ABSTRACT
This chapter utilizes the child outcome of high school dropout as the example for applying the statistical methodology. Literature on non-experimentally based statistical methods using longitudinal data suggests that, when a rigorous statistical approach is used with required longitudinal data, a nonexperimental design can be a very effective method for evaluating the impact of ongoing programs. Matching techniques going beyond the usual linking with Social Security numbers have been employed to insure high-quality linkage over time and program. The inclusion of Chicago Public Schools, juvenile justice, and unemployment insurance data takes the Integrated Database on Children and Family Services in Illinois beyond the social service program domain. A key concern of a nonexperimental study in determining the effect of welfare reform is selection bias. The major challenge is to develop an analysis model that can be used to isolate the effect of welfare reform apart from that of other changes in the environment.
