Following Bell (1997), we first present a simplified ideal model which can be used to simulate underlying data for the small area (county-level) estimation currently used in the SAIPE project. We then describe two different small-domain estimation methodologies which can be used on such data: (i) a mixed-effect linear-model fit to the logarithms of sampled counts, with zero-counts discarded, and (ii) a mixed-effect unit-level logistic regression model fit to the sampled counts. The methods, both of which are based on slightly misspecified models, are compared via simulation. Initially, sampling weights are ignored, but then it is shown how they can be included in both the aggregated-linear and the unit-level models.