BOND YIELD CURVE CONSTRUCTION USING HYBRID NM-SPIRAL OPTIMIZATION FOR NELSON-SIEGEL MODEL AND NELSON-SIEGEL-SVENSSON MODEL AND USING SMOOTHING SPLINE

• Main Author: Hanzel Hartono, Vincent
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74569
• Institution: Institut Teknologi Bandung

The yield curve has an important role in the economy as a reference for determining the fair value of a bond, a benchmark for bond yields, and for summing up economic expectations. Several models that can be used to construct yield curve are the Nelson-Siegel (1987), the Nelson-Siegel-Svensson (1994), and the smoothing spline. These three models are applied to USA and Canadian government zerocoupon bond data. Parameter estimation for the Nelson-Siegel and Nelson-Siegel- Svensson models uses Nelder-Mead optimization, spiral optimization, and hybrid NM-spiral optimization methods. The Nelder-Mead optimization method turns out to give inconsistent results, depending on the initial iteration value. Meanwhile, the spiral optimization method gives better results even though it requires a longer computation time. If the results of the spiral optimization are used as the initial values for Nelder-Mead iterations in the hybrid NM-Spiral optimization, very good results are obtained with a relatively short computation time. Then, it turns out that the smoothing spline model has the ability to approach data better, especially on data with a large variation, even though the resulting curve tends to be wavy. The Nelson-Siegel and Nelson-Siegel-Svensson models provide a smoother curve than the smoothing spline model. Meanwhile, the Nelson-Siegel-Svensson model provides a better fit than the Nelson-Siegel model.