Abstract

Statement of the problem. The objective of this dissertation is to implement and compare new approaches to measuring the effectiveness of children's mental health treatments in a community setting, especially with respect to the particular problems encountered in non-randomized studies. Static and dynamic models are proposed to address two problems associated with estimating the treatment effect. The first issue is the potential endogeneity of the treatment variables; the second is the possibility of significant dynamics in the structural model to permit certain forms of intertemporal correlation.

Study data sources. The data set for this study is the National Longitudinal Survey of Youth 1979 for Children and Young Adults (NLSY79-CHILDYA), which includes 8 biennial waves of data from children who were between the ages of 0 and 14 during the first interview in 1986. Data collected on the children's mothers collected in the National Longitudinal Survey of Youth 1979 (NLSY79) is also used.

Study methods. Static and dynamic empirical models are formulated and estimated. Several approaches to estimation of the static model are presented, including ordinary least squares, fixed effects, and random effects using traditional instrumental variables estimators and the Hausman and Taylor (1981) estimator. Consistent estimation of the dynamic panel data models involves first-differencing combined with instrumental variables methods. Results are shown from the estimation of a mental health production function for children using dynamic panel data modeling techniques including the Arellano and Bond (1991) first-differenced GMM estimator and the Blundell and Bond (1998) system GMM approach.

Results. Treatment variables are found to be endogenous in the mental health production function. Some dynamics are found to be significant and important to estimation. Evidence of treatment effectiveness of community mental health treatments is found in dynamic panel data estimation of antisocial problems. Results indicate that combining counseling or psychotherapy with psychopharmacological treatment may be effective, especially for children with mean antisocial scores in the normal range. Psychopharmacological treatment without counseling or psychotherapy appears to be an effective treatment for children with mean antisocial scores in the dysfunctional range.

Conclusion. Dynamic panel data estimation techniques provide a useful new technique for estimating the effectiveness of children's mental health treatments in a community setting. Evidence is found that dynamics in the model are significant. Using the dynamic model with first-differenced GMM estimators, evidence is found that community mental health treatments are effective in reducing antisocial behaviors among youths.