Working Paper: NBER ID: w12377
Authors: David S. Salkever; Stephen Johnston; Mustafa C. Karakus; Nicholas Ialongo; Eric Slade
Abstract: Statistical risk factor models are often proposed for screening high-risk children to participate in early intervention programs. Recent contributions to the program evaluation literature demonstrate the need for incorporating judgments about relative importance of false positives versus false negatives in screening. This paper formalizes these judgments as commensurable economic costs and benefits and applies them to demonstrate an approach to participant selection motivated by the standard cost-benefit criterion of maximizing expected net benefits. Implications of this approach are explored using data from a mental health prevention trial. We illustrate the response of expected net benefits to the choice of a selection risk level, the sensitivity of the optimal selection risk level to per participant cost/benefit magnitudes, and the use of the target-efficiency approach for choosing among alternative risk-factor models. Several strategies that directly incorporate expected net benefit maximization as a criterion in the model estimation process are also examined.
Keywords: child mental health; program evaluation; risk factor models; intervention selection; cost-benefit analysis
JEL Codes: I12; D61
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| different selection risk levels (D81) | expected net benefits (H43) |
| optimal selection risk (D81) | expected net benefits (H43) |
| high selection risk (D81) | low intervention costs (D23) |
| high selection risk (D81) | treats fewer positive cases (C22) |
| low selection risk (G11) | treats more positives (C32) |
| low selection risk (G11) | higher costs (J32) |
| optimal selection risk (D81) | sensitivity to costs and benefits (D91) |
| risk factor models (C52) | varying expected net benefits (D81) |
| target-efficiency approach (D61) | maximizes expected net benefits (D61) |