, we fitted three ARFIMA models to Sukuk's AYS and found that they have better explanatory power compared to the first-order ARIMA models. Furthermore, our 260 steps-ahead dynamic forecasting results show that the ARFIMA models are better for predicting future yield spreads. Such findings suggest to account for long memory in investing decisions and projecting future yields and spreads. Our results should be useful to Sukuk market participants whose success depends on the ability to forecast Sukuk's yield spreads movements, and anticipate the prospective default risk.