Global financial planning requires representative scenarios drawn from
stochastic forecasting systems such as Russell's vector autoregressiv
e model, Wilkie's cascade approach, Towers Perrin's global CAP:Link, a
nd the catastrophic event simulations for earthquakes and hurricanes.
We discuss the role of a dynamic forecasting system in the context of
asset and liability management. We also describe a quasi-random sampli
ng procedure for maximizing the precision of recommendations derived f
rom a dynamic decision strategy. Empirical results, for a large offsho
re reinsurance company over a 5 year planning horizon, show the benefi
ts of careful scenario selection. (C) 1998 Elsevier Science B.V.