At the beginning of the month I introduced Donald Baumann on this topic of continuing interest, and have been in contact with him since then. Here is Donn's response to my post, setting out his view of the current state of the data.
"Over-representation comes from several sources and we have evidence for some of them. According to the latest National Incidence Study (NIS 4), African American families are more likely to maltreat their children under certain conditions. However, if you look closely at the Texas CPS data on substantiated maltreatment, controlling for poverty, single parenthood, young parents, number of children and other factors, a picture of racial bias emerges.
What we find is that even at lower levels of risk of future maltreatment (which includes prior reporting), workers are more likely to substantiate African Americans than Anglos. What this means is that both bias and maltreatment may play a part in all of this but no one knows what that part is. In our view, poverty is the key because it is viewed differently for African Americans than it is for Anglos and, as a result, the threshold for substantiation is lower for African Americans than Anglos. At present, however, no one knows what features of poverty are related to substantiated maltreatment and what are related to bias.
One major source of over-representation that we know about is the higher reporting rates for African Americans, relative to Anglos. This higher rate appears to move through what we refer to as the CPS Decision-Making Continuum from intake to service provision and placement of children into care. Thus, some of the higher rates of disproportionality results from reporting and that contribution to disproportionality continues to move through the system.
So why is there so much disproportionality at reporting? Some have argued that people who are low income are often under greater scrutiny through their contacts with social service programs, increased law enforcement in low-income neighborhoods, etc., so this likely increases reporting about these families - thus, we can't say with certainty whether "repeated contacts" are truly indicative of greater rates of maltreatment among African Americans or of unidentified maltreatment among other racial groups.
Others have argued from data such as infant mortality rates that the reporting rates appear similar to the fatality rates and, because the latter is not as subject to bias as reporting rates, the reporting rates are real.
Finally, we don't know much about the potential for reporting bias. It may be that, much like the substantiation decision, the decision to report contains some disparate features that are related to actual maltreatment and some that are related to bias.
Another source of over-representation is decision-making bias that contributes to the higher rates of placements into care for African American children relative to Anglos and the lower rates of reunification for African American children relative to Anglo children. Regarding the latter two sources (flow in and out of the system) we controlled for the same factors as above for substantiation and; in the case of reunification, drug abuse and inadequate housing are additional factors that are controlled for.
Our findings are similar to those with regard to substantiation. What this means is that both underlying risk of maltreatment and racial bias play a part in how cases flow in and out of the CPS system of care and disproportionality is the result."