ABSTRACT
Although the central aim of much foster care research has been to determine how variations in placement experiences influence different children, much of it has been limited by a failure to appreciate the complex interaction between child and placement characteristics (Courtney 1993; Farmer 1996). Many of these problems are attributable not so much to any lack of understanding of the issues involved but to unsatisfactory methodological designs. As indicated in Chapter 4, the most significant limitation of these designs, and of cross-sectional designs in general, is that they tend to
over-sample children with longer placement histories. In studies of placement disruption, for example, conclusions have been drawn solely from snapshot samples of children already in care, despite the fact that children who are sampled from populations already in care are more likely to have been in care longer, and to have experienced more placement changes than other foster children not selected in this manner (Courtney 1994; Fanshel and Shinn 1978; Fernandez 1999; Goerge 1990). Many correlational studies also fail to consider the lack of independence between predictor variables, or they attempt to infer causality on the basis of covariance, and they rarely look for moderating or mediating variables. In studies that show a positive relationship between the number of placement changes and behavioural adjustment, for example, it may be tempting to conclude that placement instability results in problematic behaviour. But, as pointed out in Chapter 4, if older children display more conduct problems and have also been in care longer, it is hardly surprising that we should find associations between challenging behaviours and placement disruption (Courtney 1994).
