No abstract

This paper establishes an original methodology to forecast macro-economy based on the term structure of credit spreads. It combines the traditional Svensson model with genetic algorithms to obtain the interest rate term structures of government bonds and corporate bonds, and calculates credit spreads as their differences. The principal component analysis is used to derive three factors of the term structure of credit spreads: level, slope and curvature. Based on these three factors and several macroeconomic variables including the consumer price index, exchange rate, and the growth rate of industrial and broad money, VAR models are developed and tested to forecast macroeconomic variables. Based on monthly transaction data from Shanghai Stock Exchange in China covering the period from are taken as out-of-sample data to test the models' robustness by comparing the estimates of the forecasting model with the actual values. The empirical results confirm that our VAR models can predict the changes of China's macro-economy well, which indicates that the term structure of credit spreads contains information of future changes of macroeconomic variables. We believe this result has significant implications for macro-economy policy-makers.

In this paper, we develop a new term structure model of interest rates with combinatorial optimization method based on four classical models: polynomial spline model, exponential spline model, Nelson-Siegel model and Svensson model. Genetic algorithms are employed to solve the combinatorial optimization model. Then, we make some empirical comparisons of five models using daily bond data from Shanghai Stock Exchange in China covering the years from 2004 to 2009. The results show that the combinatorial optimization model outperforms the other models in most of the statistical indicators. Besides, the combinatorial model has good adaptability and robustness which are applicable in Chinese bond market.