I'm looking into foster care policy and I'm coming across an empirical problem. There is a broad consensus that finding a relative of a foster child (often referred to as kin) leads to empirically better for outcomes for that child. For example:
Research shows that living with relatives is better for children and benefits them in several ways.
Improves children’s well-being
Research confirms that compared to children in nonrelative care, children in kinship homes fare better, as measured by several child well-being factors.5 Children in the care of relatives experience increased stability, with fewer placement changes, decreased likelihood of disruption and not as many school changes. Relatives are more likely than nonrelatives to support the child through difficult times and less likely to request removal of problematic children to whom they are related. The children themselves generally express more positive feelings about their placements and are less likely to run away.
Increases permanency for children
Kin caregivers also provide higher levels of permanency and children experience less reentry into foster care when living with kin. Relatives are more likely to provide a permanent home through guardianship, custody or adoption. Currently about 32% of children adopted from foster care are adopted by relatives. Another 9% exit foster care to some form of guardianship with kin. Under the Fostering Connections Act, 33 states, the District of Columbia, and six tribes have taken the option to operate federally funded Guardianship Assistance Programs designed for children and youth who have been in foster care with a relative for at least six months. This subsidized permanency option allows existing kin caregivers to become legal guardians of children with much-needed financial assistance and without the need to remain in the foster care system.
Improves behavioral and mental health outcomes
Children in kinship homes have better behavioral and mental health outcomes. One study showed children in kinship care had fewer behavioral problems three years after placement than children placed into traditional foster care. This study also found children who moved to kinship care after a significant time in foster care were more likely to have behavioral problems than children in kinship care from the outset. The long-term effects of these relationships was also studied and the formation of a close relationship with an adult, such as a kinship caregiver, was found to predict more positive mental health as an adult.
So this all makes some theoretical sense. However, there is another plausible explanation. The population of children who have relatively easy to find kin willing to take them is different from the population of children who either have no kin or have kin, but their relatives are not willing to take them. This is known as selection bias. The pool of children unable to be placed with kin might have much more trauma and worse disabilities, leading to: behavioral issues, lack of permanency (it's harder to find someone with the skills and love to take them), and worse behavioral/mental health outcomes. The specifics of complicated psychological problems from abuse and neglect are not really possible to accurately collate as data and account/control for, especially not in a large data set. Kin have knowledge of these problems and can make the call as to whether they are willing to take the placements, while non-relative foster parents have little information and just have to accept whatever foster children they are placed with. It's also plausible that parents without any relationships might be more abusive on average or take longer to be found by authorities than parents who have relatives and friends interested in the well-being of the children.
This seems like a pretty big problem for any empirical research on the subject to overcome. Obviously there is not going to be a double blind study that randomly assigns children with waiting grandparents to a non-relative home. Since this kind of random assignment isn't possible, a researcher would need to find some sort of statistical instrument or discontinuity to generate enough pseudo-randomness to make a causal link compelling. I don't have access to many studies cited, but those I have seen haven't used such an instrument and don't seem to highlight this problem as a critical limitation of their work.
Is there any research that takes this problem seriously?