Estimating Yield Curves In Turkey: Factor Analysis Approach

In this paper, we perform factor analysis on yield curves estimated by McCulloch and Nelson-Siegel methods. We estimate factors using nominal volume-weighted average monthly zero-coupon yields data from the Turkish Secondary Government Securities market. Our main aim is to characterize each monthly yield curve by three factors and forecast yield curves using time series properties of each factor. According to loadings of each factor, we label the factors as level, slope and curvature, respectively. We also examine their explanatory power in different sub-samples and explore their time series properties using an unrestricted VAR. We next forecast yield curves using AR-GARCH and random walk processes for the factors and compare their relative performance. We find encouraging results regarding explanatory power of three factor model and superior forecasting power of the AR-GARCH specification.