Abstract: Penalized quantile regression with the SCAD was introduced into the spline function to build the term structure of interest rates of treasure bills. This method can automatically select the optimal quantile and complete the models knot selection and parameter estimation at the same time. The out-of-sample forecasting results show that, compared with the traditional methods, the new method can select the appropriate model effectively, increase the robustness of parameter estimation, improve forecast accuracy and enhance the pricing precision of the term structure of interest rates.

Key words: term structure, spline function, knot selection, Q-SCAD