A model is needed for estimating the costs of orbital spacecraft lying 10-15 years, or more, into the future. Such a model likely could not anticipate the nature of changes in technology and system configurations over this time span. We therefore chose to focus on macroscopic trends, using a short list of high level technical and performance variables obtained for a great many past missions. A number of modeling approaches were evaluated. To gauge models' predictive accuracy, while by necessity calibrating them on past data, they were tested on missions lying as far into the future as practical. We found several approaches that could usually estimate within about 50% of actual final cost with no apparent degradation over time.