A term structure dynamic model with correlated residuals: A comparative analysis

Recently, modelling the term structure of interest rates has gained particular relevance in the economic and financial literature and has become one of the main research topics. Nowadays, the growth of financial markets and emerging derivative instruments require the development of new techniques for estimating and forecasting interest rates that could be adapted to reality. Term structure modelling should consider two significant dimensions, time and maturity, although, in the literature, these were treated separately or analyzed through multivariate techniques. An alternative geostatistical approach was based on the use of a correlation function which was assumed to be dependent only on the temporal lag. Differently from the existing contributions, the aim is to propose a geostatistical model for the term structure of spot rates, where the joint evolution concerning time and maturity is considered. In particular, three hypotheses on the trend component of the random field are assumed: i) constant, ii) dependent only on time, and iii) dependent on time and maturity. Finally, a comparison among the predictive performance of models in that different hypotheses is proposed.