Logistic regression models for the probability of child abuse reports to child protective services were estimated using maximum likelihood methods for 511 black and 2,499 white cases of child abuse. The models were estimated with and without correction for selection bias, using a two-step procedure proposed by Heckman. Results revealed substantial differences in the characteristics of black and white victims by source of report and by type of maltreatment. Results also revealed sizeable differences within each racial group between sampled agencies and nonsampled agencies. Sample selection bias affected the estimation of both black and white reporting rates, but in different ways. White victims of child maltreatment were more likely to be reported if they were among the lower class, known by a law enforcement or medical agency, or female. White victims of physical or sexual abuse were more likely to be reported than were victims of emotional maltreatment. In contrast, black maltreated children were more likely to be known by child protective services if they lived in rural areas, were older, or experienced physical abuse. Black victims were less likely to be reported if they experienced emotional abuse. Findings indicated that selection bias was caused by the exclusion of family, friends, and neighbors in the NIS sample design. Such exclusion had the effect of altering the interpretation of the determinants of child abuse reporting among black families but not among white families. Thus, conclusions about racial differences in child maltreatment must be reached cautiously, given the NIS study design. Tables, appended list of variables, and 19 references (Author abstract modified)