Portfolio Immunization using Independent Component Analysis
Mariano González Sánchez, Juan M. Nave Pineda · 2010
Fixed-income portfolio managers usually use Principal Component Analysis (PCA) to nd the lowest number of factors that explain the behavior of a set of term structure key rates at a specic condence level. This technique reduces the size of the multidirectional immunization problem and improves the solution feasibility. Barber and Copper (1996) and Golup and Tilman (1997) are clear examples. However, when we use PCA factors in risk analysis we do not take into account the moments with order higher than two of the probability function of the yields, causing misspecication problems. In this paper we show how to apply the Independent Component Analysis (ICA), an alternative statistical technique to PCA, in xed-income portfolio management in order to achieve this same objective but avoiding misspecication problems and improving the immunization results.