Methods: The method used is a two-stage multilevel model. In stage one, county-level placement rates (n. of counties=1200) are calculated using Poisson event count models with variable exposure. Measures of social disadvantage including poverty, family structure, adult educational levels, residential mobility, and nativity are used to adjust the observed placement rates. The choice of covariates is guided by theory. The stage one, county-level placement rate residual is then taken as a measure of the extent to which the observed placement rate is above or below an expected rate. In counties that place fewer children relative to levels of social disadvantage, we expect the placement rate effect will be smaller than in those counties with placement rates above the expected level. To test this hypothesis, the residual from the stage one model is entered as a context effect in a multi-level discrete time hazard model (stage two). We test the hypothesis using different permanency outcomes. Model comparison is based on effect sizes, the variance components, and overall model fit, given the results from the unadjusted and adjusted models.
Results: Results from the model indicate that it is indeed important to adjust placement rates for social conditions when attempting to understand the impact entry rates have on exit rates. The main findings have to do with reunification, in which case the effect of entry rates on reunification is particularly important in counties with a larger than expected placement rate. The impact of placement rates above the expected level is smaller in the case of adoption.
Conclusions and Implications: In general the findings provide additional but refined support for the hypothesis that system performance at the backdoor (i.e., permanency) is affected by system performance at the front door (admissions). The specific contribution of the paper relates to the notion of placement rates above or below the level expected given underlying social conditions, may be the driving force behind the system effect. For policy makers hoping to reward states for their positive improvements, the findings indicate just how important a holistic system perspective is.