ABSTRACT
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Appendix 9.A: Additional Data for the Case Study . . . . . . . . . . . . . . . . . . . . 219
Appendix 9.B: Dynamic Programming Recursion for the Sample Approximation . .220
I N RECENT YEARS, A GROWING NUMBER of real-world applications of asset liabilitymanagement (ALM) with discrete-time models have emerged. Insurance companies and pension funds pioneered these applications, which include the Russell/Yasuada investment system (Carino and Ziemba 1998), the Towers/Perrin System (Mulvey 1995), the Siemens Austria Pension Fund (Ziemba 2003; Geyer et al. 2004), and Pioneer
Investment guaranteed funds (Dempster et al. 2006). In each of the applications, the
investment decisions are linked to liability choices, and the funds are maximized over time
using multi-stage stochastic programming methods. Other examples of the use of
stochastic programming to solve dynamic ALM problems are given by Dempster and
Consigli (1998) and Dondi (2005).