Existing postcensal estimates of poverty and income at the county level are considered inadequate for various reasons: The Census is rapidly dated and the March CPS is not sufficiently reliable, especially for those counties which are not sampled by CPS. The goal of The Small Area Income and Poverty Estimates (SAIPE) project is to form these estimates. We modeled the number of poor in various age categories and median household income as a function of various variables taken from administrative records. We recognize two sources of "error" -- sampling error and model error -- and apply a shrinkage estimator to obtain estimates of number of poor or median income by county. Finally, a ratio adjustment is used to make estimates consistent with the SAIPE state estimates. We describe the methods used to obtain these estimates and their standard errors and present some empirical evaluations of the models.