There are only two factors that determine the size of a prison population: how many people are coming into prison and their length of stay. The simulation model described is an attempt to estimate these two factors. The model uses microsimulation, which takes each offender individually, applies a process or number of processes, and accumulates the results. Further, the modeler can accumulate the results into subgroups; for example, the model can forecast age, sex, or race distribution in the Federal prison system into the future, even though none of the processes relate to these demographic variables. The author models scenarios that involve sentencing changes and/or changes in the number of convictions. The modeling of sentencing changes assumes that, within each crime category, each year's admission cohort will be the same (other than a proportional increase or decrease) as the other years. Changes in criminal justice policy also impact the numbers in prison populations. A good working knowledge of the criminal justice system can inform the analyst in the manipulation of the model. Some of the changes due to policy may be anticipated in the short term; however, without a crystal ball, such a limitation in inherent in any forecasting model. The author concludes that microsimulation is a successful technique for forecasting the population of the Federal prison system. This is due in part to the availability of rich data sets and the existence of a deterministic sentencing system based in the Federal sentencing guidelines. 7 figures and 3 tables