Apply Ridge Regression To The Nelson-siegel Term Structure Of Interest Rates

The term structure of interest rates is always an important field in financial research. In practice, The Nelson-Siegel model is the most widely used term structure model. Scholars commonly use Grid Search and OLS to estimate the parameters. however, both estimation methods are very unstable, some of the estimate parameters can not be reasonably explained, and the variance is large as well. Ridge Regression analysis is a dedicated method in the collinear data analysis regression, so in this article we choose ridge regression to solve these problems. Treasury trading in Shanghai Stock Exchange is active and sensitive to the market information, which is suitable for fitting the term structure of interest rates. We choose treasury trading data of Shanghai Stock Exchange from2009to2012for a total of968days as our sample. In the paper, we compare the parameter sequences from grid search method, OLS and ridge regression. In-sample fitting result shows that the parameters estimated from grid search have high volatility, and some of them even go against the common economic theory, the result also shows that the variation estimated from OLS is large. And out-of-sample forecasting result shows that ridge regression have the smallest Mean absolute percentage error. The innovation of this paper is firstly apply ridge regression to the Nelson-Siegel model, and provide a more effective method to fit the term structure of interest rates of Shanghai Stock Exchange.