This article develops and evaluates a method for deriving postcensal estimates of household income distributions for counties. A modified lognormal probability curve is used as a model of income distribution. The function is closely related to the classical lognormal model, but it contains a nonlinear component in its derivation. Simulated postcensal estimates of household income distributions are compared with 1980 census data for the counties in California. The results indicate that the modified lognormal curve approximates observed income distributions well and produces reliable postcensal estimates for areas with a wide variety of median income levels and numbers of households.