Use of reported harvests as an index to actual harvest assumes that the proportion of harvest reported is equal for all types of animals and hunters and does not vary spatially or temporally. We modeled reporting rates of white-tailed deer (Odocoileus virginianus) harvest to determine whether they varied by year, deer management unit (DMU), type of deer (antlered or antlerless), or sex. During rifle seasons in Pennsylvania, USA, from 1990 to 2001, reporting rates varied by year, DMU, and type of deer (antlered or antlerless). Harvest estimates of antlered and antlerless deer were precise for both statewide (CV <2.5%) and DMUs (CV < 24%, median CV <5.2%). For DMUs, reported harvests were poor predictors of estimated antlered harvests (median R² = 0.287) but generally acceptable for antlerless harvests (median R² = 0.909). During the 2000 and 2001 hunting seasons, statewide average predicted reporting rates ranged from 36 to 60% and varied by year, hunting season, and type of deer. Average predicted reporting rates also varied by DMU (range = 31.5–57.5%). Applying rifle-season reporting rates to other seasons resulted in overestimating harvest by 26–28%. Variability of reporting rates precluded use of reported harvests as reliable indices of actual harvest. We recommend regular estimation of reporting rates and caution against assuming a constant reporting rate, even in consistent harvest registration systems.