This paper examines the methodological problems involved in using child homicide rates as measures of the success or failure of child protection efforts. A comprehensive review of child homicide research identifies three major methodological issues: (1) The classification of child deaths as homicides is unreliable; (2) child homicides may not generally represent the endpoint of a continuum of violence ranging from inadequate parenting to maltreatment to death; and (3) child homicides are not frequent enough occurrences to effectively measure the impact of child welfare services and policies. We conclude that while improved reporting and classification procedures may lead to some useful avenues for research, relying on child homicide statistics to evaluate child welfare services may contribute to further narrowing the scope of child welfare by stressing procedures geared primarily to preventing child homicides. Child welfare services need to develop outcome measures that tap the broader mandate of improving the circumstances and well-being of children.