Estimating Philippine Dealing System Treasury (PDST) Reference Rate Yield Curves using a State- Space Representation of the Nelson-Siegel Model
​
Len Patrick Dominic M. Garces, and Ma. Eleanor R. Reserva
​
​
Abstract. This paper uses the dynamic latent factor approach to the Nelson-Siegel model for estimating yield curves for daily Philippine Dealing System Treasury (PDST) reference rates from March 2007 to April 2014. In order to estimate the time series for the level, slope, and curvature and the parameters of the VAR(1) model for the latent factor, the MARSS (Multivariate Autoregressive State Space) package in R, developed by Holmes, et al. was used. The decay factor lamda used, obtained via grid search, was that which maximizes the R^2 in regressing PDST-F rates against the factor loadings in the Nelson-Siegel model. It is shown that the level, slope, and curvature latent factors are highly correlated with their respective empirical proxies and lag-1 values, while forecasted and actual yield data are sufficiently consistent, especially on shorter maturities. Yield curve estimates are also shown to follow the prevalent yield curve shapes posited by Nelson and Siegel.