CFDS Discussion Paper 2019/4
The current state-of-the-art estimation of yield curves relies on the dynamic state space version of the Nelson and Siegel (1987) model proposed in the seminal paper by Diebold et al. (2006). However, things become difficult when applying their approach to emerging economies with less frequently bond issuance and more sparse maturity available. Therefore, the traditional state space representation, which requires dense and fixed grids of maturities, may not be possible. One remedy is to use the traditional Nelson and Siegel (1987) OLS estimation instead, though it sacrifices efficiency by ignoring the time dimension. We propose a simple augmentation of the Diebold et al. (2006) framework, which is more efficient than OLS estimation as it allows exploiting information from all available bonds and the time dependency of yields. We demonstrate the efficiency gains generated by our method in five case studies for major emerging economies including four of the BRICS.