Abstract: A multi-stage stochastic programming (MSSP) model for investment in government bond portfolios is proposed in a dynamic setting, an active dynamic strategy under uncertainty is derived. The uncertainty is modeled by discrete scenario trees including the levels, slopes and curves of interest rates, and the growth rate of broad money supply, which reflect the dynamic evolution of the term structure of interest rates. The model minimizes the CVaR of the government bond portfolio with comprehensive consideration of safety, liquidity and profitability, and achieves an effective balance between risk and the expected return. Empirical results show that the MSSP model outperforms traditional duration vector immunization approaches significantly, in terms of stronger competence in profit generation and risk control, which provides a flexible and effective decision support for active management of government bond investment by financial institutions.

Key words: government bonds investment, stochastic programming, scenario generation, term structure of interest rates, dynamic Nelson-Siegel model