This paper presents an improved bounding-based procedure for estimating expected production costs for a multiarea power system. The method builds upon the approach in [I]. It uses a state space characterization of equipment outages and loads, combined with linear optimization methods, standard load duration curve-based production costing techniques, and clustering analysis. The method defines upper and lower bounds to the true expected cost, and tightens those bounds by partitioning the outage and load spaces. Methods for constructing outage and load partitions are compared and applied to large multiarea systems. Application issues, such as marginal cost estimation and the effect of transmission outages, are also addressed.