The Term Structure of Government Bond Yields in an Emerging Market: Empirical Evidence from Pakistan Bond Market
The Term Structure of Government Bond Yields in an Emerging Market: Empirical Evidence from Pakistan Bond Market
复制标题
新兴市场政府债券收益率的期限结构：来自巴基斯坦债券市场的经验证据
DOI：
--
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
Muhammad Nishat
中科院分区：
文献类型：
--
作者：
Wali Ullah;Muhammad Nishat
利率期限结构的准确建模在宏观经济学和金融学中，特别是在货币政策背景下至关重要，因为其因素对于预测未来的增长和通货膨胀非常重要。本文研究了在新兴经济体的背景下，所谓的纳尔逊-西格尔模型（DNS）及其扩展版本（DNS-GARCH和DNS-EGARCH）能够最佳拟合收益率曲线并预测其未来路径的程度。该研究扩展了早期的工作（Koopman，et al. 2010），通过查看更精细的波动率建模规范，如E-Gestival，并评估了考虑模型中随时间变化的波动率在样本预测方面的预测作用。对于样本内拟合，即使在新兴市场，所有三个模型都非常适合曲线。然而，DNS-EGESTO模型比其他两个模型更好地拟合曲线。此外，基于Nelson-Siegel函数形式的收益率曲线的所有三个规格在所有三个指定的预测范围内都优于基准AR（1）预测。DNS在1个月的预测中比基于波动性的扩展模型更精确，而其他两个模型在6个月和12个月的预测中优于标准DNS。
The accurate modeling of the term structure of interest rates is of vital importance in macroeconomics and finance in general and in the context of monetary policy in particular, as its factors are important in predicting future growth and inflation. This paper investigates the extent to which the so called Nelson-Siegel model (DNS) and its extended version that accounts for time varying volatility (DNS-GARCH and DNS-EGARCH) can optimally fit the yield curve and predict its future path in the context of an emerging economy. The study expands the earlier work (Koopman, et al. 2010) by looking at more elaborate specifications for volatility modeling such as E-GARCH and also evaluates the predictive role of considering the time-varying volatility in the model in terms of out-of-sample forecasting. For the in-sample fit, all three models fit the curve remarkably well even in the emerging markets. However, the DNS-EGARCH model fits the curve slightly better than the other two models. Moreover, all three specifications of the yield curve that are based on the Nelson-Siegel functional form, outperform the benchmark AR(1) forecasts at all three specified forecast horizons. The DNS comes with more precise forecasts than the volatility based extended models for the 1-month ahead forecasts, while the other two outperform the standard DNS for 6- and 12-month horizons.