We assessed land use and demographic data as predictors discriminating between counties experiencing large or small first epizootics of rabies among raccoons (Procyon lotor). Monthly county reports of raccoons testing positive for rabies were obtained from rabies surveillance databases from Maryland, Pennsylvania, and Virginia (USA). Environmental and demographic data for the three states were obtained from public sources. On the basis of total reports of raccoon rabies during the first defined epizootic period, the 203 counties were dichotomized at the 75th percentile as having a large epizootic (≥24 rabid raccoons in the first epizootic) (51 counties) or a small epizootic or no epizootic (152 counties). A high percentage of agricultural land use [OR=9.1, 95% CI (3.6–23.1)], high water coverage in combination with low human population density [OR=8.8, 95% CI (2.9–27.0)], and low water coverage with high human population density [OR=11.7, 95% CI (4.0–34.1)] were positively associated with large rabies epizootics. Counties with more than 15% of mixed forest were less likely to experience large epizootics than were counties with ≤15% of mixed forest [OR=0.3, 95% CI (0.1, 0.9)]. A combination of land use and human population density measures provided the best model for determining epizootic size and may be important predictors of epizootic behavior and risk of exposure to this reservoir species.