Title: Bond portfolio optimization using regime switching dynamic Nelson Siegel models
Authors: Takeshi Kobayashi - NUCB Business School (Japan) [presenting]
Naoki Makimoto - University of Tsukuba (Japan)
Abstract: Markowitz's approach of portfolio selection is applied to US government bond portfolios. We use dynamic Nelson Siegel yield curve modes to estimate expected value of portfolios, and variances, and covariances of portfolios. We also extend the dynamic Nelson Siegel model to a Markov switching latent variable model that allows for discrete changes in the stochastic process followed by the interest rates. We consider switching the following four parameters: (i) a loading parameter, (ii) time-varying volatility, (iii) an unconditional mean and (iv) a matrix of the autoregressive process coefficient. We derive a model of a myopic single-period investment strategy in a transaction cost conscious mean-variance framework with dynamic factor models under regime switches. We compare the investment performance of these models with that of the standard dynamic Nelson and Siegel model and find that regime switching dynamic factor models are useful in constructing bond portfolios which realize higher risk adjusted return during yield curve regime shifts.