Understanding the causal impacts of taking youth on the margins of risk into foster care is an element of the evidence-base on which policy development for this crucial function of government relies. Yet, there is little research looking at these causal impacts; neither is there much empirical work looking at long-term outcomes. This paper focuses on estimating the impact of placing 16 to 18 year old male youth into care on their rates of high school graduation, and post-majority income assistance receipt and incarceration. Two distinct sources of exogenous variation are used to generate instrumental variables, the estimates from which are interpreted in a heterogeneous treatment effects framework as local average treatment effects (LATEs). And, indeed, each source of exogenous variation is observed to estimate different parameters. While both instruments are in accord in that placement in foster care reduces (or delays) high school graduation, the impact of taking youth into care on income assistance use has dramatically different magnitudes across the two margins explored, and, perhaps surprisingly, one source of exogenous variation causes an increase, and the other a decrease, in the likelihood of the youth being incarcerated by age 20. Our results suggest that it is not enough to ask whether more or fewer children should be taken into care; rather, which children are, and how they are, taken into care matter for long-term outcomes.